Every prediction on the Index — all 329 on one page, newest first. Prefer filters and countdowns? Use the Index. The raw data is also available as JSON.
PendingEnergy & Data CentersSaid Sep 18, 2026Deadline Sep 18, 2029
“China will be "completely self-sufficient" in chips within two to three years. Chinese chipmakers' products might not be as good as Nvidia's, but they are used to competing on lower margins — and US export controls create a powerful incentive for investment in Chinese innovation. "The export controls do not work."”
Verdict: Resolves by September 2029 on whether China's semiconductor supply (including advanced logic and the tools to make it) is effectively domestically sourced; as of the prediction date China still depended heavily on foreign EUV lithography and HBM.
PendingJobs & EconomySaid Sep 17, 2026Deadline Dec 31, 2027
“I expect Nvidia to sell twice as many chips as this next year as we do this year.”
Verdict: Huang attributed the forecast to AI adoption spreading across industries; resolution depends on Nvidia's reported chip unit volumes for the coming year versus 2026.
“Given the accelerating rate of AI capability development, it's my worry that in 6-12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage).”
Verdict: Written after an incident in which autonomous OpenAI agents reached the internet without permission and conducted cyberattacks on Hugging Face; the outer 12-month bound falls in September 2027.
PendingJobs & EconomySaid Sep 12, 2026Deadline Dec 31, 2026
“OpenAI will not go public in 2026: "We're not rushing into an IPO. I actually think that, given everything happening with safety, right now would be an ill-advised moment to go public, and we don't feel pressure on that... I would say not 2026."”
Verdict: OpenAI filed confidentially for an IPO in June 2026; Altman pointed to safety and alignment work as the reason for delay and declined to commit to 2027. Resolves at year-end 2026.
PendingEnergy & Data CentersSaid Sep 9, 2026Deadline Dec 31, 2026
“Electricity consumption reaches record levels in our forecast, driven by data center development and increased manufacturing activity in the commercial and industrial sectors: we expect electricity sales in the United States to total 4,135 billion kilowatthours in 2026, an increase of almost 2% over 2025, and 4,211 billion kilowatthours in 2027.”
Verdict: The 2026 calendar year has not closed; EIA also expects record total generation of 4,368 billion kWh in 2026. Distinct from the December 2025 STEO entry, which forecast total power demand rather than retail electricity sales.
PendingJobs & EconomySaid Sep 2026Deadline Dec 31, 2026
“AI lab revenue is "the single most important data point in the market today": the leading labs — OpenAI, Anthropic and xAI — must lift their combined annualized revenue from roughly $100 billion to at least $180 billion by year-end to keep the AI trade intact.”
Verdict: Gerstner based the ~$100 billion combined run-rate estimate on figures circulating in July 2026; resolution depends on the labs' reported annualized revenue at the end of 2026.
PendingEnergy & Data CentersSaid Aug 26, 2026Deadline Jan 31, 2028
“Nvidia will grow revenue about 70% in fiscal 2028 (the year ending January 2028) — roughly $673 billion, versus the ~44% analysts modeled: "Even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%."”
Verdict: Resolves when Nvidia reports fiscal 2028 results in early 2028; the figure is explicitly supply-constrained, with Huang saying unconstrained demand would be "a lot higher."
PendingEnergy & Data CentersSaid Aug 24, 2026Deadline Dec 31, 2027
“SPARC is now 'about 80% complete,' and the company expects to show net energy gain in 2027.”
Verdict: CFS has raised roughly $4 billion to date; Mumgaard told Canary Media in July 2026 that SPARC should come together in early-to-mid 2027, at which point he expects to demonstrate net energy gain.
PendingJobs & EconomySaid Aug 19, 2026Deadline Dec 31, 2030
“If we could take two, three, four points of our cost down from making smarter, better decisions over a series of several years — my gosh — you look at a margin at Delta, you go from a 10% margin to a 15% margin, it's a 50% improvement in your profitability. These things are billions of dollars substantial... I never refer to AI as artificial. I always call it augmented intelligence. It's going to make our people smarter and better.”
Verdict: Bastian framed the gain as coming 'over a series of several years,' scored here against an end-of-2030 horizon; Delta reported an 8.8% adjusted operating margin in Q2 2026 and guided to 11-13% for Q3 2026.
PendingEnergy & Data CentersSaid Aug 10, 2026
“As Jensen said, each gigawatt costs $50 to $60 billion to build out, and we're talking about in the United States alone, we're going to need over 70 gigawatts of power to fuel this.”
Verdict: Fink gave no completion date for the 70+ GW of US AI power capacity, so the claim is tracked as an open capacity/cost estimate; it was made alongside Nvidia's MOUs with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilize roughly $500 billion of compute financing.
PendingEnergy & Data CentersSaid Jul 30, 2026Deadline Dec 31, 2026
“We now expect to spend approximately $220 billion in capital expenditures in 2026 — higher memory costs pushing the number up from our prior estimate of about $200 billion — and we still will not have enough capacity to meet demand.”
Verdict: Supersedes the February 2026 ~$200 billion guide; AWS backlog was reported at $496 billion in the same quarter. Resolves on Amazon's full-year 2026 capex.
PendingJobs & EconomySaid Jul 30, 2026Deadline Dec 31, 2029
“"I think that in 2028 or 2029, you'll see it [quantum computing] have a measurable impact on our top line and bottom line."”
Verdict: Made the same week IBM and Algorithmiq announced research they billed as a demonstration of quantum advantage; IBM does not yet break out material quantum revenue.
PendingJobs & EconomySaid Jul 30, 2026Deadline Dec 31, 2039
“"By the end of the 2030s, we are now pretty convinced this is a trillion dollars of value" — quantum computing will represent a trillion dollars of economic value by the end of the 2030s.”
Verdict: Krishna tied the figure to quantum applications in batteries, materials, fusion energy and drug discovery; the 'value' metric is not precisely defined, which will complicate eventual scoring.
PendingEnergy & Data CentersSaid Jul 22, 2026Deadline Dec 31, 2026
“We are updating our full year 2026 CapEx guidance range to $195-205 billion, up from our previous estimate of $180-190 billion. The increase in the range is primarily due to an acceleration in the delivery of capacity to meet growing demand.”
Verdict: Third 2026 capex guide of the year (after $175-185B in February and $180-190B in Q1); Q2 2026 capex alone was $44.9 billion. Resolves on full-year 2026 results.
PendingEnergy & Data CentersSaid Jul 22, 2026Deadline Dec 31, 2026
“We now expect to have at least 125 GW of gas equipment under contract by year-end 2026.”
Verdict: Gas power backlog plus slot reservations stood at 116 GW at the end of Q2 2026 (up from 100 GW in Q1), and the target was raised from the 110 GW guided in April; analysts have suggested the market expects 130–140 GW.
PendingEnergy & Data CentersSaid Jul 22, 2026Deadline Sep 30, 2026
“We remain on track to deliver 20 GW of annual gas turbine output in the third quarter of 2026.”
Verdict: Q3 2026 ends September 30, 2026; the run-rate milestone will be verifiable with GE Vernova's Q3 2026 results in late October.
PendingEnergy & Data CentersSaid Jul 22, 2026Deadline Dec 31, 2030
“To meet this demand... we are targeting 24 GW of annual gas turbine output in 2028, and we are implementing actions to produce 30 GW in 2030 — largely within the existing factory footprint.”
Verdict: The 30 GW-by-2030 capacity plan is a step up from the previously communicated 24 GW ceiling; at least one sell-side analyst has said 'the jury's out' on whether it can be executed capital-efficiently.
PendingHealthcareSaid Jul 16, 2026Deadline Dec 31, 2026
“In the quarter, we committed to eliminating, by the end of this year [2026], 30% of prior authorization volume and nearly two-thirds of prior authorization requirements for pediatric care... AI is both an enabler and accelerant to this effort.”
Verdict: Builds on UnitedHealthcare's May 5, 2026 announcement to eliminate an additional 30% of remaining prior authorizations and its May 29, 2026 pledge to remove nearly two-thirds of pediatric prior authorization requirements by year-end; first tranche of reductions took effect October 1, 2026.
PendingJobs & EconomySaid Jul 14, 2026Deadline Dec 31, 2040
“By around 2040 there will be roughly one billion AI-powered humanoid robots in operation and 100 trillion AI agents at work, with the AI-related market accounting for about 20% of combined global GDP (about 7 quadrillion yen); the center of labor will shift from humans to robots within the next decade or so.”
Verdict: As of September 2026, the New York Times reported Anthropic is on track to exceed a $100 billion annualized revenue run rate by the end of 2026 (up from $65 billion in July) — but annualized run rate is not the same as $100 billion of revenue booked during the year, so resolution will depend on which measure is applied.
PendingJobs & EconomySaid Jun 27, 2026Deadline Dec 31, 2028
“So what's the '28 number? Is it $200 billion? Is it $300 billion? ... It will be very profitable at that scale because it'll be inference-dominated and people are reporting they have 85% [gross margins on inference].”
Verdict: Target date is more than two years out; Anthropic was reported at roughly a $65 billion annualized run rate in July 2026 and on track for over $100 billion by year-end 2026.
PendingJobs & EconomySaid Jun 15, 2026Deadline Jun 15, 2027
“It might not be inconceivable that Micron, the supplier of HBMs, might be more valuable than Meta in the next 6 to 12 months. Whatever is the bottleneck will command the price.”
Verdict: Micron briefly overtook Meta's market cap on June 25, 2026 after a blowout fiscal Q3 (revenue up 346% year over year), but the crossover did not hold: as of September 2026 Meta sits near $1.6–1.7T versus roughly $1.0–1.1T for Micron. The 6–12 month window runs to mid-June 2027.
PendingEnergy & Data CentersSaid Jun 4, 2026Deadline Dec 31, 2029
“Helion is best positioned to generate electricity from fusion for customers this decade, not the next, and we have the right technology and strategy to build the commercial fusion market over the long term.”
Verdict: Helion's Orion plant, contracted to supply Microsoft starting in 2028, is the vehicle for delivering fusion electricity to customers before 2030.
PendingEnergy & Data CentersSaid Jun 2026Deadline Dec 31, 2030
“This report updates the 2024 Data Center Energy Usage Report and estimates that data centers could account for 11.8% of total U.S. electricity by 2030 (reference case 649 TWh), with a scenario range of 9.5% to 15.3% of total U.S. electricity use by 2030 — compared with the 2024 report's 6.7%–12.0% by 2028 estimate.”
Verdict: Target year 2030 is in the future. Note the report is titled '2025 Update' but is self-cited as Shehabi 2026 and references the NERC 2025 Long-Term Reliability Assessment published in 2026.
PendingEnergy & Data CentersSaid May 20, 2026Deadline Dec 31, 2028
“On power, the near-term shortage starts to ease in 2027 and 2028 as new sources of energy come online — and orbital compute solves it in the long term.”
Verdict: Deadline is end of 2028; resolution will turn on whether US data-center power constraints measurably loosen in 2027-28 as new generation comes online.
PendingJobs & EconomySaid May 7, 2026Deadline May 7, 2028
“The AI-fueled bull market has another year or two to run — it feels like 1999, about a year before the dot-com peak: "Just imagine the stock market went up another 40%. The stock market GDP is going to probably be good lord 300%, 350%. You just know that there'll be some breathtaking kind of corrections."”
Verdict: Jones said he had recently added to AI positions and put the current cycle roughly 50-60% through a typical four- to five-and-a-half-year productivity boom; judge by whether the AI-led equity rally peaks before roughly May 2028.
PendingEnergy & Data CentersSaid May 5, 2026Deadline Dec 31, 2030
“Following 7 gigawatts of new load agreements signed during the first quarter, primarily in Ohio and Texas, AEP's incremental load is expected to grow to 63 GW by 2030 — up from 56 GW a quarter earlier, nearly 90% of it data centers — backed by signed agreements with well-capitalized industrial customers, hyperscalers and data centers.”
Verdict: AEP Texas accounts for 41 GW of the commitments; AEP has since expanded the figure to 69 GW through 2030. Resolves at end-2030.
PendingEnergy & Data CentersSaid May 5, 2026Deadline Dec 31, 2036
“The capital expenditures for this [the AI infrastructure buildout] — $10 trillion. That's just for the US. Over the next 10 years we will be rewiring the global economy... laying cloud AI factories and data centers. Governments' deficits mean much of this money is going to come from the private sector — pension funds, 401(k)s, insurance companies and private credit.”
Verdict: Ten-year horizon running to 2036; Fink made the same $10 trillion / private-savings-financing argument the following day (May 6, 2026) at Texas State Technical College in Waco, a clip of which Snopes fact-checked as genuine.
PendingJobs & EconomySaid May 4, 2026Deadline Dec 31, 2029
“I think Nvidia will be the first $10 trillion company. Whether it comes all at once or whether it comes over the course of the next two to three years, we're patient — we're going to own this, it's one of our largest holdings.”
Verdict: Nvidia's market capitalization was roughly $4.8 trillion when the prediction was made; the implied two-to-three-year window runs to about 2029.
PendingJobs & EconomySaid May 2026Deadline Dec 31, 2026
“Inflation surprises lower (not higher), the dollar rallies (not crashes), the rolling recession ends, AI drives structural deflation the bond market is only starting to price, and a 30+ year base in US capital spending just broke to the upside — with official CPI readings expected to 'surprise on the low side' over the next six to nine months.”
Verdict: Judged on end-2026 US CPI prints versus consensus at the time of the call, the trade-weighted dollar, and whether the housing/manufacturing 'rolling recession' ends; Wood framed the inflation leg as a six-to-nine-month call.
PendingEnergy & Data CentersSaid Apr 30, 2026Deadline Dec 31, 2030
“As a result of these trends, I'm excited to announce that we are increasing our large reciprocating engine capacity from 2x 2024 levels to nearly 3x 2024 levels — and our target for power generation sales is now more than 3x by 2030 from a 2024 baseline.”
Verdict: Caterpillar says it will invest heavily in the capacity expansion from 2027 through 2029 and expects positive cash payback by the end of 2030; power generation retail sales grew 72% year over year in Q2 2026.
PendingJobs & EconomySaid Apr 30, 2026Deadline Dec 31, 2030
“We expect a compound annual growth rate for sales and revenues between 6% and 9% from 2024 to 2030, and we now anticipate low double-digit growth for full year 2026 sales and revenues.”
Verdict: The 2024–2030 CAGR target runs to the end of the decade; the nearer-term 2026 component was already raised again at Q2 2026 (August 4, 2026), when Caterpillar guided to mid-to-high-teens full-year sales growth after a record $20.5 billion quarter.
IncorrectCapabilitiesSaid Apr 22, 2026Deadline Aug 31, 2026
“Start of production [for Optimus at Fremont] is, we're assuming, somewhere around the late July, August time frame — though initial output will be 'quite slow' and the production rate this year is 'literally impossible to predict.'”
Verdict: The late-July/August window was missed: Tesla's Q2 2026 update (July 22) said Fremont construction had begun with production only 'anticipated later this year,' Musk said on the call that Tesla 'will soon start production' with Optimus, and August 2026 reporting described the start as pushed to 'later this year.'
PendingCapabilitiesSaid Apr 22, 2026Deadline Dec 31, 2026
“"We strongly believe that our partners will achieve the first examples of quantum advantage this year, leveraging IBM hardware... We continue to make progress in quantum and remain on track to deliver the first large-scale fault-tolerant quantum computer by 2029."”
Verdict: In July–August 2026 IBM and partners including Algorithmiq published demonstrations they described as verified quantum advantage on Heron hardware, but whether these count as genuine advantage remains contested — IBM's own quantum lead has framed quantum advantage as a scientific debate rather than a binary milestone.
PendingSafety & RiskSaid Mar 26, 2026Deadline Dec 31, 2026
“Congress could pass bipartisan artificial intelligence legislation within months — a national framework built on the White House's four-page blueprint pairing child-safety measures with federal preemption of state AI laws: "We've gotten a very good reception from Capitol Hill. This is an area where I think we're willing and happy to work with Democrats."”
Verdict: Made in the same interview in which Sacks disclosed he had exhausted his 130-day special-government-employee limit and was moving to co-chair PCAST. As of late September 2026 no federal AI framework has been enacted: the preemption text released by a bipartisan House group on June 4, 2026 remains a discussion draft rather than an introduced bill, and while the House passed the KIDS Act child-safety package on June 29, 2026, it awaits Senate action.
PendingEnergy & Data CentersSaid Feb 26, 2026Deadline Dec 31, 2030
“Data centers could consume 9%–17% of U.S. electricity generation by 2030 — more than double current use of 4%–5% — with eight states seeing much higher percentages; these projections are about 60% higher than EPRI projected in 2024.”
PendingJobs & EconomySaid Feb 19, 2026Deadline Dec 31, 2027
“We reiterate our expectation of positive Adj. EBITDA for the full fourth quarter of 2026, with 2027 expected to be our first full year of Adj. EBITDA profitability; we first noted these expectations in 2022, underscoring the predictability of our model and a consistent record of execution.”
Verdict: Q4 2026 has not yet been reported; on the Q2 2026 call management reiterated the target (implied Q4 2026 adjusted EBITDA of roughly $8 million) and credited AI-driven claims automation for a record-low 5% LAE ratio versus an industry average of about 9%. Full resolution requires FY2027 results.
PendingAGI & SuperintelligenceSaid Feb 13, 2026Deadline Jun 30, 2028
“If you are a sophomore now, you will graduate into a world with AGI in it.”
Verdict: Resolves when the class of 2028 graduates in June 2028; the Daily reports Altman framed AGI as likely to arrive 'within the next few years.' Note the ceremony date is reported inconsistently by the Daily itself (Friday Feb. 13 vs. 'Saturday evening').
PendingAGI & SuperintelligenceSaid Feb 12, 2026Deadline Dec 31, 2030
“ARC-4 is in the works, to be released early 2027. ARC-5 is also planned. The final ARC will probably be 6-7. The point is to keep making benchmarks until it is no longer possible to propose something that humans can do and AI can't. AGI ~2030.”
Verdict: Notable because Chollet is widely cast as an AGI skeptic; the post drew attention (including from OpenAI's Noam Brown) for putting his AGI estimate within about five years.
PendingEnergy & Data CentersSaid Feb 5, 2026Deadline Dec 31, 2026
“With such strong demand for our existing offerings and seminal opportunities like AI, chips, robotics, low earth orbit satellites, we expect to invest about $200 billion in capital expenditures across Amazon in 2026, and anticipate strong long-term return on invested capital.”
Verdict: Context: Amazon's 2025 capex was roughly $131 billion; the $200 billion plan was already raised to about $220 billion in July 2026, so the figure will likely be exceeded, but it resolves on full-year 2026 results.
PendingEnergy & Data CentersSaid Feb 4, 2026Deadline Dec 31, 2026
“For the full year 2026, we expect CapEx to be in the range of $175 billion to $185 billion, with investments ramping over the course of the year. We're investing in AI compute capacity to support frontier model development by Google DeepMind, and to meet Cloud customer demand.”
Verdict: Context: Alphabet spent $91.4 billion in 2025; it has already raised 2026 guidance twice since this forecast — to $180-190 billion and then $195-205 billion in July 2026 — so the original range looks likely to be exceeded, but it resolves on full-year 2026 results.
PendingJobs & EconomySaid Feb 2026Deadline Dec 31, 2030
“Klarna's headcount will fall from about 3,000 to under 2,000 by 2030 as AI scales across operations: "I want to be honest about the fact that I do think there's going to be a very big shift... it's going to be a lot of turmoil in this."”
Verdict: Resolves on Klarna's reported headcount at end-2030; the company has already gone from roughly 7,000 to about 3,000 employees through cuts and unreplaced attrition.
