“Stocks are not up or down because of jobs or consumer sentiment. They are going straight up because they have been going straight up. On a two letter thesis that everyone thinks they understand... Feeling like the last months of the 1999-2000 bubble.”
Burry compared the Philadelphia Semiconductor Index's run (up ~65% in 2026 at the time) with the run-up into the March 2000 top; as of late September 2026 the S&P 500 and Nasdaq were still setting records, with no bubble peak yet identifiable.
“Bought January 2027 Nvidia put options at the $115 strike for $3.30 while continuing to hold $100-strike puts from earlier — a position sized at roughly 3% of notional value, on the view that Nvidia's valuation leaves almost no room for disappointment: 'borrowing costs could easily get that high or higher if Nvidia stock really starts falling.'”
Options expire January 2027; Burry subsequently escalated the trade, shorting Nvidia outright at $198.09 by June 30, 2026 and adding December 2026 and June 2027 puts in August 2026.
“Bought fresh out-of-the-money Palantir put options expiring in March 2027 with strike prices in the low-to-mid $100s — rebuilding the bearish bet while Palantir traded near $175 — on the thesis that the stock is grossly overvalued and that Palantir underreports stock-based compensation at roughly $5 billion annually.”
Resolves on the March 2027 options expiration: the bet pays off only if Palantir falls below roughly $100–$130 by then.
“Replaced the Micron short with puts at the June 2027 expiry and in the $500 strike price range — a bet that Micron falls sharply from its 2026 highs as the AI memory cycle breaks.”
Micron traded near $1,000 in August 2026 after a ~670% run; resolves at the June 2027 expiration.
“I have little doubt the next few quarters will set still higher and higher net investment/GDP marks, possibly even eclipsing that aftermath of the 2000 tech stock peak. [S&P 500 net capital investment — capex less depreciation — stood at about 2.07% of nominal GDP as of June 30, 2026, a level exceeded only once in nearly four decades.]”
Resolves against quarterly S&P 500 net investment (capex less depreciation) as a share of nominal GDP through mid-2027.
“Swapped the Nebius short for put options at the June 2027 expiration in the 'double digit strike price' range: 'Fundamentally, I am moving timelines up. As such, I want more leverage in my short positions... Nothing says leverage like options, in this case put options, which are relatively cheap due to exceptionally tight volatility measures such as the VIX.'”
Resolves at the June 2027 expiry: Nebius must fall into the double-digit strike range (from roughly $200+ around the time of the trade) for the position to pay off.
“Covered the Nvidia short and replaced it with September 2027 puts struck in the mid-$100s — less than half Nvidia's $229 closing price — a leveraged bet that Nvidia falls sharply by September 2027.”
Resolves at the September 2027 option expiration; distinct from Burry's earlier January 2027 $115-strike Nvidia puts.
“Replaced and rolled the Palantir short and put position into an enlarged put position centered at a September 2027 expiration with strikes in the low $100s.”
Resolves at the September 2027 option expiration; an enlargement and roll of his earlier March 2027 Palantir puts.
“Replaced the iShares Semiconductor ETF (SOXX) short position with September 2027 puts 'in the low $400s.'”
SOXX traded well above $500 in 2026; resolves at the September 2027 expiration.
“Folded the QQQ put position into an enlarged and rolled Nasdaq 100 Index put position, with September 2027 strikes in the $24,000s — implying a roughly 20% decline in the index from about 30,300.”
Resolves at the September 2027 NDX expiration.
“Fundamentally, I am moving timelines up. As such, I want more leverage in my short positions. Better timelines make leverage more palatable... As of today, put options completely replaced my short positions. [Burry said weekend research left him believing] the bubble in AI may burst sooner than later [and that he is] more confident than ever before [that his thesis will] play out over the next year — moving up a previous 'base case' of 2028.”
Burry moved his AI-bust base case up from 2028 to 'over the next year' and swapped outright shorts for leveraged put positions expiring June–December 2027; resolvable by roughly end-September 2027.
“A chip earning more than it costs is a sign of scarcity due to memory and power shortages, not GPU die shortages and not durability. [Once those component shortages ease and newer, more efficient architectures like Vera Rubin ramp up, rental prices for legacy A100/H100-class hardware are expected to drop significantly.]”
As of early October 2026, A100/H100 rental rates were still firm or rising (median on-demand H100 about $3.42/GPU-hour in late September 2026) amid memory and power bottlenecks; the test is whether legacy rental rates fall materially as those shortages ease and Vera Rubin volume ramps through 2027.
“Over the next two years this shortage will blow off as production catches up, and memory will have a down cycle again.”
Burry made the call while adding to short positions in Micron, Nebius, SOXX and Palantir, arguing the AI-era memory shortage is temporary and that rising Chinese production capacity will restore the sector's historic boom-and-bust cyclicality.
“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%.”
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.
“When the write-offs come, perhaps in 2028 or 2029, these commitments discussed in Part IV may be so large that a relatively small write-off has a bigger impact than we can now imagine. After all, the growth rate of these off-balance sheet commitments and exposures is eye-watering.”
Resolves on whether Amazon, Meta, Alphabet, Microsoft and Oracle take material impairments or write-offs on AI infrastructure commitments in 2028 or 2029.
“Let's all take a moment to understand how self-serving it is for OpenAI, Anthropic and other execs of big hyperscalers to talk of slowing things down. 1. LLMs are not AI and won't be AGI. There is nothing AI to slow down. 2. Competition is coming up fast, slowing benefits incumbents.”
Open-ended claim with no deadline; resolves only if and when an LLM-based system is generally accepted as AGI (or the paradigm is definitively superseded).
“They will be lights out and accompanied by the kind of hype that tells you FOMO has almost all the people that would buy it already in it. [On Nvidia's fiscal Q2 2027 earnings, reported August 26, 2026.]”
Nvidia reported Q2 FY2027 revenue of $96.2 billion on August 26, 2026, up 106% year over year and above the ~$92 billion consensus, and guided Q3 to $108 billion with 70% FY2028 growth; shares jumped 8.7% the next day, adding more than $400 billion in market value — a 'lights out' quarter followed by exactly the wave of buying enthusiasm Burry described.