AI bubble burst in 2026?

- [2026-10-16] SOXX closed at $493.39, down 4.1% on the day. The iShares page shows a 52-week range of 268.01–655.22. The close is about 25% below the 52-week high of 655.22, which is probably near the all-time high. The 40% trigger would be roughly $393. iShares also shows a forward split with a record date of Nov 3 and split-adjusted trading from Nov 5. That could complicate how a "closing price" is read. https://www.ishares.com/us/products/239705/ishares-phlx-semiconductor-etf - [2026-10-17] A GuruFocus item (not fetchable; snippet only) says SOXX fell for a ninth straight day, the longest streak since May 2006. It also says the ETF dropped nearly 12% on the week, its worst week since April 2025. https://www.gurufocus.com/news/9128917/ishares-semiconductor-etf-soxx-hits-ninth-straight-day-of-losses-amid-inflation-concerns - [2026-10-16] NVDA fell about 1.4% to roughly $214. GuruFocus reports a market cap near $5.16T, against a peak of $5.76T on Oct 6. That implies a drawdown of about 10–12% from the high. The 50% trigger would be roughly $118–120. https://www.gurufocus.com/news/9128645/nvidia-nvda-faces-selloff-amid-ai-stock-decline - [2026-10-15] NVDA closed at $217.23, down 1.8%. Drivers were a chip selloff on hot inflation data, rising yields, and memory-chip weakness (Micron, SK Hynix). CoreWeave fell more than 16% after a chart of its bond yields circulated. https://finance.yahoo.com/markets/stocks/article/nvidia-stock-sinks-as-chip-sell-off-deepens-on-hot-inflation-data-203235137.html - [2026-10-17] The Silicon Data H100 neo-cloud index (SDH100RT) reads $2.82 per GPU-hour, up 6.4% over 7 days. This is far above the $1.00 trigger. Third-party trackers put it at $2.81 on Oct 7, and the 2025 low was about $1.70 on reserved contracts. The page's FAQ text still says $2.53, which is stale. https://www.silicondata.com/products/silicon-index/h100 - [2026-10-13] Broadcom CEO Hock Tan said rising rates could make access to capital the biggest constraint for AI companies. Last week, OpenAI told investors its revenue was $50B, below the widely reported $70B. CNBC said the lower figure came from stripping out cloud-partner revenue, not from weaker demand. Chip stocks fell on both. https://www.cnbc.com/2026/10/13/anthropic-q3-financials-ipo.html - [2026-10-13] Anthropic has filed confidentially for an IPO. Reuters reports a November listing, but that is unconfirmed. Valuation talk was about $2T, though an aggregator reported investors discussing about $1.5T on Oct 16. Neither OpenAI nor Anthropic is near bankruptcy, and no OpenAI acquisition is reported. https://www.cnbc.com/2026/10/13/anthropic-q3-financials-ipo.html - [2026-10-15] Yahoo reports Anthropic quarterly revenue of $16.8B and a $20B cut to estimates of OpenAI's 2026 revenue. https://finance.yahoo.com/technology/ai/articles/openais-2026-revenue-estimate-just-032000457.html - [2026-10-15] Regulatory pressure is rising. The FTC sent civil investigative demands to OpenAI and Anthropic (reported by Axios), and eight states sued OpenAI. The EU AI Office also sent information requests. https://jpatoconnell.substack.com/p/the-and-report-42-october-17-2026 - [2026-10-16] Benzinga reports AI-linked CoreWeave bonds yielding about 16% and a historian comparing the boom to the 1873 railroad bubble. It also says Polymarket traders put a defined AI downturn by Dec 31 at roughly 7%. That figure is from an article and may be stale. https://www.benzinga.com/markets/prediction-markets/26/10/62410507/ai-boom-railroad-bubble-1873-nvidia - [2026-10-05] Bloomberg quotes DBS saying NVDA trades at 17x forward earnings, with about 70% earnings growth expected next year. https://www.bloomberg.com/news/articles/2026-10-05/nvidia-s-valuations-show-ai-rally-isn-t-a-bubble-dbs-says - Scheduled before close: the SOXX split on Nov 4–5 and Anthropic's reported November IPO. Nvidia's next earnings date (typically late November) was not confirmed in my sources. BOTTOM LINE: As of Oct 17, a sharp AI/chip selloff is underway, driven by rate and financing worries. Even so, none of the six triggers is met, and only SOXX is down a meaningful amount (about 25% of the 40% needed). NVDA is down about 10–12% against a 50% threshold, and the H100 index is at $2.82 against $1.00. Three simultaneous triggers by Dec 31 would need an extreme collapse. NOT FOUND: Current drawdowns from all-time highs for TSM, ASML, AVGO, ANET and SMCI. The search limit was hit before I could check them. The exact NVDA all-time-high closing price. The exact date of Nvidia's next earnings report.