PendingSelf-DrivingSaid Jan 28, 2026Deadline Dec 31, 2026
“We expect to have fully autonomous vehicles in probably somewhere between a quarter and half of the United States by the end of the year, pending regulatory approval... even if it is city-by-city, we expect to be in dozens of major cities by the end of the year.”
Verdict: As of mid-2026 Tesla had expanded robotaxi service to a handful of additional metros (Dallas, Houston, Phoenix, Miami, Orlando, Tampa, Las Vegas) with fleets still reported in the dozens per city, well short of the year-end target.
PendingEnergy & Data CentersSaid Jan 28, 2026Deadline Dec 31, 2026
“We anticipate 2026 capital expenditures, including principal payments on finance leases, to be in the range of $115-135 billion, with year-over-year growth driven by increased investment to support our Meta Superintelligence Labs efforts and core business.”
Verdict: Benchmark: Meta's total costs and expenses were $117.69 billion in 2025; resolves with full-year 2026 results.
PendingJobs & EconomySaid Jan 28, 2026Deadline Dec 31, 2026
“"Our generative AI book of business now stands at more than $12.5 billion... We enter 2026 with momentum and in a position of strength, giving us confidence in our full-year expectations of more than 5 percent constant currency revenue growth and an increase of about $1 billion in year-over-year free cash flow" (2025 free cash flow was $14.7 billion).”
Verdict: Tracking poorly: after a July 14, 2026 profit warning and Q2 revenue of $17.16 billion (up just 1%), IBM cut its 2026 constant-currency revenue growth target to 4–5% on July 22, 2026, while maintaining the ~$1 billion free cash flow increase.
PendingHealthcareSaid Jan 20, 2026Deadline Dec 31, 2026
“Isomorphic Labs expects to have its first clinical trials by the end of 2026.”
Verdict: This is the revised version of Hassabis's missed end-2025 target; as of September 2026 no Isomorphic Labs trial start has been announced, and the deadline runs to 31 December 2026.
PendingCapabilitiesSaid Jan 20, 2026Deadline Jun 30, 2026
“Netflix plans to offer a wider array of ad formats this year, including enhanced interactive video ads that will roll out globally in the second quarter [of 2026], after testing modular interactive video ads in the U.S. and Canada in late 2025.”
Verdict: Netflix said in November 2025 it was testing modular interactive video ads in the U.S. and Canada with 'encouraging' early results and reaffirmed a global Q2 2026 rollout on the January 2026 earnings call; no independent confirmation that the worldwide rollout was completed within Q2 2026 was found, so the outcome cannot yet be judged.
Partially correctEnergy & Data CentersSaid Jan 17, 2026Deadline Apr 30, 2026
“The Colossus 2 supercomputer for @Grok is now operational. First Gigawatt training cluster in the world. Upgrades to 1.5GW in April.”
Verdict: Colossus 2 is indeed the largest tracked AI data center, but the gigawatt claim was premature and the April upgrade slipped: Epoch AI's satellite analysis in January 2026 found only ~350 MW of cooling capacity and expected 1 GW no earlier than May, and as of September 2026 Epoch still estimated Colossus 2 at about 946 MW of IT power, projecting ~1,531 MW only in Q1 2027.
PendingJobs & EconomySaid Jan 14, 2026Deadline Dec 31, 2026
“After holding within a tight range of about 213,000 employees since the end of 2023, Bank of America's headcount will decline in 2026: the No. 1 expense lever is to 'work the headcount through operational excellence and applications of new technologies, including AI' — every time someone leaves, we evaluate whether the role needs to be replaced.”
Verdict: Tracking toward the prediction so far: on the Q1 2026 call (April 15, 2026) management said headcount was down about 1,070 from year-end 2025 through attrition, but the full-year 2026 figure is not yet known.
PendingJobs & EconomySaid Jan 9, 2026Deadline Dec 31, 2026
“Yes. AI will increase demand for knowledge workers, not decrease it. I would refer you to Aaron Levie's post called Jevons Paradox for Knowledge Workers — as the cost of a resource goes down, the aggregate demand for it actually increases. In the past it's been very expensive to generate code... I think it's going to increase massively now because the cost of generating code is coming down so much.”
Verdict: Offered as Sacks' contrarian belief for 2026, directly opposing the entry-level white-collar displacement thesis (e.g. Amodei's "bloodbath" and Calacanis' own 2026 pick that young white-collar workers are the biggest losers); resolvable against 2026 US professional/knowledge-work employment data at year end.
PendingJobs & EconomySaid Jan 9, 2026Deadline Dec 31, 2026
“[On 2026 US real GDP growth:] I think the lower bound is five. I think the upper bound is six-two [6.2%]... if we print six, the only country in the modern world that we think of as a quasi-peer competitor that has printed six is the Chinese. [Driven by AI-led productivity, deregulation and tax cuts.]”
Verdict: Not resolvable until the full-year 2026 figure is published, but BEA data so far show real GDP growth of 2.1% annualized in Q1 2026 and 1.5% in Q2 2026 — far below the 5-6.2% range.
PendingEnergy & Data CentersSaid Jan 9, 2026Deadline Dec 31, 2026
“I will pick copper. We are still completely underestimating how short we are in terms of the global demand-supply dynamics of a handful of critical elements that we need... the asset that is set up to go absolutely parabolic is copper — it is the most useful, cheap, amenable, conductive material that we have.”
Verdict: Tracking well so far: copper set successive records in 2026, crossing about $14,779 per tonne on the LME in early September 2026 (roughly +45-48% year over year), but the full-year outcome is not yet settled.
PendingEnergy & Data CentersSaid Jan 7, 2026Deadline Dec 31, 2027
“SPARC will nearly be completed by the end of 2026 and will produce its first plasma energy in 2027.”
Verdict: First plasma had earlier been guided to 2026 before slipping to 2027; CFS reiterated the 2027 first-plasma target in February 2026.
PendingSelf-DrivingSaid Jan 7, 2026Deadline Dec 31, 2028
“Ford will introduce eyes-off (Level 3) driving technology in 2028, upgrading BlueCruise from hands-free to eyes-off and debuting it on a roughly $30,000 all-electric vehicle built on Ford's Universal EV platform: "Autonomy shouldn't be a premium feature."”
Verdict: Deadline is end of 2028; the first Universal EV platform vehicle, a midsize electric pickup, is slated for 2027 with eyes-off following in 2028.
PendingEnergy & Data CentersSaid Dec 24, 2025Deadline Dec 24, 2030
“xAI will have more AI compute than everyone else combined in <5 years.”
Verdict: Posted in response to SemiAnalysis compute rankings; days later Musk announced a third Memphis building taking the Colossus site toward 2 GW, while Epoch AI's tracked 86 AI data centers totalled about 13 GW of IT power as of September 2026.
PendingSelf-DrivingSaid Dec 16, 2025Deadline Dec 31, 2027
“Waymo will be "meaningful" to Alphabet's financials as soon as 2027, as the company eyes "aggressive" expansion in the coming year.”
Verdict: Deadline is end of 2027. At the time of the prediction Waymo operated paid service in Austin, the Bay Area, Phoenix, Atlanta and Los Angeles, with ~450,000 weekly paid rides and a planned 2026 expansion to 11 more US metros plus London.
IncorrectSelf-DrivingSaid Dec 9, 2025Deadline Dec 30, 2025
“There will be Tesla Robotaxis operating in Austin with no one in them, not even anyone in the passenger seat, in about three weeks. I think it's pretty much a solved problem, we're just going through validation right now.”
Verdict: Three weeks later Austin robotaxi rides still arrived with a safety monitor up front; Tesla only began offering public rides with no one in the vehicle on January 22, 2026, about six weeks after the promise and after at least four similar missed projections.
PendingEnergy & Data CentersSaid Dec 9, 2025Deadline Dec 31, 2026
“U.S. power demand will climb to a record 4,199 billion kilowatt-hours in 2025 and further increase to 4,267 billion kWh in 2026, up from the 2024 record of 4,110 billion kWh, driven in part by data centers dedicated to artificial intelligence and cryptocurrency.”
Verdict: The 2025 leg looks essentially correct — EIA later reported a second straight record of roughly 4,195 billion kWh for 2025 — while the 2026 figure resolves at year-end; mid-2026 STEO vintages put 2026 demand near 4,268 billion kWh.
CorrectEnergy & Data CentersSaid Dec 9, 2025Deadline Dec 31, 2025
“We expect to reach 80 GW of combined gas turbine slot reservation agreements and backlog by year-end 2025 — a book of business that stretches into 2029.”
Verdict: GE Vernova's Q4 2025 results (reported January 28, 2026) showed gas power equipment backlog and slot reservation agreements rising from 62 GW to 83 GW sequentially — above the 80 GW the company guided to three weeks earlier.
PendingEnergy & Data CentersSaid Dec 9, 2025Deadline Dec 31, 2026
“I expect gas turbine reservations to be sold out through 2030 by the end of 2026.”
Verdict: Tracking ahead so far: on the Q2 2026 call (July 22, 2026) Strazik said most 2030 slots were already sold and that more than half of 2031 slots should be under contract by year-end 2026.
PendingEnergy & Data CentersSaid Dec 8, 2025Deadline Dec 31, 2035
“NextEra will build 15 gigawatts of new power generation for data center hubs by 2035 — a 'fairly conservative' target: 'Quite frankly, based on what we're seeing today, we'll be disappointed if we don't do more' — with potential upside to 30 gigawatts.”
Verdict: Announced alongside partnerships with Google (three gigawatt-scale campuses) and Meta; resolves in 2035.
PendingEnergy & Data CentersSaid Dec 8, 2025Deadline Dec 31, 2032
“The data center hubs will help NextEra meet its target of building four to eight gigawatts of new gas-fired generation by 2032 — and considerably more by 2035 — out of a gas development pipeline that has 'now eclipsed 20 gigawatts.'”
Verdict: Grok 4.20 missed the 3–4 week window by a wide margin: it did not reach users until a limited beta on 17 February 2026, with the full rollout and general availability following in March 2026 — roughly two to three months later than promised. (The claim that it beat GPT-5.1 and Gemini 3 Pro came from third-party Alpha Arena trading-sim reports, not from Musk.)
PendingAGI & SuperintelligenceSaid Dec 1, 2025Deadline Dec 31, 2029
“Recursive self improvement is when it's learning on its own. This is not true today. Today, when you set up one of these huge data centers, you have to tell it to learn. But the belief is that this is coming — within four years.”
Verdict: Deadline is end of 2029. The Crimson reports the timeline as four years; other write-ups of the same talk render it as 'two to four years,' so four years is the generous reading. Schmidt added that AI generating medical conjectures and discovering new facts 'looks like it's very close.'
PendingEnergy & Data CentersSaid Dec 1, 2025Deadline Dec 31, 2032
“Filling a one-gigawatt AI data center costs about $80 billion today, so the roughly 100 GW being announced implies about $8 trillion of capital — requiring roughly $800 billion of annual profit just to service it, and on a five-to-seven-year payback that means $1–2 trillion of incremental annual revenue: "That much incremental revenue I don't believe is there... It's my view that there's no way you're going to get a return on that" (compounded by AI accelerators that must be used within about five years before being replaced).”
Verdict: Resolves against the five-to-seven-year payback window Krishna invoked (i.e. by roughly 2032): the test is whether the AI buildout produces the $1–2 trillion of incremental annual revenue his arithmetic requires. Krishna simultaneously argued AI is not a bubble and that generative AI will unlock trillions of dollars of enterprise productivity.
CorrectCapabilitiesSaid Nov 24, 2025Deadline Jan 23, 2026
“Executive Order "Launching the Genesis Mission": "Within 60 days of the date of this order, the Secretary [of Energy] shall identify and submit to the APST a detailed list of at least 20 science and technology challenges of national importance that the Secretary assesses to have potential to be addressed through the Mission."”
Verdict: DOE identified 26 Genesis Mission National Science and Technology Challenges — exceeding the 20 minimum — which were published on February 12, 2026 following the APST's 30-day review window, timing consistent with an on-time submission (the exact internal submission date to the APST was not made public).
PendingCapabilitiesSaid Nov 24, 2025Deadline Aug 21, 2026
“"Within 270 days of the date of this order, the Secretary shall... seek to demonstrate an initial operating capability of the [American Science and Security] Platform for at least one of the national science and technology challenges" — i.e. a working Genesis Mission AI science platform by August 21, 2026.”
Verdict: The deadline has passed but the outcome is not yet publicly adjudicable: DOE contracted Accenture Federal Services in April 2026 for an 'early operating capability' sprint on the critical-minerals (CM2US) challenge and demonstrated a platform build at the July 22, 2026 Genesis Mission Summit, but DOE has not publicly confirmed that formal initial operating capability was demonstrated by August 21, 2026, and independent coverage of the summit called the mission 'not yet measurable.'
PendingCapabilitiesSaid Nov 19, 2025Deadline Jun 30, 2026
“Gemini 3 shows that scaling laws for pretraining are intact. This is the most important AI datapoint since the release of o1. This means that Blackwell models are likely to show a significant increase in performance when they come out in the second quarter of 2026. GPT-5 was not evidence of a slowdown in scaling laws.”
Verdict: Blackwell-trained frontier models in fact shipped earlier than Q2 2026 — OpenAI's GPT-5.3 Codex (February 2026) was trained and served entirely on GB200 NVL72, and xAI's Grok 4.5/4.6 were built on GB200/GB300 — but whether they represent the 'significant increase in performance' Baker predicted is a judgment call that needs a benchmark-based resolution.
PendingJobs & EconomySaid Nov 10, 2025Deadline Dec 31, 2028
“Understating depreciation by extending useful life of assets artificially boosts earnings — one of the more common frauds of the modern era... Yet this is exactly what all the hyperscalers have done. By my estimates they will understate depreciation by $176 billion 2026-2028. By 2028, ORCL will overstate earnings 26.9%, META by 20.8%.”
Verdict: Resolves against hyperscaler depreciation schedules and reported 2026-2028 earnings, in particular whether Oracle's and Meta's 2028 earnings prove overstated by roughly the quoted percentages under shorter server useful lives.
CorrectJobs & EconomySaid Nov 6, 2025Deadline Dec 31, 2025
“We expect to end this year above $20 billion in annualized revenue run rate.”
Verdict: OpenAI CFO Sarah Friar subsequently said the company ended 2025 with more than $20 billion in annualized revenue, and August 2026 Bloomberg reporting on a $40B+ run rate described it as roughly double the end-2025 figure. Numbers are company-reported rather than audited.
PendingJobs & EconomySaid Nov 6, 2025Deadline Dec 31, 2030
“We expect to... grow to hundreds of billions [in annualized revenue] by 2030. We are looking at commitments of about $1.4 trillion over the next 8 years.”
Verdict: Open until end of 2030; OpenAI's run rate was reported above $40 billion in August 2026, so the trajectory is early but the target requires roughly another order of magnitude.
PendingJobs & EconomySaid Nov 6, 2025Deadline Dec 31, 2065
“It's going to affect every application, every job, every customer interface. My guess is the developed world will be working three and a half days a week in 20, 30, 40 years, and have wonderful lives.”
Verdict: Dimon gave a 20-to-40-year range, so the entry is graded at the far end (2065) on average hours worked in developed economies; a ~3.5-day norm would be roughly a 30% cut from today's full-time week.
IncorrectSelf-DrivingSaid Oct 31, 2025Deadline Dec 31, 2025
“We're scaling up the number of cars — probably we'll have a thousand cars or more in the Bay Area by the end of this year, probably 500 or more in the greater Austin area.”
Verdict: By late November Musk had already cut the Austin target to roughly 60 vehicles, and independent tracking at year-end found only about 30-42 robotaxis operating in Austin; the Bay Area fleet also fell far short of 1,000 and still used safety drivers.
CorrectCapabilitiesSaid Oct 29, 2025Deadline Dec 31, 2025
“Looking forward to the Gemini 3 release later this year!”
Verdict: Google released Gemini 3 (Gemini 3 Pro, plus the Antigravity coding platform) on November 18, 2025, comfortably within the calendar year; Deep Think and Flash variants followed in December.
CorrectEnergy & Data CentersSaid Oct 29, 2025Deadline Dec 31, 2025
“We currently expect 2025 capital expenditures, including principal payments on finance leases, to be in the range of $70-72 billion, increased from our prior outlook of $66-72 billion (and up from the $60-65 billion guided in January 2025).”
Verdict: Meta reported full-year 2025 capital expenditures, including principal payments on finance leases, of $72.22 billion — essentially on target, a rounding margin above the top of the $70-72 billion range.
PendingEnergy & Data CentersSaid Oct 29, 2025Deadline Dec 31, 2026
“Our current expectation is that capex dollar growth will be notably larger in 2026 than 2025. We also anticipate total expenses will grow at a significantly faster percentage rate in 2026 than 2025, with growth driven primarily by infrastructure costs.”
Verdict: Benchmark: 2025 capex grew about $33 billion (from $39.23B to $72.22B); Meta's January 2026 guidance of $115-135 billion implies growth of $43-63 billion, so the claim looks on track but resolves only with full-year 2026 results.
CorrectCapabilitiesSaid Oct 28, 2025Deadline Sep 30, 2026
“We think it is plausible that by September of next year [2026] we have an intern-level AI research assistant.”
Verdict: On 6 September 2026 OpenAI said it had hit the 'automated research intern' goal — a system that carries out well-defined research tasks under human direction, including ones taking a skilled researcher days — and reported roughly 3.1 agent-workdays per human workday internally. The milestone is OpenAI's own self-assessment and has not been independently audited.
PendingAGI & SuperintelligenceSaid Oct 28, 2025Deadline Mar 31, 2028
“By March 2028 we have a legitimate AI researcher — a system capable of autonomously delivering on large research projects.”
Verdict: Still open; OpenAI said in September 2026 it was making 'strong progress' toward the March 2028 target while conceding it does not yet know how to safely reach full recursive self-improvement.
PendingJobs & EconomySaid Oct 28, 2025Deadline Dec 31, 2026
“'This is how much business is on the books. Half a trillion dollars worth so far' — cumulative Blackwell and Rubin revenue and bookings, including networking, across calendar 2025 and 2026.”
Verdict: Analysts read the figure as implying calendar-2026 data center sales well above consensus. At GTC in March 2026 Huang raised the number, saying NVIDIA now sees roughly $1 trillion of AI chip demand through 2027; final resolution needs full calendar-2026 results.
PendingSelf-DrivingSaid Oct 27, 2025Deadline Dec 31, 2026
“By the end of 2026, you should expect us to be offering 1 million trips per week.”
Verdict: Waymo was reported at roughly 450,000 paid driverless trips per week in January 2026 — about an 80% increase over six months — while expanding to additional US cities, so reaching 1 million weekly trips by December 2026 remains open.
PendingCapabilitiesSaid Oct 22, 2025Deadline Dec 31, 2026
“We're going to be building a million-unit Optimus production line, hopefully with the production start towards the end of next year [2026].”
Verdict: Deadline is end of 2026. Tesla decommissioned the Fremont Model S/X lines in May 2026 and began converting them for a line designed for 1 million robots/year, but start of production had slipped from 'late July or August' to 'later this year' as of the Q2 2026 update.
IncorrectCapabilitiesSaid Oct 22, 2025Deadline Mar 31, 2026
“We'll actually have a production-intent [Optimus V3] prototype ready to show off in Q1, probably February or March... It won't even seem like a robot. It will seem like a person in a robot suit.”
Verdict: No V3/Gen 3 prototype was shown in Q1 2026. Musk confirmed the delay on March 31, 2026, then on the April 22, 2026 Q1 call pushed the unveil to 'probably middle of this year' (later, closer to production start), citing polish and copycat concerns; it remained unrevealed as of September 2026.
PendingSelf-DrivingSaid Oct 22, 2025Deadline Dec 31, 2028
“GM will bring eyes-off driving to market in 2028, debuting on the Cadillac ESCALADE IQ electric SUV, beginning with highways before expanding to other models and environments.”
Verdict: GE Vernova ended 2025 with 83 GW of gas power equipment backlog plus slot reservation agreements (up from 62 GW at the end of Q3), comfortably exceeding the ~70 GW guided in October.
PendingEnergy & Data CentersSaid Oct 22, 2025Deadline Dec 31, 2030
“In North America alone, we're seeing significant investment in Electrification, where we expect our combined serviceable addressable market to grow at approximately 10% compounded growth rate, doubling in size by 2030 — driven by expanded electrification needs, data center growth and rising digitization.”
Verdict: GE Vernova's electrification equipment backlog has continued to climb (above $40 billion by Q2 2026), but the 2030 market-doubling claim cannot be judged until the end of the decade.
PendingJobs & EconomySaid Oct 21, 2025Deadline Dec 31, 2027
“Amazon's robotics and automation team projects that automation will let the company avoid hiring more than 160,000 U.S. workers it would otherwise need by 2027, saving about 30 cents on each item Amazon picks, packs and delivers (an estimated $12.6 billion in savings from 2025 through 2027).”
Verdict: Amazon disputes the framing, saying leaked documents give an incomplete picture and that it is still hiring at operations facilities; resolution will be judged against Amazon's U.S. operations headcount trajectory and disclosed cost-per-unit savings through 2027, with the avoided-hiring figure inherently counterfactual.
PendingJobs & EconomySaid Oct 21, 2025Deadline Dec 31, 2033
“Amazon executives told the board that robotic automation — with the robotics team's ultimate goal of automating 75% of Amazon's operations — could keep the company's U.S. headcount essentially flat through 2033 even as it sells twice as many products, meaning more than 600,000 U.S. jobs it would otherwise have had to fill will not be needed.”
Verdict: Amazon's U.S. workforce was about 1.2 million when the documents leaked; the company called the leak 'incomplete and misleading.' Resolvable by 2033 against Amazon's reported U.S. headcount versus unit volume growth.
PendingAGI & SuperintelligenceSaid Oct 18, 2025Deadline Dec 31, 2026
“Grok 5 will be AGI or something indistinguishable from AGI.”
Verdict: Target date is an editorial inference from xAI's own Grok 5 schedule (training start September 2025, release originally signalled for end-2025/early-2026). As of September 2026 Grok 5 has still not been released — it slipped from Q1 to Q2 2026 and remains in training — so the claim cannot yet be judged.
PendingAGI & SuperintelligenceSaid Oct 18, 2025Deadline Dec 31, 2035
“AGI is still a decade away. My AI timelines are about 5-10X pessimistic w.r.t. what you'll find in your neighborhood SF AI house party or on your twitter timeline, but still quite optimistic w.r.t. a rising tide of AI deniers and skeptics.”
Verdict: Horizon runs to roughly 2035; Karpathy's bar is an AI you can delegate to like an employee.
PendingEnergy & Data CentersSaid Oct 17, 2025Deadline Oct 17, 2026
“OpenAI needs $400 billion in the next 12 months to pay for the compute and data centers it has publicly committed to — and to complete these deals by the end of 2026, OpenAI needs large swaths of this money by February 2026.”
Verdict: The 12-month window closes in October 2026. As of mid-2026 OpenAI had closed a March 2026 round at an $852 billion post-money valuation and was reported to have received roughly $50 billion of a $122 billion round — far below $400 billion — while raising its planned compute spending to a reported $750 billion through 2030.
PendingJobs & EconomySaid Oct 8, 2025Deadline Oct 8, 2027
“I'm far more worried than others about a serious market correction — if the market is pricing in 10%, I'd say it's more like 30%; it could be six months, could be two years. AI is real and in total will pay off, just like cars and TV paid off, but most people involved in it didn't do well, and a lot of the money going in now will probably be lost.”
Verdict: The two-year window runs to October 2027. As of September 2026 the S&P 500 sits near record highs (~7,600) after an ~8% drawdown in late February–March 2026 that did not reach the conventional 10% correction threshold.
PendingCapabilitiesSaid Oct 6, 2025Deadline Dec 31, 2026
“The xAI game studio will release a great AI-generated game before the end of next year [2026].”
Verdict: Deadline is December 31, 2026; xAI's game studio, announced in late 2024 and staffing up through 2025, had not shipped a title as of late 2026.
PendingCapabilitiesSaid Oct 6, 2025Deadline Dec 31, 2026
“Grok will make a movie that is at least watchable before the end of next year [2026] and really good movies in 2027.”
Verdict: First deadline is December 31, 2026 for an 'at least watchable' Grok-made movie, with a follow-on 2027 target for 'really good movies'; as of late 2026 xAI had released Grok Imagine video generation but no Grok-produced film.
PendingSafety & RiskSaid Oct 2025Deadline Dec 31, 2035
“Hyperintelligent machines could pose an existential threat to humanity within the next decade: AI systems trained on human language and behaviour could develop their own 'preservation goals', making them in effect competitors to the species that created them.”
Verdict: Bengio framed major AI risks as arriving in roughly five to 10 years while cautioning they could come sooner; he launched the nonprofit LawZero in June 2025 with $30 million to build 'non-agentic' safe-by-default systems.
PendingJobs & EconomySaid Sep 29, 2025Deadline Dec 31, 2030
“Reduction of 4,000 administrative jobs by 2030 through digitalization, automation, and process consolidation: "In particular, the profound changes brought about by digitalization and the increased use of artificial intelligence will lead to greater efficiency in many areas and processes... the Lufthansa Group plans to cut a total of around 4,000 jobs worldwide by 2030, the majority of which will be in Germany." The focus will be on administrative rather than operational roles.”
Verdict: Announced at Lufthansa Group's first company-wide Capital Markets Day in six years, alongside a raised mid-term adjusted EBIT margin target of 8-10% from 2028 and plans for 230+ new aircraft by 2030; Spohr told a staff town hall days earlier that becoming "leaner in admin" meant roughly 20% fewer people in group administration. Resolution depends on group administrative headcount reported through 2030.
“"It's very clear that AI is going to change literally every job. Maybe there's a job in the world that AI won't change, but I haven't thought of it." Walmart plans to keep its global headcount roughly flat at about 2.1 million over the next three years, even as revenue grows, with some jobs and tasks eliminated and new ones created.”
Verdict: Walmart's global headcount was about 2.1 million when the prediction was made; the flat-headcount horizon runs to roughly late 2028 and can be checked against Walmart's annual associate counts.
Partially correctEnergy & Data CentersSaid Sep 23, 2025Deadline Dec 31, 2025
“This puts us on a clear path to securing the full $500 billion, 10-gigawatt commitment we announced in January by the end of 2025, ahead of schedule.”
Verdict: Across all partners OpenAI blew past 10GW on paper — Altman said on October 28, 2025 that the company had roughly $1.4 trillion and ~30GW of infrastructure commitments. But the Stargate vehicle itself did not close it out: The Information reported in February 2026 that OpenAI fell short of locking in 10GW through SoftBank and Oracle by the end of 2025, with OpenAI citing 'well over 8GW' and $400B+ in play as of early 2026.
CorrectEnergy & Data CentersSaid Sep 15, 2025Deadline Jul 4, 2026
“"I'm pretty confident we'll have one of them before July 4th of next year, and several more during the year. It's not the back-end electricity production, but it's the whole nuclear system running, demonstrating how it will work" — the U.S. will have at least one next-generation small reactor operating before July 4, 2026.”
Verdict: Four privately built advanced test reactors reached criticality under DOE authorization by July 4, 2026 (Antares, Valar Atomics, Deployable Energy and Aalo Atomics), matching Wright's framing of a full system demonstration rather than electricity production — though these are kilowatt-scale test units, not operating commercial SMRs.
PendingHealthcareSaid Sep 12, 2025Deadline Dec 31, 2027
“In the next couple of years, I'd like to see that [the roughly ten years and billions of dollars it takes to design one drug] cut down in a matter of months, instead of years. That's what I think is possible. Perhaps even faster.”
Verdict: Resolves against demonstrated design-to-candidate timelines at Isomorphic Labs and peers by end-2027; Isomorphic has not yet started a clinical trial.
PendingEnergy & Data CentersSaid Sep 9, 2025Deadline May 31, 2030
“We expect Oracle Cloud Infrastructure revenue to grow 77% to $18 billion this fiscal year—and then increase to $32 billion, $73 billion, $114 billion, and $144 billion over the subsequent four years.”
Verdict: The FY2026 waypoint looks roughly on track: quarterly IaaS revenue ran $3.3B, $4.1B and $4.9B in Q1–Q3 FY26, with Q1 FY27 IaaS up 121% year-over-year. The $144 billion FY2030 figure resolves in May 2030.
CorrectEnergy & Data CentersSaid Sep 9, 2025Deadline Mar 31, 2026
“Over the next few months, we expect to sign-up several additional multi-billion-dollar customers and RPO is likely to exceed half-a-trillion dollars.”
Verdict: Remaining performance obligations were $455 billion when Catz spoke; one quarter later, on December 10, 2025, Oracle reported Q2 FY26 RPO of $523 billion, up 438% year-over-year — above half a trillion dollars and well inside the stated window. RPO reached $638 billion by Q4 FY26.
PendingEnergy & Data CentersSaid Sep 2025Deadline Dec 31, 2033
“OpenAI's audacious long-term goal is to build 250 gigawatts of capacity by 2033.”
Verdict: For scale: 250GW is roughly a third of US peak power demand and more than India's total capacity; OpenAI was operating on the order of a gigawatt-scale footprint when the memo was written.
PendingJobs & EconomySaid Aug 27, 2025Deadline Dec 31, 2030
“AI infrastructure spending will reach $3 trillion to $4 trillion by the end of the decade: 'As the AI revolution went into full steam... the capex spend has doubled to $600 billion per year. There's five years between now and the end of the decade, and $600 billion only represents the top four hyperscalers.'”
Verdict: Huang reiterated the same $3–4 trillion 2030 forecast at the Goldman Sachs Communacopia + Technology Conference on September 10, 2026, citing a new layer of computing and the end of Moore's Law.
Partially correctEnergy & Data CentersSaid Aug 20, 2025Deadline Dec 31, 2025
“The AI datacenters to be built in 2025 will suffer $40 billion of annual depreciation, while generating somewhere between $15 and $20 billion of revenue.”
Verdict: The capex/depreciation arithmetic held up directionally — 2025 AI capex landed near his ~$400 billion estimate, implying roughly $40 billion of annual depreciation on a 10-year life (and more on the 3-6 year schedules now used) — and pure-play AI revenue for calendar 2025 was in his range (OpenAI reported $13.07 billion of 2025 revenue, Anthropic roughly $9 billion run-rate exiting the year). But industry-wide AI revenue including hyperscaler AI cloud lines was estimated well above $20 billion, so the revenue side of the comparison is disputed.
CorrectCapabilitiesSaid Aug 8, 2025Deadline Oct 8, 2025
“They no longer have anything like a clear technical lead. GPT-5 is unlikely to be ahead of the pack for more than a couple months. (And Grok 4 Heavy is already better on the ARC-AGI-2 measure.)”
Verdict: GPT-5 never held an uncontested lead: Grok 4 already beat it on ARC-AGI-2 at launch, and competing frontier releases within weeks to a few months (Anthropic's Claude 4.5 line in late September and Google's Gemini 3 Pro in November 2025) took the top of major leaderboards, prompting OpenAI's own GPT-5.1 refresh.
PendingCapabilitiesSaid Jul 31, 2025Deadline Dec 31, 2026
“We're making good progress on a more personalized Siri, and as we've said before, we expect to release these features next year [2026].”
Verdict: Deadline is end of 2026. The personal-context/on-screen-awareness Siri features were first promised at WWDC 2024, delayed in March 2025, and were reported to be targeted at a spring 2026 (iOS 26.4) release before slipping toward the WWDC 2026 / iOS 27 cycle.
IncorrectEnergy & Data CentersSaid Jul 30, 2025Deadline Dec 31, 2025
“Helion remains on track to become the first fusion company in the world to generate electricity from its Polaris test reactor later in 2025: 'Early testing has been encouraging, and they expect to demonstrate electricity this year.'”
Verdict: No electricity demonstration was announced by the end of 2025. Polaris' announced milestones in February 2026 were measurable D-T fusion and 150 million °C plasmas, and by September 2026 Helion had softened its language and hedged on when it would sell fusion electricity.
PendingEnergy & Data CentersSaid Jul 29, 2025Deadline Dec 31, 2030
“AEP expects to interconnect 24 gigawatts of new load by 2030 — 18 gigawatts of it data centers, all backed by signed customer agreements — and will raise its five-year capital plan from $54 billion to about $70 billion to serve it: 'Demand for power is growing at a pace I have not seen in my 45-year energy career.'”
Verdict: Tracking: AEP subsequently raised its capital plan to about $72 billion (Oct 2025) and $78 billion, and its contracted incremental load by 2030 to 63 GW (Q1 2026); the load itself resolves at end-2030.
CorrectCapabilitiesSaid Jul 27, 2025Deadline Dec 31, 2025
“GPT-5 by itself won't be a general purpose artificial general intelligence capable of taking on arbitrary tasks. Without external aids it won't be able to drive a car reliably; it won't be able to reliably guide a robot like Optimus to be anything like as versatile as Rosie the Robot. Reasoning about the physical, psychological and mathematical world will still be unreliable, fluent hallucinations will still be common, and 'alignment' will continue to be a critical, unsolved problem.”
Verdict: GPT-5 launched in August 2025 and was not an AGI: it could not drive a car or run a humanoid robot, hallucinations and reasoning failures persisted, and alignment remained open. The rollout was poorly received — Sam Altman conceded OpenAI had 'totally screwed up' the launch — and Marcus's assessment that it 'came late and fell short' was echoed by many observers.
IncorrectSelf-DrivingSaid Jul 23, 2025Deadline Dec 31, 2025
“I think we'll probably have autonomous ride-hailing in probably half the population of the US by the end of the year. That's at least our goal, subject to regulatory approvals.”
Verdict: At the end of 2025 Tesla's robotaxi service ran only in Austin and the Bay Area, almost entirely with human safety monitors on board, covering a negligible share of the US population; commentators put actual driverless coverage at roughly 0% versus the promised 50%.
“[On how long until an AI is as good as one of Phil Tetlock's superforecasters:] A year ago, I said two to three years. Right now, I'm going to say one to two years, which is the same prediction. I think you're way too pessimistic in your timetable.”
Verdict: Deadline is mid-2027. The Forecasting Research Institute's October 2025 ForecastBench update judged Cowen closer to the mark than Silver, with extrapolation implying parity by late 2026; as of the January 2026 leaderboard superforecasters still led the best models by about 0.017 Brier points.
“[Asked whether he wanted to revise his prediction that it would take at least 10 to 15 years for AI to be as good as human superforecasters:] I think it's probably about right. I would say, relative to a year ago, AI is about at the 40th percentile of progress I would've expected.”
Verdict: Long-dated claim (parity no sooner than ~2035-2040). In the earlier 2024 episode Silver's stated over was '15 or 20' years; FRI's October 2025 ForecastBench update judged Cowen's timeline closer, though superforecasters still topped the leaderboard in January 2026.
PendingEnergy & Data CentersSaid Jul 22, 2025Deadline Jul 22, 2030
“The @xAI goal is 50 million in units of H100 equivalent-AI compute (but much better power-efficiency) online within 5 years.”
Verdict: For scale, 50 million H100-equivalents at H100 power draw would imply roughly 35 GW; xAI's largest site, Colossus 2, was estimated by Epoch AI at about 946 MW of IT power and roughly 1.1 million H100-equivalents as of September 2026.
PendingEnergy & Data CentersSaid Jul 20, 2025Deadline Dec 31, 2025
“we will cross well over 1 million GPUs brought online by the end of this year! very proud of the team but now they better get to work figuring out how to 100x that lol”
Verdict: OpenAI does not disclose an audited GPU count, so the milestone cannot yet be independently judged; contemporaneous reporting noted the figure refers to aggregate capacity across partners and sites rather than a single cluster.
Partially correctCapabilitiesSaid Jul 17, 2025Deadline Dec 31, 2025
“We're also going to be introducing interactivity in the second half of the year — a pretty significant milestone for us — on top of Netflix's own ad-tech stack, consistent with our goal to roughly double the ads business this year.”
Verdict: The revenue half was met — Netflix more than doubled advertising revenue in 2025 (about $1.5B in Q4 alone) — but interactivity arrived only in limited form: interactive/modular video ad formats began testing in the U.S. and Canada in November 2025, with the full global rollout deferred to Q2 2026.
IncorrectJobs & EconomySaid Jul 16, 2025Deadline Dec 31, 2025
“SoftBank will deploy one billion AI agents internally this year [2025] — active 24 hours a day, 365 days a year and interacting with each other — with roughly 1,000 AI agents needed to replace each employee, because employees have complex thought processes.”
Verdict: SoftBank came nowhere near a billion: in December 2025 SoftBank Corp. publicised 2.5 million internally created AI agents as a landmark achievement — roughly 0.25% of the target — and Son had already conceded at announcement that the goal was 'challenging' and that the orchestration software did not yet exist.
PendingCapabilitiesSaid Jul 14, 2025Deadline Dec 31, 2026
“We're calling the first one Prometheus and it's coming online in '26. We're also building Hyperion, which will be able to scale up to 5GW over several years. We're building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan.”
Verdict: Prometheus is under construction at Meta's New Albany, Ohio campus (gas turbines plus nuclear supply deals signed with Vistra, TerraPower and Oklo in January 2026); Meta still says it expects the cluster online sometime in 2026, so the claim cannot be resolved before 31 December 2026.
PendingJobs & EconomySaid Jul 10, 2025Deadline Dec 31, 2025
“So today, we're about 3% of domestic. Our goal is to have about 20% by the end of the year... We're letting the machine go ahead and set prices in a very controlled way. This is a full reengineering of how we price and how we will be pricing in the future.”
Verdict: Delta has never publicly confirmed reaching the 20% figure; the plan drew a July 2025 letter from Senators Gallego, Warner and Blumenthal, and press coverage a year later still described AI-set fares on roughly 3% of tickets, so the outcome cannot yet be judged definitively.
PendingEnergy & Data CentersSaid Jul 7, 2025Deadline Dec 31, 2030
“Blackouts could increase by 100 times in 2030 relative to today's averages if the U.S. continues to shutter reliable power plants and fails to add firm capacity: in the 'plant closures' scenario annual loss-of-load hours rise from 8.1 hours per year today to 817 hours — 'more than 800 hours per year' — with 104 GW of firm capacity set to retire by 2030 and an additional 100 GW of new peak-hour supply needed, 50 GW of it attributable to data centers.”
Verdict: 2030 is in the future, and the figure is a scenario-conditional modeling result (the 'plant closures' case); clean energy analysts have disputed the modeling assumptions.
PendingJobs & EconomySaid Jul 2025Deadline Dec 31, 2030
“Almost certainly within the next five years, any economically valuable job humans can do, AI will be able to do 80% of it — with a few exceptions, like heart surgery or brain surgery.”
Verdict: Khosla framed the near term (to 2030) as showing up mainly as productivity gains, with more radical labor displacement after that.
PendingCapabilitiesSaid Jun 27, 2025Deadline Dec 31, 2027
“Three years ago, Copilot was already creating 40% of code in those files where it was enabled, and I predicted that 80% of code would be written by AI in 2027. Three predictions of where we are going next: 1. 90% of code will be written by agents.”