- [2026-10-16] NVDA closed at $214.13. Its record close was $239.24 on 2026-10-06, so it is about 10.5% below that. Trigger 1 needs a 50% fall, which means a close of roughly $119.6 or lower. The 52-week low was $164.27 (2026-03-30), a smaller drawdown than the trigger needs. https://stockanalysis.com/stocks/nvda/history/ and https://www.macrotrends.net/stocks/charts/NVDA/nvidia/stock-price-history - [2026-10-16] SOXX closed at $514.71 on 2026-10-15, nearly 14% below its October high, after eight straight down sessions. That is the longest losing streak in 16 years per Barchart, though another outlet says since September 2015. The trigger needs a 40% fall, so SOXX has about 26 points still to fall. Fetched: https://finance.yahoo.com/markets/stocks/articles/ai-bubble-cracking-chip-stocks-150000631.html (also https://www.tokenpost.com/news/investing/33336) - [2026-10-16] Memory stocks are much weaker than SOXX. SK Hynix is 57.8% below its record high, Kioxia 65.7% and WDC 57.2%. None of these is on the market's list (TSM, ASML, AVGO, ANET, SMCI), so they don't count toward the triggers. This shows that 50% drawdowns can happen quickly in AI-linked names. UBS called the sell-off a "recalibration of expectations" rather than a fundamental break. https://www.benzinga.com/markets/tech/26/10/62388904/chip-bloodbath-1-trillion-memory-makers-micron-sandisk-sk-hynix - [2026-10-16] The H100 Silicon Index (SDH100RT, neo-cloud) was $2.83 per GPU-hour, up 6.8% over 7 days. The $1.00 trigger is about 65% below that level. The page also shows an FAQ figure of $2.53. Fetched: https://www.silicondata.com/products/silicon-index/h100 - [2026-10-15] H100 rentals have risen in 2026. The index ran about $2.00 in Dec 2025, $2.57 on 2026-05-04 and $2.75 on 2026-07-27. CME forward-curve data shows $2.53–2.81 in mid-October. I found no sign that the index has been near $1.00. https://www.silicondata.com/blog/h200-vs-h100-rental-prices-may-july-2026 and https://cryptobriefing.com/cme-group-gpu-rental-forward-curve-pricing/ - [2026-10-16] Sell-off drivers reported: doubts about OpenAI revenue growth, Anthropic annualized revenue of $67.2B (only modestly above $65B at end-July), inflation and rate fears, and neocloud debt stress (CoreWeave). Big Tech capex of about $800B is still planned. Nothing found on OpenAI or Anthropic bankruptcy, or on an OpenAI acquisition. https://www.benzinga.com/markets/tech/26/10/62388904/chip-bloodbath-1-trillion-memory-makers-micron-sandisk-sk-hynix and https://www.tokenpost.com/news/investing/33336 - [2026-10-02] A comparable earlier drawdown: Nvidia fell more than $1T in value over a two-month selloff to a late-July low, then rallied 23% to a record on 2026-10-02. Earlier in 2026, the 52-week low of $164.27 on 2026-03-30 was about 32% below the 2026 high. Neither episode came near 50%. https://www.bloomberg.com/news/articles/2026-10-02/nvidia-hits-first-record-since-may-as-value-nears-6-trillion - [2026-10-17] Reference class: the Nasdaq Composite peaked on 2000-03-10 and fell 77.9% to 2002-10-09, a span of about 31 months. NVDA fell 89.7% from its 2002-01 peak to 2002-10, and the S&P 500 fell about 49%. Even in the biggest bust, the sector-wide 40–50% drawdowns took roughly a year or more to build from the peak. This time the peak was only 2026-10-06, and the market closes 2027-01-01, about 11 weeks away. https://eco3min.fr/en/macro-financial-regimes/crisis-hub/dot-com-crash-2000/ and https://totalrealreturns.com/n/NVDA - [2026-10-17] Nasdaq has had roughly seven drawdowns of 20% or more since 1985, and only two went beyond 50% (2000–02 and 2007–09 at -55.8%). A 40% fall in a semiconductor ETF is rarer still. The rule needs three of six triggers at once (or within 90 days). Two triggers, bankruptcy of OpenAI or Anthropic and an OpenAI acquisition, are essentially non-market events with no sign of happening. That means the three-trigger threshold is effectively all of: NVDA -50%, SOXX -40%, a supplier -50%, and the H100 at $1. https://drawdownalerts.com/learn/nasdaq-drawdown-history/ - [2026-10-17] Sentiment context: Michael Burry reportedly holds puts on SOXX, QQQ, NVDA and others, expiring Jan 2027, and compares the rally to 1999–2000. This is a secondary-source report and counts as an opinion, not evidence of a downturn. https://intellectia.ai/blog/semiconductor-bubble-warning-2026 - [2026-10-17] Media framing: headlines such as "Is the AI Bubble Cracking?" appear often. One tracker counted 26 flips between "boom" and "bubble" headlines from May 2023 to now, and none of them led to a downturn of the size this contract requires. https://stocktitan.net/articles/ai-boom-or-bubble-market-story-flips BOTTOM LINE: Historically, three simultaneous 40–50% collapses across Nvidia, the chip ETF, a major supplier and GPU rental prices have occurred only in full-scale busts like 2000–02, and those took years to unfold. Today none of the six triggers has been met, NVDA is about 10% off its high and SOXX about 14%. The H100 index is rising, not falling, and under 11 weeks remain. NOT FOUND: Any current forecasts or expert probability estimates for this specific contract. Current prices for TSM, ASML, AVGO, ANET and SMCI versus their highs. SOXX's exact all-time high. A reference-class frequency for "three of six triggers" events.