Verdict: Deadline is end of 2027; the 90%-by-agents half of the statement carries no date, so the entry resolves on the 80%-of-all-code-written-by-AI-in-2027 figure.
PendingEnergy & Data CentersSaid Jun 25, 2025Deadline Dec 31, 2027
“"With PJM's action... we're on track to make history ahead of schedule, helping America achieve energy independence, supercharge economic growth, and win the global AI race" — Three Mile Island Unit 1 / the Crane Clean Energy Center will restart as early as 2027, about a year ahead of the original 2028 target.”
Verdict: As of September 2026 Constellation still targets 2027; FERC granted the Eddystone capacity-rights waiver in June 2026, the NRC issued a finding of no significant environmental impact, and a $1 billion DOE loan closed in November 2025.
CorrectEnergy & Data CentersSaid Jun 20, 2025Deadline Jul 4, 2026
“The Energy Department's new Reactor Pilot Program will expedite testing of advanced nuclear reactor designs under DOE authority, outside the national laboratories and without NRC licensing, "ensuring at least three reactors achieve criticality by July 4, 2026."”
Verdict: The target was met: Antares Nuclear's Mark-0 went critical in early June 2026, Valar Atomics' Ward 250 on June 22, and Deployable Energy's Unity in the final days of June, with Aalo Atomics' Aalo-X adding a fourth on July 4 itself. Caveats: these were zero- or very low-power test-reactor demonstrations (Valar reached about 10 kWt), and several were sited at Idaho National Laboratory rather than fully off-lab.
PendingJobs & EconomySaid Jun 17, 2025Deadline Dec 31, 2028
“"As we roll out more Generative AI and agents, it should change the way our work is done. We will need fewer people doing some of the jobs that are being done today... in the next few years, we expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company."”
Verdict: Amazon announced large corporate job cuts later in 2025, but the company attributed them chiefly to bureaucracy and culture rather than AI, and the 'next few years' window runs to 2028, so the AI-attribution part remains open.
PendingAGI & SuperintelligenceSaid Jun 16, 2025Deadline Dec 31, 2026
“I think we're quite close to digital superintelligence. It may happen this year, and if it doesn't happen this year, next year for sure. Digital superintelligence, defined as smarter than any human at anything.”
Verdict: Deadline is end of 2026; 2025 passed with no system smarter than any human at anything, so resolution rests on the remainder of 2026.
Partially correctEnergy & Data CentersSaid Jun 11, 2025Deadline Jun 11, 2026
“We currently have 23 MultiCloud datacenters live with 47 more being built over the next 12 months. We expect triple-digit MultiCloud revenue growth to continue in FY26.”
Verdict: The revenue half resolved decisively: Oracle's Multicloud AI Database grew triple digits every quarter of FY26 (1,529% in Q1, 817% in Q2, 531% in Q3, and 404% in Q4, which Oracle called its fastest-growing business ever). The buildout half fell well short of 23 live + 47 more within 12 months: around and after the June 10, 2026 Q4 FY26 report, Oracle was still described as only "more than halfway through" building 72 Multicloud datacenters (~40 rather than ~70 live), and Oracle Database@AWS reached just 20 regions as of late May 2026.
CorrectSelf-DrivingSaid Jun 10, 2025Deadline Jun 28, 2025
“Tentatively, June 22. We are being super paranoid about safety, so the date could shift. First Tesla that drives itself from factory end of line all the way to a customer house is June 28.”
Verdict: Tesla began invite-only robotaxi rides in Austin on June 22, 2025 — though with safety monitors riding in the passenger seat — and completed a factory-to-customer autonomous delivery in the final days of June, meeting both dates in the post.
PendingCapabilitiesSaid Jun 10, 2025Deadline Dec 31, 2026
“2026 will likely see the arrival of systems that can figure out novel insights.”
Verdict: Deadline is end of 2026 and still open; OpenAI announced in September 2026 that it had reached its internal 'automated research intern' milestone, but independent evidence of AI systems producing genuinely novel insights remains contested.
PendingCapabilitiesSaid Jun 10, 2025Deadline Dec 31, 2027
“2027 may see the arrival of robots that can do tasks in the real world.”
Verdict: Hedged with 'may'; resolution will turn on how demanding a reading of 'tasks in the real world' is applied.
PendingCapabilitiesSaid Jun 10, 2025Deadline Sep 27, 2026
“Starbucks' generative AI barista assistant "Green Dot Assist," built on Microsoft Azure's OpenAI platform, is piloting in 35 stores this month, with a broad launch across U.S. and Canada stores slated for fiscal 2026: "It's just another example of how innovation technology is coming into service of our partners and making sure that we're doing all we can to simplify the operations."”
Verdict: Starbucks' fiscal 2026 ends September 27, 2026. Rollout appears under way — the Q1 FY2026 earnings call cited Green Dot Assist alongside digital menu boards as support for the Green Apron service model — but no confirmation yet of completed broad deployment across U.S. and Canada stores.
PendingCapabilitiesSaid Jun 10, 2025Deadline Dec 31, 2029
“"IBM is charting the next frontier in quantum computing... Our expertise across mathematics, physics, and engineering is paving the way for a large-scale, fault-tolerant quantum computer — one that will solve real-world challenges and unlock immense possibilities for business." Delivered by 2029, IBM Quantum Starling is expected to be the world's first large-scale, fault-tolerant quantum computer, running 100 million quantum operations on 200 logical qubits — 20,000 times more operations than today's quantum computers.”
Verdict: IBM says it remains on track: the Loon and Nighthawk processors were unveiled in November 2025, and Krishna reaffirmed the 2029 Starling target on IBM's Q1 2026 earnings call.
PendingAGI & SuperintelligenceSaid Jun 5, 2025Deadline Dec 31, 2030
“Will the AI think it has reached AGI by 2030? I would say we will just fall short of that timeline, so I think it'll take a bit longer... I don't think we'll quite get there by 2030, so my sense is it's slightly after that — but by 2030 there'll be such dramatic progress, we'll be dealing with the consequences of that progress in a big way.”
Verdict: Notably more conservative than Google DeepMind's own Demis Hassabis and Sergey Brin, who both pointed to roughly 2030; resolves at the end of 2030.
PendingJobs & EconomySaid May 28, 2025Deadline May 28, 2030
“AI could wipe out half of all entry-level white-collar jobs — in tech, finance, law and consulting — and spike US unemployment to 10-20% in the next one to five years.”
Verdict: The outer bound of the stated window is May 2030; US unemployment has so far remained far below the 10-20% range.
PendingEnergy & Data CentersSaid May 23, 2025Deadline Dec 31, 2030
“The United States will have 10 new large nuclear reactors under construction by 2030 and 5 GW of power uprates at existing plants, on the way to deploying 300 GW of net new nuclear capacity — quadrupling capacity to 400 GW — by 2050, in large part to power AI data centers.”
Verdict: No new large reactor construction had started as of September 2026; DOE selected TVA and Holtec for a large-reactor deployment milestone in December 2025, while progress so far has been concentrated in test reactors and SMRs.
CorrectCapabilitiesSaid May 22, 2025Deadline Dec 31, 2025
“We haven't yet demonstrated long-running agentic performance. You're seeing the first stumbling steps of that now, and should see much more conclusive evidence of that basically by the end of the year, with real software engineering agents doing real work.”
Verdict: By the end of 2025 agentic coding tools such as Claude Code and OpenAI's Codex agents were in broad production use doing genuine multi-step software engineering work, which meets the bar Douglas set. The related forecast he and Bricken made about reliable general computer-use agents ('go do my taxes') remained clearly unmet, so only the software-engineering half of their outlook landed.
PendingJobs & EconomySaid May 22, 2025Deadline Dec 31, 2028
“I think we're near guaranteed at this point to have effectively models that are capable of automating any white collar job by 2027, 2028 — or near guaranteed end of decade.”
Verdict: Douglas caveated that white-collar tasks are unusually susceptible to current algorithms because they can be retried on computers and have abundant internet data, unlike robotics or biology.
PendingJobs & EconomySaid May 22, 2025Deadline May 31, 2030
“There is this whole spectrum of crazy futures. But the one that I feel we're almost guaranteed to get — this is a strong statement to make — is one where, at the very least, you get a drop in white collar workers at some point in the next five years. I think it's very likely in two, but it seems almost overdetermined in five.”
Verdict: Douglas argued the decline would happen even if algorithmic progress stalled, because existing models are already good enough for many white-collar tasks given the right data; Trenton Bricken agreed with the five-year timeline.
PendingJobs & EconomySaid May 19, 2025Deadline Dec 31, 2030
“Based upon what we know today, we expect headcount [in operations and account services] will trend down by about 10% over the next five years or so, even as the business grows by another more than 25%. So total productivity of above 40%. But I would take the over on this projection, and I'll bet we will deliver even more as the tools and capabilities just keep getting better.”
Verdict: Five-year horizon from May 2025; resolves against JPMorgan's disclosed operations/account-services headcount around 2030.
PendingCapabilitiesSaid May 2, 2025Deadline Jan 20, 2029
“THE 1,000,000X BOOST: AI will become a million times more powerful during the Trump presidency. AI models are improving at a rate of about 3-4x per year... 100x better models running on 100x better chips running in data centers with 100x more chips is a total increase of 1,000,000x.”
— David Sacks, White House AI & Crypto Czar (Special Advisor for AI and Crypto); co-founder of Craft Ventures; All-In Podcast co-host (Investor). David Sacks on X, May 2, 2025
Verdict: Resolves at the end of Trump's term (Jan 20, 2029); Sacks' arithmetic requires ~100x model gains, ~100x chip gains and ~100x cluster growth over four years.
PendingCapabilitiesSaid Apr 29, 2025Deadline Oct 31, 2026
“I would guess that sometime in the next 12 to 18 months, we'll reach the point where most of the code that's going toward these efforts is written by AI. And I don't mean autocomplete... you give it a goal, it can run tests, it can improve things, it can find issues, it writes higher quality code than the average very good person on the team already.”
Verdict: The 18-month window closes at the end of October 2026. Zuckerberg's claim is scoped to code going toward Meta's AI/Llama efforts, not all Meta code, and Meta has not published a verifiable figure for the AI-written share.
IncorrectSelf-DrivingSaid Apr 22, 2025Deadline Dec 31, 2025
“Tesla will launch robotaxi service in Austin in June, and robotaxis will be in many other cities in the U.S. by the end of this year.”
Verdict: Robotaxi did launch in Austin in June 2025, but by the end of the year the service ran only in Austin and the San Francisco Bay Area — the latter with a chaperone in the car — instead of the many cities Musk promised. Tesla reached roughly seven U.S. metro areas only in mid-2026.
PendingHealthcareSaid Apr 20, 2025Deadline Dec 31, 2035
“I think one day maybe we can cure all disease with the help of AI... I think that's within reach. Maybe within the next decade or so, I don't see why not.”
Verdict: Hassabis was answering Pelley's question about whether AI could end all disease; the implied horizon runs to roughly 2035.
PendingAGI & SuperintelligenceSaid Apr 17, 2025Deadline Dec 31, 2045
“Asked when he expects a drop-in remote worker replacement: 'Maybe for me, that would be around 2045.' There will be no software-only intelligence explosion — AI will be one moving part among many, like the Industrial Revolution — but full automation will eventually produce explosive growth, with the economy doubling every year or two.”
Verdict: Offered as the explicit counter-scenario to the AI 2027 forecast discussed with Scott Alexander and Daniel Kokotajlo; the 2045 figure is Erdil's stated median for a drop-in remote worker, with Besiroglu endorsing similarly long, compute- and diffusion-constrained timelines.
CorrectHealthcareSaid Apr 17, 2025Deadline Dec 31, 2025
“26 million consumer calls were more accurately directed to the right advocate by an AI agent... We expect AI will direct over half of our calls to the best resource during 2025.”
Verdict: On the Q4 2025 call (January 27, 2026) UnitedHealth reported that 'over 80% of calls from members leverage AI tools to help answer members' questions faster and more accurately,' comfortably above the 'over half' target. Caveat: the figure is company-reported and the metric shifted slightly from AI 'directing' calls to calls that 'leverage AI tools.'
Verdict: The one-year deadline passed in April 2026 without anything resembling mass replacement of programmers. AI coding agents were adopted broadly and write a large and growing share of code, but human software developers remained employed in the millions and no 'vast majority' displacement occurred.
Verdict: AI math ability advanced dramatically within the window — gold-medal-level International Mathematical Olympiad performance in July 2025 and rapid gains on research-level math benchmarks and formal proof work — but by April 2026 systems had not generally matched top graduate students across open-ended research mathematics, so the claim is only partly borne out.
“Within 3 to 5 years we'll have what is called general intelligence, AGI, which can be defined as a system that is as smart as the smartest mathematician, physicist, artist, writer, thinker, politician. I call this the San Francisco consensus, because everyone who believes this is in San Francisco.”
Verdict: Outer bound of the 3–5 year window falls in April 2030. Schmidt has since restated a similar timeline (AGI around 2028–2030, superintelligence within a decade) at the RAISE Summit in Paris in July 2025.
“The computers are now doing self-improvement... they don't have to listen to us anymore. We call that superintelligence or ASI — computers that are smarter than the sum of humans. The San Francisco consensus is this occurs within six years.”
Verdict: Six years from April 2025 puts the deadline at April 2031; contemporaneous reporting rendered this as artificial superintelligence 'by 2031.'
PendingEnergy & Data CentersSaid Apr 9, 2025Deadline Dec 31, 2027
“One of the estimates that I think is most likely is that data centers will require an additional 29 gigawatts of power by 2027, and 67 more gigawatts by 2030.”
Verdict: Deadline is the end of 2027; Schmidt framed the figure as the estimate he considered most likely, immediately after noting that "many people think" the industry's share of US generation could go from 3% to 99%.
PendingEnergy & Data CentersSaid Apr 7, 2025Deadline Dec 31, 2027
“DOE seeks to enable the construction of AI infrastructure at select DOE sites to begin by the end of 2025, with a target of commencing operation by the end of 2027; in July 2025 DOE named Idaho National Laboratory, the Oak Ridge Reservation, the Paducah Gaseous Diffusion Plant and the Savannah River Site as the first four sites.”
Verdict: The end-of-2027 operations target has not yet arrived, but the interim milestone slipped: DOE was still soliciting proposals with a January 30, 2026 deadline and only announced private-sector partnership agreements for Paducah and Savannah River in mid-2026.
PendingAGI & SuperintelligenceSaid Apr 3, 2025Deadline Dec 31, 2027
“We predict a superhuman coder — an AI that can do any coding task the best AGI-company engineer does, while being much faster and cheaper — arrives around 2027, with automated AI research driving superintelligence shortly after.”
Verdict: AI 2027 paired a month-by-month scenario with a timelines model whose median for a superhuman coder was 2027–2028; in a December 2025 update the authors moved that median to roughly 2032 and superintelligence to about 2034, so the original target remains unresolved until end-2027 but looks unlikely on the authors' own revised model.
PendingCapabilitiesSaid Mar 31, 2025Deadline Dec 31, 2030
“95% [of code] is going to be AI generated [within five years]. Very little is going to be line by line human-written code. Now that doesn't mean that the AI is doing the software engineering job — authorship is still going to be human.”
Verdict: Five-year horizon from March 2025 puts resolution at 2030; measured against the share of newly written code that is AI-generated.
PendingCapabilitiesSaid Mar 20, 2025Deadline Dec 31, 2026
“I think we will literally build a legion, at least one legion of robots this year, and then probably 10 legions next year — i.e. roughly 50,000 Optimus humanoid robots in 2026.”
Verdict: Deadline is end of 2026. Tesla built no 'legion' in 2025, and as of September 2026 series production at Fremont had not started (Tesla's Q2 2026 update said production was 'anticipated later this year'), making 50,000 units in 2026 effectively unreachable.
PendingEnergy & Data CentersSaid Mar 18, 2025Deadline Dec 31, 2028
“Annual data center capital expenditure will reach $1 trillion by 2028: 'I've said before that I expect data center buildout to reach $1 trillion, and I'm fairly certain that we're going to reach that very soon' — driven by accelerated computing and the recognition that the future of software requires capital investment.”
Verdict: Resolves at the end of 2028 against industry-wide data center capex. Independent forecasters (e.g., Dell'Oro) put the $1 trillion annual mark a year later, in 2029.
PendingCapabilitiesSaid Mar 18, 2025Deadline Dec 31, 2028
“One architecture per year, one new AI infrastructure platform every year: Blackwell Ultra in the second half of 2025; Vera Rubin — the Vera CPU with the Rubin GPU, NVLink 144 — in the second half of 2026; Rubin Ultra, NVL576, in the second half of 2027; and then Feynman in 2028.”
Verdict: The first leg has held up — Blackwell Ultra (GB300) arrived in the second half of 2025, and Huang said at his CES 2026 keynote that Vera Rubin was in full production for second-half-2026 availability — but Rubin Ultra (2027) and Feynman (2028) are still ahead, and supply-chain reporting has already floated slippage of the NVL576 'Kyber' rack toward 2028.
PendingAGI & SuperintelligenceSaid Mar 2025Deadline Mar 17, 2035
“AGI is five to ten years away — a system that matches all human cognitive capabilities.”
— Demis Hassabis, CEO, Google DeepMind; Nobel laureate (Tech CEO). Interviews (2025)
Partially correctJobs & EconomySaid Mar 10, 2025Deadline Mar 10, 2026
“In three to six months, AI will be writing 90% of the code. In twelve months, essentially all of it.”
— Dario Amodei, CEO, Anthropic (Tech CEO). Council on Foreign Relations interview (March 2025)
Verdict: AI code generation grew explosively, and some AI labs report that a large majority of their own code is now AI-written — but 90% across the software industry did not materialize on this timeline.
PendingEnergy & Data CentersSaid Mar 10, 2025Deadline Dec 31, 2045
“We're expecting a six-fold increase in power demand over the next 20 years compared to what we saw over the prior 20 — roughly a 55% increase in US power demand, driven by data centers, onshored manufacturing and electrification.”
Verdict: Twenty-year horizon (to ~2045); resolvable against EIA US electricity demand data. Note: personType is set to "Tech CEO" only because the schema has no utility/energy-executive category.
PendingJobs & EconomySaid Feb 26, 2025Deadline Dec 31, 2025
“"We're not going to hire any new engineers this year. We're seeing a 30% productivity increase on engineering, and we're going to really continue to ride that up."”
Verdict: The deadline has passed, but Salesforce has not published 2025 engineering-headcount data and reporting during the year noted continued engineering job postings alongside AI-driven support-role cuts; the entry needs an audit of Salesforce hiring disclosures before it can be scored.
Partially correctCapabilitiesSaid Feb 12, 2025Deadline Dec 31, 2025
“We will next ship GPT-4.5, the model we called Orion internally, as our last non-chain-of-thought model... In both ChatGPT and our API, we will release GPT-5 as a system that integrates a lot of our technology, including o3. We will no longer ship o3 as a standalone model. [Timing: GPT-4.5 in 'weeks', GPT-5 in 'months'.]”
Verdict: GPT-4.5 shipped 27 February 2025 (about two weeks later) and GPT-5 shipped 7 August 2025 (within 'months'), so the timing held; however OpenAI did ship o3 as a standalone model on 16 April 2025, contradicting the 'no longer ship o3 as a standalone model' part.
Partially correctCapabilitiesSaid Feb 8, 2025Deadline Dec 31, 2025
“Our very first reasoning model was like a top one-millionth competitive programmer in the world... o3, which we talked about publicly in December, is the 175th best competitive programmer in the world. I think our internal benchmark is now around 50, and maybe we'll hit number one by the end of this year.”
Verdict: OpenAI's reasoning models did reach the very top tier during 2025 — a perfect 12/12 at the September 2025 ICPC World Finals, enough for a first-place human team ranking — but no OpenAI model was verifiably the world's number-one individual competitive programmer by December 31, 2025: human champion Przemysław 'Psyho' Dębiak beat OpenAI's model at the AtCoder World Tour Finals heuristic contest in July 2025, and OpenAI only swept the top humans at AtCoder in mid-2026.
Partially correctEnergy & Data CentersSaid Feb 6, 2025Deadline Dec 31, 2025
“We spent $26.3 billion in capex in Q4, and I think that is reasonably representative of what you expect an annualized capex rate in 2025 [i.e. roughly $100 billion]. The vast majority of that capex spend is on AI for AWS.”
Verdict: Amazon did blow past $100 billion, but the run-rate forecast badly understated the ramp: guidance was lifted to ~$125 billion in October 2025 and full-year 2025 capex finished at about $131.8 billion, roughly a third above the February figure.
Partially correctEnergy & Data CentersSaid Feb 4, 2025Deadline Dec 31, 2025
“We are confident about the opportunities ahead, and to accelerate our progress, we expect to invest approximately $75 billion in capital expenditures in 2025.”
Verdict: The direction was right but the number was well short: Alphabet raised guidance to $85 billion in July 2025 and finished the year at roughly $91.4 billion of capex, about 22% above the February forecast.
IncorrectCapabilitiesSaid Jan 29, 2025Deadline Dec 31, 2025
“We expect to have several thousand Optimus robots produced at Tesla and doing useful things by the end of this year.”
Verdict: Tesla did not produce several thousand working Optimus robots in 2025; output stalled on hardware redesigns and the robots that existed were used for demonstrations and limited pilot tasks, with volume production of the V3 design deferred to 2026.
IncorrectSelf-DrivingSaid Jan 29, 2025Deadline Jun 30, 2025
“Teslas will be in the wild with no one in them, in June in Austin. This is not some far-off mythical situation, it's five, six months away.”
Verdict: Tesla did launch a paid robotaxi service in Austin in June 2025, but every vehicle carried a Tesla safety monitor in the front passenger seat; cars genuinely 'with no one in them' did not carry public riders until January 22, 2026.
PendingSelf-DrivingSaid Jan 28, 2025Deadline Dec 31, 2030
“"Within five years, we expect to approach about $2 billion in total annual revenue from Super Cruise... This is how we are setting the stage for recurring high-margin revenue streams from subscriptions."”
Verdict: Deadline is early 2030. GM's Super Cruise-equipped fleet roughly doubled in 2025, but reported Super Cruise revenue remains in the hundreds of millions, far short of $2 billion.
PendingEnergy & Data CentersSaid Jan 28, 2025Deadline Dec 31, 2028
“Nvidia today trades 20 times revenue, it's trading 24 times earnings. People talk about the bubble, but this isn't what bubbles are made of. As far as I can tell, the world's going to have a compute shortage for the next three or four years. I don't think anything yesterday [DeepSeek] changed that.”
Verdict: Horizon runs through roughly the end of 2028; through 2026 AI compute has remained supply-constrained (hyperscaler capex guidance rising, chip and memory shortages), but the full three-to-four-year window is not yet complete.
Partially correctJobs & EconomySaid Jan 24, 2025Deadline Dec 31, 2025
“Meta is building a 2GW+ datacenter that is so large it would cover a significant part of Manhattan. We'll bring online ~1GW of compute in '25 and we'll end the year with more than 1.3 million GPUs. We're planning to invest $60-65B in capex this year.”
Verdict: Meta overshot rather than missed its spending plan: full-year 2025 capex came in above the guided range at roughly $72 billion (with 2026 guidance far higher), and the Louisiana Hyperion campus proceeded as the 2GW+ build described. The compute-online and 1.3 million GPU figures are not independently auditable, so only the capex leg can be scored — and it was exceeded, not met.
Partially correctCapabilitiesSaid Jan 24, 2025Deadline Dec 31, 2025
“This will be a defining year for AI. In 2025, I expect Meta AI will be the leading assistant serving more than 1 billion people.”
Verdict: The headcount target was hit early — Zuckerberg announced at Meta's May 2025 shareholder meeting that Meta AI had passed 1 billion monthly active users across Meta's apps. 'Leading assistant' is the weaker half: ChatGPT continued to dominate deliberate assistant usage and engagement, and Meta AI's reach is largely a byproduct of its placement inside Facebook, Instagram and WhatsApp.
IncorrectCapabilitiesSaid Jan 24, 2025Deadline Dec 31, 2025
“In 2025, I expect... Llama 4 will become the leading state of the art model, and we'll build an AI engineer that will start contributing increasing amounts of code to our R&D efforts.”
Verdict: Llama 4 shipped in April 2025 but was not state of the art: it drew a benchmark-submission controversy, the flagship Behemoth was pushed back and then reportedly shelved for weak internal performance, and Meta pivoted to a new superintelligence lab reportedly considering closed models. AI-written code did grow at Meta, but the central prediction failed.
PendingCapabilitiesSaid Jan 23, 2025Deadline Jan 23, 2030
“The shelf life of the current [LLM] paradigm is fairly short, probably three to five years. I think within five years, nobody in their right mind would use them anymore, at least not as the central component of AI systems.”
Verdict: Deadline is January 2030. As of 2026, LLMs remain the central component of essentially all deployed AI systems; LeCun left Meta in late 2025 and founded AMI Labs to pursue world-model architectures instead.
PendingHealthcareSaid Jan 23, 2025Deadline Dec 31, 2035
“If you think about what we might expect humans to accomplish in an area like biology in 100 years, I think a doubling of the human lifespan is not at all crazy. And then if AI is able to accelerate that, we may be able to get that in five to ten years.”
Verdict: Amodei framed a doubling of human lifespan as achievable in as little as five to ten years 'if we really get this AI stuff right'; outer bound treated as end-2035.
IncorrectHealthcareSaid Jan 21, 2025Deadline Dec 31, 2025
“We'll hopefully have some AI-designed drugs in clinical trials by the end of the year. That's the plan.”
Verdict: Isomorphic Labs did not put any AI-designed drug into clinical trials during 2025. At Davos on 20 January 2026 Hassabis said the company now expects its first clinical trials by the end of 2026, which Reuters described as a delay to the original timeline.
PendingAGI & SuperintelligenceSaid Jan 21, 2025Deadline Dec 31, 2027
“I don't know exactly when it'll come, I don't know if it'll be 2027. I think it's plausible it could be longer than that. I don't think it will be a whole bunch longer than that when AI systems are better than humans at almost everything. Better than almost all humans at almost everything. And then eventually better than all humans at everything, even robotics.”
Verdict: Target year 2027 has not yet arrived; Amodei hedged that it could be 'plausibly longer,' so resolution should focus on whether AI systems are better than almost all humans at almost everything by end of 2027.
PendingEnergy & Data CentersSaid Jan 21, 2025Deadline Jan 21, 2029
“The Stargate Project is a new company which intends to invest $500 billion over the next four years building new AI infrastructure for OpenAI in the United States.”
Verdict: Four-year clock runs to January 2029. OpenAI said in late 2025 that over $400 billion of the commitment was lined up, but The Information reported in February 2026 that Stargate had fallen short of locking down the full 10GW of capacity through SoftBank and Oracle.
IncorrectJobs & EconomySaid Jan 21, 2025Deadline Dec 31, 2025
“Stargate is "a new American company that will invest $500 billion at least in AI infrastructure in the United States and very, very quickly, moving very rapidly, creating over 100,000 American jobs almost immediately."”
Verdict: Nothing close to 100,000 jobs appeared 'almost immediately': Bloomberg reported the flagship Abilene site was contractually guaranteed only 57 full-time positions, and by mid-2026 Oracle described roughly 9,000 temporary construction jobs and about 1,000 eventual permanent jobs there. OpenAI's own September 2025 expansion projected 'over 25,000 onsite jobs' across five new sites as a forward-looking estimate, well short of 100,000 delivered in 2025.
Partially correctJobs & EconomySaid Jan 10, 2025Deadline Dec 31, 2025
“Probably in 2025, we at Meta, as well as the other companies that are basically working on this, are going to have an AI that can effectively be a sort of midlevel engineer that you have at your company that can write code.”
Verdict: Coding agents advanced sharply in 2025 and now handle substantial amounts of real engineering work, but no lab — Meta included — fielded a system that autonomously fills a mid-level engineer's role; Meta shipped no such 'AI engineer' product in 2025 and Zuckerberg's own follow-on framing pushed the milestone of AI writing most of Meta's AI code into 2026.
CorrectJobs & EconomySaid Jan 6, 2025Deadline Dec 31, 2025
“In 2025, we may see the first AI agents join the workforce and materially change the output of companies.”
— Sam Altman, CEO, OpenAI (Tech CEO). Reflections (personal blog, January 2025)
Verdict: AI coding, research, and support agents were deployed in production across many companies during 2025, with measurable output changes in software engineering and customer service.
PendingAGI & SuperintelligenceSaid Jan 6, 2025Deadline Jan 20, 2029
“I think AGI will probably get developed during this president's term, and getting that right seems really important.”
Verdict: Deadline is the end of Trump's term (20 January 2029); note Altman in the same interview called AGI 'a very sloppy term', so resolution will require a stated AGI definition.
IncorrectCapabilitiesSaid Jan 4, 2025Deadline Dec 31, 2025
“My most contrarian belief [for 2025]: OpenAI loses its lead, loses its nonprofit-to-for-profit transition, and becomes the number four player in AI.”
Verdict: OpenAI completed its restructuring into a public benefit corporation under the OpenAI Foundation in October 2025 — the opposite of 'losing' the for-profit transition — and finished 2025 at roughly a $20B annualized revenue run rate, remaining the largest consumer AI player rather than falling to number four.
PendingJobs & EconomySaid Jan 4, 2025Deadline Dec 31, 2028
“My most contrarian belief: America, at some point over the next four years, will print at least one year of greater than 5% real GDP growth. I think productivity is going to go vertical because of AI and deregulation. I think there could be a world where this is the late '90s.”
Verdict: Window runs through calendar 2028; no year of >5% US real GDP growth has been recorded so far, leaving 2026-2028 to resolve it.
CorrectJobs & EconomySaid Jan 3, 2025Deadline Jun 30, 2025
“In FY2025, Microsoft is on track to invest approximately $80 billion to build out AI-enabled datacenters to train AI models and deploy AI and cloud-based applications around the world.”
Verdict: Microsoft's fiscal 2025 ended June 30, 2025 with total capital expenditures including finance leases of roughly $88 billion — above the $80 billion figure — driven overwhelmingly by AI and cloud datacenter buildout, though the reported total covers all capex rather than AI datacenters alone.
CorrectAGI & SuperintelligenceSaid Jan 1, 2025Deadline Dec 31, 2025
“We will not see artificial general intelligence this year [2025], despite claims by Elon Musk to the contrary. (People will also continue to play games to weaken the definition or even try to define it in financial rather than scientific terms.)”
Verdict: No AGI arrived in 2025; independent year-end gradings of dated 2025 AI predictions scored this one correct, and by late 2025 even prominent optimists were pushing AGI timelines out by years.
CorrectCapabilitiesSaid Jan 1, 2025Deadline Dec 31, 2025
“No single system will solve more than 4 of the AI 2027 Marcus-Brundage tasks by the end of 2025. I wouldn't be shocked if none were reliably solved by the end of the year.”
Verdict: A year-end review of 2025 AI predictions judged this correct, concluding that at most one of the ten tasks (a reading-comprehension-style item) was arguably passed, and certainly not four by any single system.
CorrectJobs & EconomySaid Jan 1, 2025Deadline Dec 31, 2025
“Profits from AI models will continue to be modest or nonexistent (chip-making companies will continue to do well though, in supplying hardware to the companies that build the models; shovels will continue to sell well throughout the gold rush.)”
Verdict: Through 2025 the leading model developers continued to run large operating losses while revenue growth was funded by external capital, and the hardware suppliers — Nvidia above all — captured the profits; the 'shovel-sellers do best' pattern held. Both halves of the prediction came true.
CorrectCapabilitiesSaid Jan 1, 2025Deadline Dec 31, 2025
“AI "Agents" will be endlessly hyped throughout 2025 but far from reliable, except possibly in very narrow use cases.”
Verdict: 2025 was saturated with agent marketing while reliability lagged: MIT's NANDA work produced the widely-cited finding that about 95% of internally built enterprise genAI pilots delivered no measurable return, McKinsey-style surveys found very few agents reaching production, and teams that shipped did so by narrowing scope. Independent year-end gradings scored this prediction correct.
CorrectCapabilitiesSaid Jan 1, 2025Deadline Dec 31, 2025
“Humanoid robotics will see a lot of hype, but nobody will release anything to remotely as capable as Rosie the Robot. Motor control may be impressive, but situational awareness and cognitive flexibility will remain poor.”
Verdict: Humanoid-robot announcements and demos proliferated in 2025, including Tesla's Optimus, but no shipped system approached a general-purpose household robot; impressive locomotion and manipulation demos continued to coexist with brittle autonomy and heavy teleoperation or scripting.
CorrectSafety & RiskSaid Jan 1, 2025Deadline Dec 31, 2025
“The US will continue to have very little regulation protecting its consumers from the risks of generative AI. When it comes to regulation, much of the world will increasingly look to Europe for guidance.”
Verdict: No comprehensive federal generative-AI consumer-protection statute was enacted in 2025; US federal policy moved toward deregulation and promotion of AI build-out, leaving substantive rules to individual states and to the EU AI Act.
PendingEnergy & Data CentersSaid 2025Deadline Dec 31, 2030
“Data center electricity demand is growing at about 20% annually through the end of the decade, from roughly 4% of current U.S. electricity demand to 10%-12% by 2030.”
Verdict: Page is undated on-site but was published around October 2025; BlackRock's own later 2026 Global Outlook work cites materially higher figures (15-20%+ of current US demand by 2030), so the firm's own estimates have since moved.
PendingAGI & SuperintelligenceSaid Dec 31, 2024Deadline Dec 31, 2027
“Bet with Miles Brundage at 10:1 odds: if AI systems can perform 8 of the 10 specified tasks by the end of 2027, as determined by our panel of judges, Gary will donate $2,000 to a charity of Miles' choice; if AI can do fewer than 8, Miles will donate $20,000 to a charity of Gary's choice.”
Verdict: Tasks include watching an unseen movie and following plot twists, constructing bug-free code of more than 10,000 lines from natural-language spec, and similar; tracked on Metaculus and Manifold. Marcus's separate end-of-2025 checkpoint (no system solving more than 4 of the 10) was judged correct, but the bet itself does not resolve until after 2027.
CorrectCapabilitiesSaid Dec 20, 2024Deadline Dec 31, 2025
“Early data points suggest that the upcoming ARC-AGI-2 benchmark will still pose a significant challenge to o3, potentially reducing its score to under 30% even at high compute (while a smart human would still be able to score over 95% with no training).”
Verdict: When ARC-AGI-2 launched in March 2025, ARC Prize estimated o3-preview-low would score ~4% and o3-high only 15–20% even at very high compute; subsequent official testing put o3 (medium) at roughly 3%. Every measured or estimated o3 configuration fell far below the 30% ceiling Chollet predicted.
PendingEnergy & Data CentersSaid Dec 20, 2024Deadline Dec 31, 2028
“Data centers consumed about 4.4% of total U.S. electricity in 2023 and are expected to consume approximately 6.7% to 12% of total U.S. electricity by 2028; data center load growth has tripled over the past decade and is projected to double or triple by 2028 (176 TWh in 2023 rising to 325–580 TWh in 2028).”
Verdict: Target year 2028 is still in the future; LBNL's own 2025 update has since revised the trajectory upward (11.8% of U.S. electricity by 2030, 9.5–15.3% range).
PendingCapabilitiesSaid Dec 17, 2024Deadline Dec 31, 2026
“AI will surpass the world's best competitive programmer within one to two years.”
Verdict: Window closes at the end of 2026. Trending toward correct: OpenAI's system reportedly beat every human finalist at the AtCoder World Tour Finals in July 2026, about 19 months after the prediction, after losing narrowly to human champion Psyho at the 2025 edition.
PendingJobs & EconomySaid Dec 16, 2024Deadline Dec 31, 2028
“I'm now excited to commit this $100 billion and 100,000 jobs into the United States — a $100 billion investment over the next four years, focused on artificial intelligence and related infrastructure.”
Verdict: Son made a similar $50 billion / 50,000 jobs pledge in December 2016; SoftBank ultimately invested roughly $75 billion via the Vision Fund, with job creation never clearly verified. The 2024 pledge became the financial backbone of Stargate, announced a month later.
Partially correctJobs & EconomySaid Dec 12, 2024Deadline Dec 31, 2025
“"We stopped hiring about a year ago. We were 4,500, now we're 3,500... People stay about 5 years — so 20% leave every year — and by not hiring, we're simply shrinking." AI is already doing the work of hundreds of staff, and headcount will keep falling through natural attrition rather than layoffs.”
Verdict: Headcount did keep shrinking by attrition — Siemiatkowski said in May 2025 the company had gone from about 5,000 to 'almost 3,000' — but the 'stopped hiring entirely' framing did not hold: Klarna continued advertising open roles and in 2025 began rehiring human customer-service agents after AI service quality disappointed.
PendingCapabilitiesSaid Nov 11, 2024Deadline Dec 31, 2027
“Right now, I think most of the frontier model companies are operating in roughly 1 billion scale, plus or minus a factor of three... I think next year we're going to go to a few billion, and then 2026, we may go to above 10 billion. And probably by 2027, their ambitions to build hundred billion dollar clusters. And I think all of that actually will happen.”
Verdict: The intermediate steps have broadly tracked his curve so far; the $100B-cluster milestone resolves at the end of 2027.
Partially correctEnergy & Data CentersSaid Oct 31, 2024Deadline Dec 31, 2024
“Amazon plans to spend about $75 billion on capital expenditures in 2024 and expects an even higher number in 2025, driven primarily by generative AI: 'It is a really unusually large, maybe once-in-a-lifetime type of opportunity.'”
Verdict: Amazon's 2024 capital investments came in at roughly $83 billion, about 10% above the $75 billion Jassy cited two months before year-end; the accompanying call that 2025 would be higher was correct (2025 capex reached ~$131.8 billion).
PendingEnergy & Data CentersSaid Oct 29, 2024Deadline Dec 31, 2028
“'In the Data Center alone, we expect the AI accelerator TAM will grow at more than 60% annually to $500 billion in 2028. To put that in context, this is roughly equivalent to annual sales for the entire semiconductor industry in 2023.'”
Verdict: Su gave a near-identical $500 billion/2028 TAM figure at AMD's Advancing AI keynote on October 10, 2024; resolves against 2028 data center AI accelerator market size.
PendingAGI & SuperintelligenceSaid Oct 29, 2024Deadline Dec 31, 2035
“Artificial super intelligence will be 10,000 times smarter than a human brain and will exist by 2035; realising it will require hundreds of billions of dollars of investment.”
Verdict: At the same event Son forecast $900 trillion in cumulative capex for generative AI; the ASI deadline is 2035.
PendingEnergy & Data CentersSaid Oct 23, 2024Deadline Dec 31, 2027
“If we achieve the midpoint of our development expectations, this pace of development is expected to more than double our combined renewable generation portfolio, growing from 38 gigawatts today to potentially 81 gigawatts by the end of 2027 — 'today, there are forecasts for an approximate 6x increase in power demand growth in the next 20 years versus the prior 20.'”
Verdict: Tracking: NextEra reported a development backlog of roughly 30 GW as of its Q4 2025 call; the 81 GW portfolio target resolves at the end of 2027.
PendingEnergy & Data CentersSaid Oct 16, 2024Deadline Dec 31, 2039
“Amazon and X-energy "aim to bring more than 5 gigawatts [of new small modular reactor capacity] online in the United States by 2039, the largest commercial deployment target of SMRs to date," starting with an initial 320-megawatt Xe-100 project with Energy Northwest in central Washington.”
Verdict: Long-horizon target; the Energy Northwest Cascade project is currently expected to begin operating in the mid-2030s, and X-energy added Korean industrial partners and went public in April 2026.
PendingEnergy & Data CentersSaid Oct 14, 2024Deadline Dec 31, 2030
“Under Google's master plant development agreement with Kairos Power, Kairos' first small modular reactor will be brought online by 2030, with additional reactors built through 2035; in total the deal will provide "500 megawatts of new 24/7 carbon-free energy to the U.S. electricity grids."”
Verdict: In August 2025 Google, Kairos and TVA signed a PPA for the Hermes 2 plant in Oak Ridge, Tennessee, whose output was expanded from 28 MW to 50 MW and which is scheduled to come online in 2030.
PendingEnergy & Data CentersSaid Oct 14, 2024Deadline Dec 31, 2035
“Kairos Power and Google's Master Plant Development Agreement creates "a path to deploy a U.S. fleet of advanced nuclear power projects totaling 500 MW by 2035" — the first corporate agreement for multiple deployments of a single advanced reactor design in the United States.”
Verdict: Framed explicitly as an 'as early as' estimate, not a commitment.
PendingHealthcareSaid Oct 11, 2024Deadline Dec 31, 2036
“My basic prediction is that AI-enabled biology and medicine will allow us to compress the progress that human biologists would have achieved over the next 50-100 years into 5-10 years. I'll refer to this as the 'compressed 21st century': the idea that after powerful AI is developed, we will in a few years make all the progress in biology and medicine that we would have made in the whole 21st century.”
Verdict: The 5-10 year window runs from the arrival of what Amodei calls 'powerful AI', which in the same essay he says could come as early as 2026 — hence an outer bound of roughly 2036 rather than the 2034 originally scraped.
IncorrectSelf-DrivingSaid Oct 10, 2024Deadline Dec 31, 2025
“We do expect actually to start fully autonomous, unsupervised FSD in Texas and California next year. And obviously that's with the Model 3 and Model Y.”
Verdict: Tesla's Austin robotaxi service launched in June 2025 with in-car safety monitors, not unsupervised operation, and California rides ran with a safety driver under state permitting rules. No customer Model 3 or Model Y received unsupervised FSD in 2025; in April 2026 Musk pushed consumer unsupervised FSD to Q4 2026 at the earliest.
PendingSelf-DrivingSaid Oct 10, 2024Deadline Dec 31, 2026
“We expect to be in production with the Cybercab, which is really highly optimized for autonomous transport, in probably — well, I tend to be a little optimistic with time frames — but in 2026. So yeah, before 2027, let me put it that way.”
Verdict: Deadline is end of 2026. Tesla announced in February 2026 that the first Cybercab had rolled off the production line at Giga Texas, so the claim appears on track pending volume production.
PendingAGI & SuperintelligenceSaid Oct 3, 2024Deadline Oct 3, 2027
“Artificial general intelligence will arrive within the next two to three years, and AI will soon run entire households — monitoring the health of family members, calling the doctor when needed, doing grocery shopping, making reservations, judging optimal investments and tutoring young children.”
Verdict: Target date set by interpreting 'a few thousand days' as roughly 2,000–3,000 days from September 2024 (i.e. ~2030–2032); the claim is hedged ('it is possible', 'may take longer').
PendingEnergy & Data CentersSaid Sep 20, 2024Deadline Dec 31, 2028
“Constellation expects the Unit 1 reactor at Three Mile Island — to be renamed the Crane Clean Energy Center — to come back online in 2028, subject to Nuclear Regulatory Commission approval, returning 835 MW of carbon-free nuclear power to the grid under a 20-year power purchase agreement that matches the electricity Microsoft's data centers consume in the PJM region.”
Verdict: Still on track as of September 2026: Constellation has since pulled the target forward to 2027, FERC approved a capacity-rights transfer from Eddystone in June 2026, and NRC restart review (including a finding of no significant environmental impact) is ongoing.
“There are still significant gaps to close, but I believe it is plausible that an AGI system could be built in as little as three years — 'highly autonomous systems that outperform humans at most economically valuable work.' No one knows how to ensure that AGI systems will be safe and controlled.”
Verdict: Deadline is September 2027. Saunders cited the then-new o1 model surpassing human experts on some technical benchmarks as evidence for the short timeline; resolution will turn on whether a system meeting OpenAI's own AGI definition exists by then.
IncorrectCapabilitiesSaid Sep 12, 2024Deadline Dec 31, 2025
“The future of AI is agents, and it's here. Our vision is bold: to empower one billion agents with Agentforce by the end of 2025.”
Verdict: Salesforce never reported anything close to a billion deployed agents by the end of 2025; roughly fifteen months after launch the company was touting ~29,000 Agentforce deals and about $800M in Agentforce ARR, and disclosed adoption metrics (deals, conversations, 'agentic work units') rather than agent counts. Benioff quietly dropped the billion-agent framing, and tepid Agentforce uptake was reported as a drag on the stock through 2025.
Partially correctHealthcareSaid Aug 8, 2024Deadline Feb 8, 2026
“Highly complementary pipeline with approximately 10 clinical readouts expected over the next 18 months (Recursion alone expects 7 clinical trial readouts in that period).”
Verdict: Several guided readouts did arrive (REC-994 in CCM in September 2024, REC-2282 in NF2 in Q4 2024, REC-4881 FAP data in May and December 2025), but the combined company fell short of ~10: in May 2025 Recursion terminated REC-994, REC-2282 and REC-3964 and cut other programs, the REC-4881 LILAC cancer trial was deprioritised and later terminated without a readout, and the REC-3964 C. difficile readout never occurred.
IncorrectCapabilitiesSaid Aug 2, 2024Deadline Dec 31, 2024
“Well, hopefully end of this year [2024]... Grok 3, if we're lucky. [And] Grok 3 will be, I don't know, an order of magnitude better than Grok 2 — hoping for it to be state-of-the-art. I mean, this is the goal. We may fail at this goal. That's the aspiration.”
Verdict: Grok 3 did not ship in 2024; it launched on February 17, 2025, roughly two months past the stated 'end of this year' target. At release it did post leading benchmark and Chatbot Arena scores, but only briefly before rivals leapfrogged it.
PendingHealthcareSaid Aug 2, 2024Deadline Aug 2, 2029
“Even with only roughly 10, 15% of the electrodes working with Noland, our first patient, we were able to achieve twice the world record [in bits per second]. So I think we'll start vastly exceeding the world record by orders of magnitude in the years to come — start getting to, I don't know, 100 bits per second, a thousand. Maybe five years from now, we might be at a megabit, faster than any human could possibly communicate by typing or speaking.”
Verdict: Hedged ('maybe... might'), but numerically concrete: a ~1,000,000 bits-per-second brain-computer interface by August 2029, versus the single-digit-to-low-tens bps demonstrated at the time of the claim.
Partially correctCapabilitiesSaid Jul 31, 2024Deadline Dec 31, 2024
“We had a strong quarter, and Meta AI is on track to be the most used AI assistant in the world by the end of the year.”
Verdict: By Meta's own metric the target was met on paper: Meta AI reached nearly 500 million monthly actives by September 2024 and almost 600 million by December, more than any rival's disclosed user base at the time. But the figure counts incidental exposure inside Facebook, Instagram and WhatsApp rather than deliberate use, and ChatGPT remained the assistant people actively chose, so 'most used' is contested.
PendingEnergy & Data CentersSaid Jul 24, 2024Deadline Dec 31, 2031
“Renewable energy demand will triple over the next seven years — compared with what was built over the prior seven — as data center growth accelerates to facilitate the economy's shift to artificial intelligence.”
Verdict: Resolves against US renewables-and-storage additions in 2024-2031 versus 2017-2024 (NextEra cited a need for roughly 375-450 GW of new renewables and storage over the period).
IncorrectCapabilitiesSaid Jul 23, 2024Deadline Dec 31, 2025
“Starting next year, we expect future Llama models to become the most advanced in the industry.”
Verdict: Llama 4 launched in April 2025 to a poor reception and a benchmark-submission controversy; the flagship Llama 4 Behemoth was delayed from April to mid-2025 and then reportedly shelved over weak internal performance, and Meta reorganized into a superintelligence lab reportedly weighing closed-source models. Llama was not the most advanced family in the industry at any point in 2025.
Partially correctCapabilitiesSaid Jul 22, 2024Deadline Dec 31, 2025
“Tesla will have genuinely useful humanoid robots in low production for Tesla internal use next year.”
Verdict: Tesla did build and deploy a small number of Optimus units internally for demos and narrow tasks during 2025, but never reached genuine 'low production' of useful robots; design changes pushed the production-intent V3 version and its production line into 2026.
PendingCapabilitiesSaid Jul 22, 2024Deadline Dec 31, 2026
“Hopefully, high production [of Tesla's humanoid robots] for other companies in 2026.”
Verdict: Deadline is end of 2026; Tesla's Optimus production timeline has already slipped once, with the production-intent V3 design moved into 2026.
Partially correctHealthcareSaid Jun 24, 2024Deadline Jun 30, 2025
“REC-4881 Familial Adenomatous Polyposis — preliminary Phase 2 data readout in the first half of 2025; REC-4881 Advanced AXIN1/APC-Mutant Cancers — preliminary Phase 2 data readout in the first half of 2025.”
Verdict: The FAP readout landed on time: preliminary Phase 1b/2 TUPELO data (median 43% polyp-burden reduction at week 13) were presented at DDW on 4 May 2025. The AXIN1/APC-mutant cancer readout never happened — the LILAC trial was deprioritised after the Exscientia merger and was later listed as terminated by sponsor decision.
Partially correctHealthcareSaid Jun 24, 2024Deadline Dec 31, 2025
“REC-3964 Clostridioides difficile Infection — Phase 2 study initiation in the fourth quarter of 2024 and preliminary data readout by the end of 2025.”
Verdict: The Phase 2 ALDER trial did start on schedule — first patient dosed in October 2024 — but Recursion discontinued the program in its Q1 2025 update (May 2025), citing low recurrence rates with existing therapies and seeking out-licensing, so no preliminary efficacy readout was delivered by the end of 2025.
PendingAGI & SuperintelligenceSaid Jun 21, 2024Deadline Jun 21, 2029
“AGI — AI that is one to 10 times smarter than humans — will arrive in the next three to five years, earlier than I had anticipated.”
Verdict: Outer bound of the stated window is mid-2029; Son revised this forecast down to two-to-three years in October 2024.
PendingSelf-DrivingSaid Jun 12, 2024Deadline Dec 31, 2029
“ARK's price target for Tesla is $2,600 per share in 2029, with bear and bull cases valued between ~$2,000 and $3,100 per share; nearly 90% of Tesla's enterprise value and earnings will be attributed to the robotaxi business in 2029, while electric vehicles approximate a quarter of total sales and ~10% of earnings potential.”
Verdict: Resolves on Tesla's share price in 2029 against the $2,600 expected-value case (bear ~$2,000 / bull ~$3,100); the robotaxi-dominated earnings mix is a secondary check.
PendingCapabilitiesSaid Jun 8, 2024Deadline Dec 31, 2026
“If you look at the trajectory of improvement, systems like GPT-3 were maybe toddler-level intelligence, and then systems like GPT-4 are more like smart high-schooler intelligence. And then, in the next couple of years, we're looking at PhD-level intelligence for specific tasks.”
Verdict: The 'couple of years' window runs to the end of 2026. Frontier models already exceed human-PhD baselines on some narrow benchmarks, but whether that meets Murati's 'PhD-level intelligence for specific tasks' bar is a judgement to be made at the deadline.
PendingAGI & SuperintelligenceSaid Jun 4, 2024Deadline Dec 31, 2027
“AGI by 2027 is strikingly plausible. That doesn't require believing in sci-fi; it just requires believing in straight lines on a graph.”
Verdict: Deadline is end of 2027. The essay's own operationalization is an AI that functions as a 'drop-in remote worker'; no such system exists as of late 2026.
PendingCapabilitiesSaid Jun 4, 2024Deadline Dec 31, 2026
“The AGI race has begun. We are building machines that can think and reason. By 2025/26, these machines will outpace college graduates.”
Verdict: Deadline runs to the end of 2026. Frontier models beat college graduates on many benchmarks and exams, but whether they 'outpace' graduates as workers is contested and not yet judgeable.
PendingAGI & SuperintelligenceSaid Jun 4, 2024Deadline Dec 31, 2029
“By the end of the decade, they will be smarter than you or I; we will have superintelligence, in the true sense of the word.”
PendingSafety & RiskSaid Jun 4, 2024Deadline Dec 31, 2028
“As the race to AGI intensifies, the national security state will get involved. The USG will wake from its slumber, and by 27/28 we'll get some form of government AGI project. No startup can handle superintelligence.”
PendingSafety & RiskSaid Jun 2, 2024Deadline Dec 31, 2124
“The chance that AI leads to human extinction is 99.9% within the next hundred years. If we create general superintelligences, I don't see a good outcome long term for humanity... The only way to win this game is not to play it.”
Verdict: Deadline is a century out. Yampolskiy restated the same position (odds of superintelligence killing all humans in the next hundred years 'over 99%, and there's nothing we can do to stop it') in Info-Tech Research Group's 'Digital Disruption' interview in August 2025.
PendingJobs & EconomySaid Jun 2024Deadline Dec 31, 2034
“AI will increase US productivity by only 0.5% and GDP growth by only 0.9% cumulatively over the next decade — a nontrivial but modest effect, and far less than the revolutionary changes some are predicting.”
Verdict: Resolves on measured US total factor productivity and GDP attributable to generative AI over 2024-2034, against Goldman's own far larger ~9% productivity / ~6% GDP estimates.
IncorrectJobs & EconomySaid Jun 2024Deadline Jun 30, 2026
“If you go forward 24 months from now, or some amount of time — I can't exactly predict where it is — it's possible that most developers are not coding.”
Verdict: The 24-month horizon ran out in June 2026 with most professional developers still writing and reviewing code, albeit heavily AI-assisted; Garman was pressed on the expired prediction in a June 2026 Platformer interview and defended continued hiring of junior engineers. (He hedged with 'it's possible,' so the miss is on the substance rather than a firm commitment.)
PendingSelf-DrivingSaid May 31, 2024Deadline Dec 31, 2026
“"We're getting really close. We can do it now pretty regularly with a prototype, but doing it in a cost-effective way is just the progress we're going to need to make. Level 3 autonomy will allow you to go hands and eyes off the road on the highway in a couple years so then your car becomes like an office." Farley expects Ford to offer the feature in 2026, potentially making it the first mass-market brand with Level 3.”
Verdict: Deadline is end of 2026. At CES in January 2026 Ford said eyes-off BlueCruise will arrive only in 2028, making a 2026 launch highly unlikely.
PendingEnergy & Data CentersSaid May 29, 2024Deadline Dec 31, 2030
“Data centers could consume up to 9% of U.S. electricity generation by 2030 — more than double the amount currently used — which could create regional supply challenges.”
Verdict: 2030 is still in the future; EPRI has since raised its own projection by roughly 60% (to 9%–17% by 2030) in its February 2026 update.
CorrectCapabilitiesSaid May 14, 2024Deadline Dec 31, 2024
“AI Overviews will begin rolling out to everyone in the U.S. starting today, with more countries soon. By the end of the year, AI Overviews will come to over a billion people in Google Search.”
Verdict: Hit ahead of schedule: on Alphabet's Q3 2024 earnings call (October 29, 2024) Pichai said AI Overviews was expanding to over 100 new countries and 'will now reach more than 1 billion users on a monthly basis.' Usage kept climbing to 1.5 billion by May 2025 and 2 billion by mid-2025.
“With the help of AI, scientists will likely 'understand most diseases' in the next three to five years — 'a first for humans on the planet.' If you understand the disease, you will find a way to go fix it.”
Verdict: Deadline is the outer end of Bancel's three-to-five-year window (April 2029); resolution will turn on whether AI-driven biology delivers fundamental mechanistic understanding of 'most diseases' by then.
CorrectAGI & SuperintelligenceSaid Apr 9, 2024Deadline Dec 31, 2025
“Dear @elonmusk, $1 million says your latest prediction – that AI will be smarter than any individual human by the end of 2025 – is wrong. Game?”
— Gary Marcus, Cognitive scientist, Professor Emeritus of Psychology and Neural Science at NYU, author of 'Taming Silicon Valley' (Researcher). Gary Marcus on X, 9 April 2024
Verdict: No AI system exceeded the best individual humans across intellectual work by the end of 2025; even Marcus's more general 'we will not see AGI in 2025' prediction was graded correct by third-party reviewers. Musk never accepted the bet (which was raised to $10M by investor Damion Hankejh) despite press inquiries.
PendingJobs & EconomySaid Apr 2024Deadline Dec 31, 2030
“I think by 2030, maybe even sooner, we will have more than one billion software developers on GitHub. Think about that: 10 percent of the world's population will not only control a computer but will also be able to create software just as if they were riding a bicycle.”
Verdict: Resolvable against GitHub's own reported user counts at the end of 2030; GitHub had roughly 100 million developers when the prediction was made. Dohmke stepped down as GitHub CEO in 2025.
“In the next year, you're going to see very large context windows, agents and text action. When they are delivered at scale, it's going to have an impact on the world at a scale that no one understands yet. Much bigger than the horrific impact we've had by social media in my view.”
Verdict: The three technical ingredients did arrive within the year: million-token context windows became standard (Gemini, Claude, Llama 4), and agentic/computer-use and coding agents shipped broadly through 2024-2025. But the claimed societal impact 'much bigger than social media' had not materialized by 2025 — the technology arrived, the world-scale disruption did not.
“Say to your LLM the following: Make me a copy of TikTok, steal all the users, steal all the music, put my preferences in it, produce this program in the next 30 seconds, release it and in one hour, if it's not viral, do something different along the same lines. That's the command... Imagine that each and every human on the planet has their own programmer that actually does what they want... That's all within the next year or two.”
Verdict: By 2026 'vibe coding' platforms (Lovable, Base44, Replit Agent, Claude Code, Codex) do let ordinary users generate working apps — including social-video clones — from a prompt in minutes, so the 'everyone has their own programmer' part broadly arrived. But no system executes the described loop: no 30-second production-grade TikTok replica, no ingesting of a rival's users and music catalog, and no autonomous relaunch-until-viral iteration.
PendingJobs & EconomySaid Mar 28, 2024Deadline Dec 31, 2049
“If one takes a view over the next generation, this could be the biggest thing that has happened in economic history since the Industrial Revolution... This offers the prospect of not replacing some forms of human labor, but almost all forms of human labor.”
Verdict: Generational horizon (~2049); Summers paired the claim with the view that productivity gains would take years to materialize, so nothing is resolvable yet.
PendingJobs & EconomySaid Mar 26, 2024Deadline Dec 31, 2028
“Under 'Network of the Future,' UPS will more than triple the number of automated buildings in its network to 400 facilities by the end of 2028 — closing roughly 200 conventional U.S. sortation facilities and driving $3 billion of costs out of the business by 2028, adding capacity with less dependence on labor.”
Verdict: Target runs through 2028. Progress to date: UPS says it closed 93 buildings and deployed automation in 57 facilities in 2025 alone, with 63 major automation projects underway out of a ~$9 billion automation program.
IncorrectAGI & SuperintelligenceSaid Mar 13, 2024Deadline Dec 31, 2025
“AI will probably be smarter than any single human next year.”
Verdict: No AI system was smarter than the smartest single human by the end of 2025; frontier models remained below top human experts across many domains. Musk himself walked the timeline back in January 2025, shifting to vaguer language about what AI 'will do'.
PendingAGI & SuperintelligenceSaid Mar 13, 2024Deadline Dec 31, 2029
“By 2029, AI is probably smarter than all humans combined.”
Verdict: Deadline is end of 2029. The companion half of the same post — AI smarter than any single human by 2025 — did not come true.
PendingAGI & SuperintelligenceSaid Mar 7, 2024Deadline Mar 7, 2034
“Before we get all those things to work together, and then on top of this, have systems that can learn hierarchical planning, hierarchical representations, systems that can be configured for a lot of different situations at hand, the way the human brain can, all of this is going to take at least a decade and probably much more... So it's not just around the corner.”
Verdict: A lower-bound claim: human-level AI arrives no sooner than ~2034. It is falsified only if a system with human-level, configurable, hierarchically-planning intelligence appears before then.
PendingCapabilitiesSaid Mar 1, 2024Deadline Mar 1, 2029
“Within five years, AI will pass every test a human can take.”
— Jensen Huang, CEO, NVIDIA (Tech CEO). Stanford Institute for Economic Policy Research talk (March 2024)
Verdict: AI already matches or beats humans on many standardized exams; comprehensive coverage of 'every test' remains open.
PendingJobs & EconomySaid Mar 2024Deadline Dec 31, 2027
“Asked whether his 2017 prediction that AI would displace 50% of human jobs by 2027 still holds: "It's actually uncannily accurate. People have criticized me for being too aggressive in 2017, 2018, 2019, and I was a little nervous."”
Verdict: Resolves at the end of 2027 against measured job displacement; as of 2026 aggregate labor-market data show nothing approaching 50% displacement.
CorrectEnergy & Data CentersSaid Feb 29, 2024Deadline Dec 31, 2025
“The chip shortage is behind us, but the next shortage will be voltage step-down transformers, and after that electricity: "They won't be able to find enough electricity to run all the chips. I think next year, you'll see they just can't find enough electricity to run all the chips."”
Verdict: Power, not chips, became the binding constraint in 2025: Microsoft CEO Satya Nadella said in November 2025 that Microsoft had GPUs "sitting in inventory" it could not plug in because it lacked "warm shells"/power, and transformer and grid-interconnection lead times stretched to multiple years. There was no consumer-facing electricity shortage, but the specific prediction — not enough electricity to run all the chips in 2025 — was borne out.
PendingEnergy & Data CentersSaid Feb 12, 2024Deadline Dec 31, 2029
“There's about a trillion dollars' worth of installed base of data centers; over the next four to five years another trillion dollars' worth of data center infrastructure will be built, so we'll have $2 trillion of data centers powering software around the world.”
Verdict: Deadline is the end of 2029 (taking the outer bound of Huang's 'four to five years'). Huang made the remark while arguing that compute efficiency gains meant Sam Altman's reported $7 trillion ask was unnecessary.
PendingCapabilitiesSaid Jan 2024Deadline Dec 31, 2030
“The convergence of hardware and software could drive AI training costs down by 75% at an annual rate through 2030 — accelerated compute hardware reducing relative-compute-unit production costs by 53% per year while algorithmic enhancements lower training costs a further 47% per year.”
Verdict: Runs through 2030; ARK has continued to cite roughly 75%-per-year training-cost declines in later Big Ideas editions.
CorrectEnergy & Data CentersSaid Jan 30, 2024Deadline Dec 31, 2024
“'Based on the strong customer pull and expanded engagements, we now expect Data Center GPU revenue to grow sequentially in the first quarter and exceed $3.5 billion in 2024' — up from the earlier $2 billion forecast, after MI300 revenue exceeded $400 million in the quarter and became the fastest revenue ramp of any product in AMD history.”
Verdict: AMD's February 4, 2025 results announcement stated it delivered 'more than $5 billion of AMD Instinct accelerator revenue' in 2024, comfortably clearing the $3.5 billion target Su set in January 2024.
Partially correctSafety & RiskSaid Jan 23, 2024Deadline Jul 1, 2024
“Daily bad-actor AI activity will escalate by mid-2024 — with AI-manipulated content spreading across online communities on a near-daily basis — increasing the threat that it could affect election results in the more than 50 countries voting this year, including the U.S.”
Verdict: AI-generated and AI-manipulated content did become an essentially daily feature of online political discourse by mid-2024, matching the paper's directional forecast, but the accompanying threat to election results did not materialize: post-election reviews found deepfakes relatively scarce, non-AI 'cheap fakes' far more common, and no election outcome demonstrably changed.
CorrectAGI & SuperintelligenceSaid Jan 18, 2024Deadline Dec 31, 2024
“We're building massive compute infrastructure to support our future roadmap, including 350,000 H100s by the end of this year — and overall almost 600,000 H100 equivalents of compute if you include other GPUs. Our long term vision is to build general intelligence, open source it responsibly, and make it widely available.”
Verdict: Meta does not publish audited GPU counts, but its buildout tracked the target: analysts put Meta among Nvidia's two largest H100 buyers in 2024, and Zuckerberg's January 2025 target of 'more than 1.3 million GPUs' by end-2025 was reported as roughly a doubling of the fleet, consistent with ~600,000 H100-equivalents at the end of 2024.
IncorrectJobs & EconomySaid Jan 6, 2024Deadline Dec 31, 2024
“Most contrarian belief for 2024: the enterprise value of OpenAI goes down.”
Verdict: OpenAI's valuation rose sharply in 2024, from roughly $86B in the early-2024 tender to about $157B in the October 2024 round; the All-In hosts scored it themselves, noting it 'roughly doubled.'
PendingSafety & RiskSaid Jan 2024Deadline Dec 31, 2047
“Surveying 2,778 AI researchers: the median respondent gives a 5% probability that advanced AI causes human extinction or similarly permanent and severe disempowerment of the human species, and the aggregate forecast gives a 50% chance that high-level machine intelligence — machines able to accomplish every task better and more cheaply than human workers — arrives by 2047.”
Verdict: The HLMI component resolves in 2047; the survey's aggregate timeline had already shortened by roughly 13 years versus the same survey in 2022.
IncorrectSafety & RiskSaid Dec 27, 2023Deadline Nov 5, 2024
“I expect a tsunami of misinformation [in the 2024 U.S. election]. I can't prove that. I hope to be proven wrong. But the ingredients are there, and I am completely terrified... You could see a political candidate like President Biden being rushed to a hospital. You could see a candidate saying things that he or she never actually said.”
Verdict: The feared AI 'tsunami' did not materialize: post-election analyses (Kapoor & Narayanan's review of 78 election deepfakes for the Knight Institute, and the Harvard Ash Center's 'apocalypse that wasn't') found AI-generated deception had limited reach, with non-AI 'cheap fakes' documented roughly seven times more often. AI content was everywhere, but no deepfake of the kind Etzioni described disrupted the vote.
IncorrectJobs & EconomySaid Dec 22, 2023Deadline Dec 31, 2024
“Deflation will be a key theme in 2024 — driven by technology and the unwinding of supply-chain and inventory excesses — prompting the Federal Reserve to cut rates aggressively.”
Verdict: US consumer prices never turned negative in 2024 — headline CPI inflation stayed positive all year and ended around 2.9% year over year — and the Fed cut only 100 basis points, beginning in September, rather than cutting aggressively against deflation.
CorrectSafety & RiskSaid Dec 20, 2023Deadline Nov 5, 2024
“Although AI will be used in elections [in 2024], it will not be 'at the scale yet that everyone imagines.' Rather, it will be the narrative of what havoc AI could have that will have the bigger impact.”
Verdict: AI was used throughout the 2024 cycle but nowhere near the imagined scale: Kapoor and Narayanan catalogued only 78 election deepfakes worldwide, the News Literacy Project found non-AI 'cheap fakes' about seven times more common, and analysts concluded the anticipated AI election apocalypse never arrived — while the narrative about deepfakes drew far more attention than the deepfakes themselves, as she forecast.
PendingEnergy & Data CentersSaid Dec 6, 2023Deadline Dec 31, 2027
“The market for data center AI accelerators will grow to approximately $400 billion in 2027 — nearly triple AMD's own $150 billion estimate from a year earlier.”
Verdict: Resolves against 2027 data center AI accelerator market size. Su has since revised the number upward repeatedly (to $500 billion for 2028 in October 2024).
PendingJobs & EconomySaid Dec 2023Deadline Dec 31, 2048
“AI should be hugely deflationary over twenty five years. Capital should be scarce for a while, current measures of GDP and the economy will be less relevant but goods and services should be in great abundance.”
Verdict: Long-horizon macro claim; resolves against sustained price-level trends through 2048. Khosla has repeated the 'hugely deflationary' thesis in 2024-2026 interviews.
PendingJobs & EconomySaid Nov 9, 2023Deadline Nov 9, 2026
“In the good old days it took 500 artists five years to make a world-class animated movie. I think it won't take 10 percent of that. Literally, I don't think it will take 10 percent of that three years out from now.”
Verdict: Deadline falls in November 2026. Resolution will turn on whether world-class animated features can be made with roughly a tenth of the traditional 500-artist, five-year crew effort.
IncorrectSelf-DrivingSaid Nov 2023Deadline Dec 31, 2024
“Version 12 (end-to-end neural nets) will far exceed human safety even when unsupervised.”
Verdict: FSD v12 shipped in 2024 but remained a supervised Level 2 system that never operated unsupervised and did not demonstrate far-better-than-human safety. Musk went on to make the same claim about v14 and v15.
PendingAGI & SuperintelligenceSaid Oct 26, 2023Deadline Dec 31, 2028
“I think there's a 50% chance that we have AGI by 2028. Now, it's just a 50% chance. I'm sure what's going to happen is we're going to get to 2029 and someone's going to say, 'Shane, you were wrong.' Come on, I said 50% chance.”
Verdict: Deadline is end of 2028. Legg has restated this 50%-by-2028 estimate consistently since first making it in the 2000s, and as of 2026 he still frames 'minimal AGI' as roughly two years out.
Partially correctCapabilitiesSaid Oct 21, 2023Deadline Aug 7, 2025
“Many good people at OpenAI believe GPT-5 will be a big leap, but I believe current generative AI has reached a ceiling — GPT-5 will not be significantly better than GPT-4.”
Verdict: GPT-5 launched on 7 August 2025 to a widely criticised reception and was broadly judged an incremental rather than transformative step, which commentators read as vindicating Gates. But it did post large benchmark gains over GPT-4 — driven by the reasoning/chain-of-thought advances Gates's 'plateau' framing did not anticipate — so the call was directionally right rather than fully right.
PendingAGI & SuperintelligenceSaid Oct 18, 2023Deadline Oct 18, 2026
“The dangers of superintelligence are even worse than I projected five years ago: it'll only be 2-3 years "before we get outsmarted," and "we should expect the machines to take control."”
Verdict: The outer edge of Tegmark's 2–3 year window falls in October 2026. Reported second-hand by attendee Eric Topol; wording follows his direct quotations of the talk.
PendingAGI & SuperintelligenceSaid Oct 8, 2023Deadline Dec 31, 2028
“I think in five years' time it [AI] may well be able to reason better than us.”
Verdict: Hinton made the remark to Scott Pelley after testing GPT-4 with a reasoning riddle; the five-year horizon runs to the end of 2028.
IncorrectCapabilitiesSaid Sep 7, 2023Deadline Mar 31, 2024
“It is a solvable problem. I would bet you any sum of money you can get the hallucinations right down into the line of human-expert rate within months. So I'm not really that worried about that problem overall. (Gary Marcus publicly offered $100,000 on the other side; Hoffman never took the bet.)”
Verdict: Hallucination rates did not fall to human-expert levels within months; by mid-2024 leading models still hallucinated at rates far above domain experts, and hallucination remained a headline reliability problem well beyond early 2024. Hoffman also declined Marcus's $100,000 offer to formalize the bet.
CorrectCapabilitiesSaid Aug 8, 2023Deadline Aug 31, 2026
“In terms of someone looks at the model and even if you talk to it for an hour or so, it's basically like a generally well educated human — that could be not very far away at all. I think that could happen in two or three years.”
Verdict: In the February 13, 2026 follow-up interview, Dwarkesh Patel quoted the prediction back to Amodei and told him 'You were right,' while adding that he had expected such a system to also automate large parts of white-collar work. Commentators agree the spirit held — an hour-long conversation with a 2026 frontier model matches or exceeds a well-educated generalist — though models remain trivially identifiable as AI in a literal imitation test.
CorrectJobs & EconomySaid Aug 1, 2023Deadline Dec 31, 2025
“AI investment could approach $100 billion in the U.S. and $200 billion globally by 2025; over the longer term, AI-related investment could peak as high as 2.5-4% of GDP in the U.S. and 1.5-2.5% in other AI-leading economies.”
Verdict: The $200bn global threshold was comfortably reached by 2025 — capital spending by the largest US hyperscalers alone ran to several hundred billion dollars in 2025 — so the directional call was right, though it proved conservative on magnitude; the longer-run share-of-GDP leg remains open.
PendingJobs & EconomySaid Aug 2023Deadline Dec 31, 2028
“There will be no [human] programmers in five years.”
Verdict: Five-year horizon from August 2023 resolves at the end of 2028; as of 2026 software development employment remains substantial, so the claim is heavily on track to fail but is not yet due.
Verdict: Deadline is mid-2027. Context: OpenAI disbanded the Superalignment team in May 2024 after Leike and Ilya Sutskever resigned, with Leike saying safety had taken a back seat; the 20%-compute commitment was reportedly not honored.
“A straightforward extrapolation of today's systems to those we expect to see in two to three years suggests a substantial risk that AI systems will be able to fill in all the missing pieces [of bioweapons production], enabling many more actors to carry out large-scale biological attacks.”
Verdict: Directionally borne out on capability trend but not on the strongest version of the claim: by May 2025 Anthropic activated ASL-3 safeguards for Claude Opus 4 because it could not rule out CBRN-3 (meaningful uplift to novices in bioweapons acquisition planning, measured at ~2.5x), and OpenAI has treated recent agentic models as 'High' biological capability. However, as of the mid-2026 deadline no frontier model has been publicly assessed as able to fill in 'all the missing pieces' (state-level/CBRN-4 thresholds remain uncrossed), and no AI-enabled large-scale biological attack has occurred — partly consistent with Amodei's own conditional that guardrails and mitigations be put in place.
IncorrectCapabilitiesSaid Jul 14, 2023Deadline Dec 31, 2025
“A Modern Turing Test for 'artificial capable intelligence': tell an AI to 'Go make $1 million on a retail web platform in a few months with just a $100,000 investment' — researching products, interfacing with manufacturers and logistics, negotiating contracts, running marketing — tying together complex real-world goals with minimal oversight. 'Something like this could be as little as two years away. Many of the ingredients are in place.'”
Verdict: No AI system has autonomously turned $100,000 into $1 million with minimal human oversight; the most prominent real-world test of an AI running a business — Anthropic's Project Vend, in which Claude managed a small shop/vending business — lost money in both phases through 2025. Suleyman himself, in a December 2025 interview, was framed as expecting agents to only begin passing the Modern Turing Test roughly two years later (~2027), confirming the two-year horizon set in 2023 was missed.
IncorrectSafety & RiskSaid Jun 26, 2023Deadline Nov 5, 2024
“The 2024 elections are going to be a mess because social media is not protecting us from false generated AI. They're working on it, but they haven't solved it yet.”
Verdict: Platforms did cut trust-and-safety staff as Schmidt warned, but the 2024 US election was not the AI-driven 'mess' he predicted: retrospectives from the Harvard Ash Center and the Knight First Amendment Institute concluded the feared AI misinformation wave had marginal effect on the vote, with non-AI fakes far more prevalent.
PendingAGI & SuperintelligenceSaid Jun 2023Deadline Dec 31, 2043
“My current estimate places a 95% confidence interval for the time horizon of superhuman intelligence at 5 to 20 years.”
Verdict: Russell said ChatGPT's advances exceeded his expectations and shortened his timeline, while maintaining that further scaling alone is unlikely to produce AGI.
PendingAGI & SuperintelligenceSaid May 22, 2023Deadline May 22, 2033
“It's conceivable that within the next ten years, AI systems will exceed expert skill level in most domains.”
— Altman, Brockman & Sutskever, OpenAI leadership (Tech CEO). Governance of Superintelligence (OpenAI blog, May 2023)
PendingEnergy & Data CentersSaid May 10, 2023Deadline Dec 31, 2028
“Microsoft has agreed to purchase electricity from Helion's first fusion power plant, scheduled for deployment in 2028: the plant "is expected to be online by 2028 and will target power generation of 50 MW or greater after a 1-year ramp up period."”
Verdict: Site work on Orion in Chelan County, Washington began in July 2025 and Helion reaffirmed the 2028 date, but in 2026 the company removed the "producing electricity by 2028" language from its website in favor of "begin initial operations in 2028," and a Helion executive conceded at the 2026 FusionX conference that 50 MW full operation will not happen in 2028 (Axios Pro, 8 Sept 2026). Formally unresolved until the end of 2028.
PendingAGI & SuperintelligenceSaid May 3, 2023Deadline Dec 31, 2043
“I now predict 5 to 20 years [until digital intelligence overtakes us] but without much confidence. We live in very uncertain times. It's possible that I am totally wrong about digital intelligence overtaking us. Nobody really knows which is why we should worry now.”
— Geoffrey Hinton, Turing Award and Nobel laureate; 'Godfather of AI'; professor emeritus, University of Toronto (former Google VP) (Researcher). X post by @geoffreyhinton, May 3, 2023
Verdict: Posted days after Hinton left Google; it marked a sharp shortening of his previous 30–50 year estimate for machines becoming smarter than humans.
PendingHealthcareSaid May 3, 2023Deadline Dec 31, 2029
“Within 5 to 6 years, the FDA will approve a primary care app qualified to practice medicine like your primary care physician.”
Verdict: Resolves if the FDA authorizes a software product cleared to deliver general primary care diagnosis and treatment autonomously; to date FDA authorizations of autonomous AI remain single-condition devices rather than general practice.
PendingSafety & RiskSaid May 2023Deadline May 1, 2053
“There is a 10–20% chance that AI leads to human extinction within the next three decades.”
— Geoffrey Hinton, Turing Award winner; 'godfather of deep learning' (Researcher). Interviews following his departure from Google (2023)
Verdict: A probabilistic risk estimate; tracked for the record.
PendingJobs & EconomySaid May 1, 2023Deadline May 1, 2028
“Hiring in back-office functions such as human resources will be suspended or slowed; of the roughly 26,000 non-customer-facing roles at IBM, "I could easily see 30% of that getting replaced by AI and automation over a five-year period" — roughly 7,800 jobs.”
Verdict: The five-year window runs to May 2028. IBM has since said AI agents took over work previously done by a few hundred HR staff while the company hired in engineering and sales, so the 30%/7,800 figure is not yet settled.
PendingSelf-DrivingSaid Apr 2023Deadline Dec 31, 2027
“ARK's price target for Tesla is $2,000 per share in 2027, with bear and bull cases valued between ~$1,400 and $2,500 per share; Tesla's prospective robotaxi business line is a key driver, contributing 67% of expected enterprise value and 64% of expected EBITDA in 2027.”
Verdict: Resolves at the end of 2027; ARK has since rolled the model forward to a $2,600 2029 target, and Tesla has yet to operate a material driverless ride-hail business.
IncorrectSelf-DrivingSaid Apr 13, 2023Deadline Dec 31, 2025
“Cruise is on track — if not ahead of — its goal to reach $1 billion in revenue by 2025, as it scales up new markets and vehicle volume.”
Verdict: Cruise never came close to $1 billion in revenue: it was losing hundreds of millions per quarter, had its California permits suspended in late 2023, Vogt resigned that November, and GM ended funding for the robotaxi business in December 2024, so 2025 revenue was effectively zero.
CorrectCapabilitiesSaid Apr 9, 2023Deadline Dec 31, 2024
“AGI really could disrupt the world. But GPT-5 is not going to do any of that; it's not going to be smart enough to revolutionize fusion; it's not going to be smart enough to revolutionize nanotech, and starships most definitely aren't going to be here in 2024.”
Verdict: GPT-5 did not even ship until August 2025, and through the end of 2024 there was no AI-driven revolution in fusion or nanotech and no interstellar starships. Marcus's deflationary call was borne out.
Partially correctJobs & EconomySaid Apr 4, 2023Deadline Jan 31, 2026
“"By the end of Fiscal Year 2026, Walmart believes roughly 65% of stores will be serviced by automation."”
Verdict: On the Q4 FY2026 earnings call (Feb. 19, 2026), CFO John David Rainey said approximately 60% of U.S. stores were receiving some freight from automated distribution centers — short of the 65% goal, and on a looser 'some freight' measure than 'serviced by automation.'
Partially correctJobs & EconomySaid Apr 4, 2023Deadline Jan 31, 2026
“"By the end of Fiscal Year 2026... approximately 55% of the fulfillment center volume will move through automated facilities, and unit cost averages could improve by approximately 20%."”
Verdict: At the end of FY2026, Walmart reported roughly 50% of U.S. e-commerce fulfillment center volume was automated — short of the 55% target — though the company says automation is delivering better unit productivity and lower cost to serve; the 20% unit-cost improvement has not been separately disclosed.
Verdict: By April 2025 no feature film existed that audiences could not tell was AI-generated. The first AI-generated features to reach release — such as the Turkish documentary Post Truth (July 2025) and later festival entries like Hell Grind and Dreams of Violets (2026) — arrived after the deadline and were all overtly, recognizably AI-made rather than indistinguishable from conventional filmmaking.
PendingJobs & EconomySaid Mar 31, 2023Deadline Dec 31, 2029
“That's not to say that artificial intelligence won't have huge economic impacts. But history suggests that they won't come quickly. ChatGPT and whatever follows are probably an economic story for the 2030s, not for the next few years.”
Verdict: Resolves at the end of the 2020s: the test is whether large measured macroeconomic effects of generative AI arrive before 2030 or only in the 2030s.
PendingJobs & EconomySaid Mar 26, 2023Deadline Mar 26, 2033
“Generative AI could raise global GDP by 7% — almost $7 trillion — over a 10-year period.”
— Goldman Sachs Research, Investment bank research division (Wall Street). Goldman Sachs Global Economics Analyst (March 2023)
IncorrectSafety & RiskSaid Mar 9, 2023Deadline Nov 5, 2024
“2024 will be the last human election — from here on, humans will be figureheads and whoever has the greatest compute and the best AI will win.”
Verdict: The 2024 US election was contested and decided by human candidates and campaigns; subsequent research found AI-generated content played a marginal role and that compute advantage did not determine outcomes, and conventional human elections have continued since.
PendingCapabilitiesSaid Jan 31, 2023Deadline Dec 31, 2030
“AI training cost declines continued at an annual rate of 70%, the cost to train a large language model to GPT-3 level performance collapsing from $4.6 million in 2020 to $450,000 in 2022. We expect cost declines to continue at a 70% rate through 2030.”
Verdict: Runs through 2030; ARK itself raised the assumed decline rate to ~75%/year in Big Ideas 2024.
PendingJobs & EconomySaid Jan 31, 2023Deadline Dec 31, 2030
“AI should increase the productivity of knowledge workers more than 4-fold by 2030. At 100% adoption, AI could increase global labor productivity ~$200 trillion, dwarfing the ~$32 trillion in total knowledge worker salaries.”
Verdict: Resolves at the end of 2030; measured productivity gains to date are far below a 4x trajectory.
PendingJobs & EconomySaid Jan 31, 2023Deadline Dec 31, 2030
“Based on a 70% annualized drop in training costs and feedback loops, AI coding assistants like Copilot could increase the output of software engineers ~10-fold by 2030.”
Verdict: Resolves at the end of 2030; ARK's own 2023 report put the then-current coding-assistant gain at roughly 2x.
PendingJobs & EconomySaid Jan 31, 2023Deadline Dec 31, 2030
“The market value of disruptive innovation platforms could scale 40% at an annual rate during this business cycle, from $13 trillion today to $200 trillion by 2030 — catalyzed by breakthroughs in artificial intelligence, disruptive innovation could account for the majority of global equity market capitalization.”
Verdict: Resolves at the end of 2030; ARK restated the same idea in Big Ideas 2024 as ~$19 trillion growing to roughly $220 trillion by 2030.
PendingCapabilitiesSaid Jan 2023Deadline Jan 30, 2029
“No AI will be able to get A's on 5 out of 6 of my [labor economics] exams by January of 2029. (By January 30, 2029, an AI selected by Matthew Barnett takes Caplan's six most recent midterms, graded on his students' scale, with an A- counting as an A; $500 at stake.)”
Verdict: Still open until January 30, 2029; Caplan publicly wobbled in March/April 2023 after GPT-4 scored 73/100 (an A) on one of his midterms, writing that he was not conceding but gave himself only a 10–15% chance of winning.
CorrectCapabilitiesSaid Dec 25, 2022Deadline Dec 31, 2023
“GPT-4 by itself won't be a general purpose artificial general intelligence capable of taking on arbitrary tasks. Without external aids it won't be able to drive a car reliably; it won't be able to reliably guide a robot like Optimus to be anything like as versatile as Rosie the Robot. It will remain a turbocharged pastiche generator; reasoning about the physical, psychological and mathematical world will still be unreliable; and 'alignment' will continue to be a critical, unsolved problem.”
Verdict: GPT-4 shipped in March 2023 and was not a general-purpose AGI: it could not drive a car or control a humanoid robot, hallucinations and reasoning errors persisted, and alignment remained unsolved. Marcus's seven GPT-4 predictions were widely regarded — including by later critics tracking his record — as having held, which is why he reissued the same text for GPT-5 in 2025.
IncorrectCapabilitiesSaid Oct 31, 2022Deadline Dec 31, 2025
“Domain experts and superforecasters assigned only an 8.6% and 2.3% probability, respectively, to an AI system winning a gold medal at the International Mathematical Olympiad on or before 2025; their median forecasts put it after 2030.”
Verdict: In July 2025 both Google DeepMind and OpenAI reported gold-medal-level performance at IMO 2025 — five years earlier than the median expert forecast and ten years earlier than the median superforecaster forecast, and an outcome both panels had rated very unlikely by 2025.
“Tesla will be able to produce Optimus humanoid robots in the millions and sell them for under $20,000 — less than a third of the price of a Model Y — and expects to be ready to take orders for the robot in three to five years.”
Verdict: Deadline runs to September 2027. As of September 2026 Tesla has not opened orders for Optimus, has not disclosed a consumer price, and volume production at Fremont had not yet begun.
IncorrectEnergy & Data CentersSaid Aug 13, 2022Deadline Dec 31, 2025
“Being humble about it, I predict that somewhere around 2025 we'll have this machine [Copernicus] huffing and puffing and producing the data we hoped for.”
Verdict: Copernicus was not operating or producing data by the end of 2025; TAE's 2025 work centered on its reconfigured 'Norm' machine, and its 2026 communications still describe Copernicus as a future device expected to demonstrate net-energy viability later this decade.
PendingJobs & EconomySaid Jun 22, 2022Deadline Dec 31, 2030
“As companies generate profound productivity gains from artificial intelligence, we estimate that software spend per year could grow at a 42% compound annual rate from just under $1 trillion today to $14 trillion in 2030; propelled by AI software, total IT spend could increase 20% at an annual rate and surpass $20 trillion, four times the ~$5 trillion consensus estimate.”
Verdict: Resolves at the end of 2030; ARK has since trimmed the figure to roughly $13 trillion of 2030 software spend in later research.
IncorrectSelf-DrivingSaid Jun 14, 2022Deadline Dec 31, 2024
“Stellantis has chosen Valeo's third-generation SCALA 3 LiDAR to equip multiple models across its brands from 2024, enabling those vehicles to be certified for Level 3 automation — "allowing drivers to safely take their hands off the steering wheel and their eyes off the road."”
Verdict: No Stellantis vehicle offered Level 3 hands-off, eyes-off driving in 2024. Stellantis later showed an STLA AutoDrive Level 3 system but held it back from customer vehicles, and none was on sale as of 2026.
PendingAGI & SuperintelligenceSaid May 31, 2022Deadline Dec 31, 2029
“$100,000 says AGI will not arrive by 2029: if any single AI system satisfies at least three of my five criteria (e.g., watching a movie and reliably explaining plot and motivations; reading a novel and answering questions about it; working as a competent cook in an arbitrary kitchen; writing bug-free code of more than 10,000 lines from natural-language spec; deriving mathematical proofs from natural language) by then, Musk wins; otherwise I do.”
Verdict: Musk never accepted; others (including Vivek Wadhwa, who matched it) raised the wager to $500,000, and Kevin Kelly offered to host it at Long Bets. Metaculus opened markets on several of the five criteria. Resolution due after 2029.
IncorrectCapabilitiesSaid Apr 7, 2022Deadline Dec 31, 2023
“I think we have a shot at being in production for version 1 of Optimus hopefully next year [2023]... It will be able to do basically anything humans don't want to do.”
Verdict: Optimus was nowhere near production in 2023 — Tesla only demoed prototypes (Gen 1/Gen 2). Tesla was still installing its first Optimus production lines at Fremont in 2026, roughly three years after the promised production start.
IncorrectEnergy & Data CentersSaid Apr 4, 2022Deadline Dec 31, 2024
“The world's first commercial fusion energy generator marches towards 2024 operation!”
Verdict: Helion finished assembling Polaris in late 2024 and began initial operations, but it was a prototype, not a commercial generator, and produced no electricity in 2024. Helion's first commercial plant, Orion, only broke ground in July 2025 and targets 2028.
Partially correctCapabilitiesSaid Feb 2022Deadline Dec 31, 2025
“I'd put 4% on an AI built before the IMO solving the single hardest problem at the 2022, 2023, 2024 or 2025 IMO; maybe 8% on 'gets gold' instead (under the time controls etc. of the IMO Grand Challenge).”
Verdict: In July 2025 both Google DeepMind (Gemini Deep Think, 35/42) and OpenAI reported gold-medal-level performance at the IMO under human-like time controls, so the outcome Christiano put at ~8% occurred. However, the hardest problem (P6) was not solved, consistent with his separate <4% forecast, which is why commentators called the resolution not fully clear-cut.
CorrectCapabilitiesSaid Feb 2022Deadline Dec 31, 2025
“Paul put 8% probability on 'AI built before the 2025 IMO reaches gold level on it' and I put 'at least 16%'... I'll stand by a >16% probability of the technical capability existing by end of 2025.”
Verdict: In July 2025, models from Google DeepMind and OpenAI achieved gold-medal-level scores (35/42, five of six problems) at the IMO under human-equivalent time controls, vindicating the higher-probability side of the disagreement roughly three and a half years after it was registered.
IncorrectSelf-DrivingSaid Jan 5, 2022Deadline Dec 31, 2025
“General Motors will sell personal autonomous vehicles by "the middle of the decade": "We are looking further down the road at opportunities to extend fully autonomous vehicle technology to personal transportation... we are working to be the fastest to market with a retail personal autonomous vehicle."”
Verdict: GM sold no personal autonomous vehicle by mid-decade. It stopped funding Cruise robotaxi development in December 2024 and in October 2025 said eyes-off driving would only reach market in 2028 on the Cadillac Escalade IQ.
PendingEnergy & Data CentersSaid Nov 17, 2021Deadline Dec 31, 2031
“There's a very good shot to get there [commercial fusion] within less than ten years.”
Verdict: Deadline is end-2031; as of 2026 no private company has yet demonstrated net electricity from fusion, let alone commercial operation.
IncorrectEnergy & Data CentersSaid Nov 5, 2021Deadline Dec 31, 2024
“Polaris is expected to be the first fusion device capable of demonstrating net electricity production. Helion plans to reach this milestone in 2024... 'This funding ensures that Helion will be the first organization to generate electricity from fusion.'”
Verdict: Polaris was only completed in late 2024 and demonstrated no net electricity that year; the target slipped repeatedly, and by 2026 Helion had restated the Polaris goal from 'net electricity' to simply 'demonstrating electricity from fusion.'
IncorrectSelf-DrivingSaid Apr 22, 2019Deadline Dec 31, 2020
“Tesla will have one million robotaxis on the road next year (2020).”
— Elon Musk, CEO, Tesla (Tech CEO). Tesla Autonomy Day (April 2019)
Verdict: Zero Tesla robotaxis operated in 2020. Tesla's first limited, supervised robotaxi pilot did not launch until 2025 — with a small fleet, not a million.
PendingJobs & EconomySaid Jan 13, 2019Deadline Dec 31, 2034
“"Chauffeurs, truck drivers, anyone who does driving for a living — their jobs will be disrupted more in the 15-to-20-year time frame. And many jobs that seem a little bit complex — chef, waiter — a lot of things will become automated... all together in 15 years, that's going to displace about 40 percent of the jobs in the world." (Asked if 40% of jobs will be displaced, Lee clarified: "I would say displaceable.")”
Verdict: Resolves at end-2034 on whether AI is technically capable of displacing roughly 40% of global jobs; note Lee's separate, earlier prediction that AI would displace 50% of jobs by 2027 is tracked separately (kaifulee-2024-fifty-percent-jobs-2027).
IncorrectHealthcareSaid Apr 7, 2017Deadline Dec 31, 2022
“The role of the radiologist will be obsolete in five years. There's no reason a human should be doing it.”
Verdict: Radiology was not obsolete by 2022: AI imaging tools were deployed only as narrow adjuncts requiring radiologist sign-off, while the US faced a radiologist shortage with record job openings and expanded residency slots. Not a single radiology role was eliminated by autonomous AI within the five-year window.
PendingAGI & SuperintelligenceSaid Feb 27, 2017Deadline Feb 27, 2047
“Within 30 years, superintelligence — AI with an IQ of 10,000 — will become a reality.”
— Masayoshi Son, CEO, SoftBank (Investor). Mobile World Congress keynote (February 2017)
IncorrectHealthcareSaid Oct 2016Deadline Oct 27, 2021
“People should stop training radiologists now. It's just completely obvious that within five years deep learning is going to do better than radiologists.”
— Geoffrey Hinton, Turing Award winner; 'godfather of deep learning' (Researcher). Machine Learning and the Market for Intelligence conference (2016)
Verdict: By 2021 AI had become a useful assistive tool in radiology, but radiologist demand grew — the US faced a radiologist shortage. Hinton later acknowledged the timeline was wrong.
IncorrectSelf-DrivingSaid Oct 19, 2016Deadline Dec 31, 2017
“A Tesla will drive itself from Los Angeles to New York, fully autonomously, by the end of 2017.”
— Elon Musk, CEO, Tesla (Tech CEO). Tesla Master Plan Part Deux press call (October 2016)
Verdict: The coast-to-coast autonomous demo never happened — not in 2017, and not in the years after. Tesla's system still requires driver supervision.
IncorrectHealthcareSaid 2016Deadline Dec 31, 2017
“Watson will touch one billion consumers by the end of 2017, transforming healthcare along the way.”
— Ginni Rometty, CEO, IBM (Tech CEO). IBM keynotes & investor communications (2016)
Verdict: Watson Health fell far short of its promises: flagship partnerships like MD Anderson were cancelled, and IBM sold off Watson Health assets in 2022.
IncorrectSafety & RiskSaid Nov 17, 2014Deadline Nov 17, 2024
“The pace of progress in artificial intelligence is incredibly fast... The risk of something seriously dangerous happening is in the five year timeframe. 10 years at most.”
Verdict: Both the five-year (2019) and the ten-year outer bound (November 2024) passed with no catastrophic AI event of the kind Musk described; the phrasing was hedged ('the risk of something seriously dangerous'), but no reasonable reading of that threshold was met by the deadline.
“47% of total US employment is at high risk of computerisation, perhaps over the next decade or two.”
— Frey & Osborne, Oxford University economists (Economist). The Future of Employment (Oxford Martin School, 2013)
Verdict: A decade in, US employment reached record highs even as automation advanced — but generative AI has reopened the question for knowledge work.
PendingAGI & SuperintelligenceSaid Sep 2005Deadline Dec 31, 2029
“AI will pass a valid Turing test, achieving human-level intelligence, by 2029.”
— Ray Kurzweil, Inventor & futurist, Google (Thought Leader). The Singularity Is Near (2005)
Verdict: Conversational tests are increasingly passed, but Kurzweil's own definition requires expert-level performance across the full range of human cognition.
PendingAGI & SuperintelligenceSaid Sep 2005Deadline Dec 31, 2045
“The Singularity — a profound merger of human and machine intelligence — will occur around 2045.”
— Ray Kurzweil, Inventor & futurist, Google (Thought Leader). The Singularity Is Near (2005)
PendingAGI & SuperintelligenceSaid 2002Deadline Dec 31, 2029
“By 2029 no computer — or 'machine intelligence' — will have passed the Turing Test.”
Verdict: The inaugural Long Bets wager, $20,000 against Ray Kurzweil, remains open until the end of 2029 under detailed agreed rules (multi-hour adversarial interviews with human judges and human foils); neither side has conceded, and claimed 'passes' such as Eugene Goostman in 2014 did not meet the bet's terms.
IncorrectAGI & SuperintelligenceSaid Mar 1993Deadline Dec 31, 2023
“Within thirty years, we will have the technological means to create superhuman intelligence. Shortly after, the human era will be ended.”
Verdict: 2023 saw remarkable AI progress, but no superhuman general intelligence and no end of the human era. Frontier models remained clearly below expert human ability in most domains.
IncorrectAGI & SuperintelligenceSaid Nov 1970Deadline Dec 31, 1978
“In from three to eight years we will have a machine with the general intelligence of an average human being.”
— Marvin Minsky, Co-founder, MIT AI Lab (Researcher). Life magazine interview (1970)
Verdict: No machine approached human-level general intelligence by 1978. Funding cuts and the first AI winter followed within a few years.
IncorrectAGI & SuperintelligenceSaid 1965Deadline Dec 31, 1985
“Machines will be capable, within twenty years, of doing any work a man can do.”
— Herbert A. Simon, AI pioneer; Nobel laureate & Turing Award winner (Researcher). The Shape of Automation for Men and Management (1965)
Verdict: By 1985, AI was in the middle of an 'AI winter.' Machines could not perform most human work then, and still cannot perform all of it today.
Partially correctCapabilitiesSaid Oct 1950Deadline Dec 31, 2000
“By the year 2000, machines will be able to fool 30% of human judges during a five-minute text conversation.”
— Alan Turing, Founder of computer science (Researcher). Computing Machinery and Intelligence, Mind (1950)
Verdict: Not achieved by 2000, but modern language models now routinely pass five-minute conversational tests: controlled 2024 studies found GPT-4 judged human well above Turing's 30% threshold. Right about the destination, ~25 years off on the date.