YES means: at least three of the six listed triggers are met within the 90-day window the rules define. The triggers are NVDA −50%, SOXX −40%, OpenAI or Anthropic bankruptcy, OpenAI acquired, H100 index at or below $1.00 for five days, and a −50% close for TSM, ASML, AVGO, ANET or SMCI. All drawdowns are measured from all-time highs. - [2026-10-16] **The SMCI trigger may already be met on a literal reading.** Macrotrends, split-adjusted, shows SMCI's all-time-high close of $118.81 on 2024-03-13 and a close of $31.63 on 2026-10-16, about 73% below that high. The rules say "all-time high", not 52-week high. Sources that cite SMCI's 52-week high of $56.60 show only about a 44% drop. A forecaster should check how the resolver reads this. The −50% bar was cleared back in June, when one source put SMCI about 51% below its 52-week high. That makes SMCI the most likely first trigger, which leaves two more needed. https://www.macrotrends.net/stocks/charts/SMCI/super-micro-computer/stock-price-history - [2026-10-16] SMCI also has company-specific weakness. A new CFO was named on Oct 15 and the stock fell as much as 9.9% on Oct 16, to $31.63. It carries an unresolved export-control case. GuruFocus reports shares are down 41% over the past year. https://www.gurufocus.com/news/9128447/is-super-micro-computer-inc-smci-a-bargain-after-99-drop-gf-value-says-undervalued - [2026-10-16] The SOX index is about 22% below its June 22 record, per Benzinga's weekly wrap, and that is the second bear market this year. So SOXX has already had one 20%+ leg in 2026. The 10-year yield is about 5.09%, after touching 5.37% last week. September CPI was 3.7% and PPI 6.3%. A Fed hike in October is still possible. Rising rates are a macro driver that could push a further leg down. SOXX needs about −40% from its $655.95 high. https://www.benzinga.com/markets/equities/26/10/62424485/chip-stocks-endure-longest-losing-streak-since-2006-as-coreweave-credit-concerns-mount-this-week-on-wall-street - [2026-10-16] The selloff has spread beyond memory. Coherent fell 24.8% and Ciena 25% in one day, the worst for each since 2000 and 2008 respectively. KOSPI lost 14.5% in a week, its worst since Lehman. CoreWeave CDS passed 1,000bp, which implies about a 60% five-year default probability. Its 9% notes trade near 80 cents. This is the contagion channel from credit stress to equities. https://www.benzinga.com/markets/equities/26/10/62424485/chip-stocks-endure-longest-losing-streak-since-2006-as-coreweave-credit-concerns-mount-this-week-on-wall-street - [2026-10-16] Financing is tightening. Global AI debt issuance fell to $23B in September from $113B in June. Dimon confirms his bank is tightening credit. About $18B of Oracle data-center loans are stressed. Firmus shelved an IPO that sought a valuation above $30B. Gartner (Oct 14) says about $3.5T of AI infrastructure spending could be impaired by 2030. This is the mechanism for a hardware-demand shock that would hit AVGO, ANET and NVDA. https://www.axios.com/2026/10/16/wall-street-limit-ai-spending-debt - [2026-10-16] Broadcom is the other supplier to watch. It closed near $329 (−4.4%) on Oct 16. KeyBanc flagged MediaTek gaining share in Google's TPU programme and cancelled Meta custom-chip orders. Microsoft is also launching custom chips. A −50% move would need a drop to roughly half its high, and I did not find AVGO's all-time-high close. https://neoreports.substack.com/p/daily-ai-investment-brief-261017 - [2026-10-15] TSMC guided cautiously on 2027 even after a record quarter, and ASML's sold-out message did not stop a 4–5% drop that day. This shows bad-news-ignoring price action. Tech-sector CDS also hit a record. https://www.marketscreener.com/news/asml-s-sold-out-outlook-forces-investors-to-look-beyond-demand-ce785dd3de8bf027 - [2026-10-16] The relevant tail risk is in compute-neocloud credit rather than the demand side. Gartner says over 80% of neoclouds will stop operating independently by 2030. A forced-liquidation wave of H100s could pull the index down. It is currently $2.8, so a fall to $1.00 would be a roughly 65% drop. That would be the hardest trigger to reach, because there is no sign of a glut. https://www.axios.com/2026/10/16/wall-street-limit-ai-spending-debt - [2026-10-15] OpenAI finances are the most stressed non-market input. Its 2025 operating loss was about $20.9B on $13.07B revenue. A $278B cash burn is projected for 2026–30. It is seeking $30B or more at a $1.4T pre-money valuation, and a delay or down-round would revive the worries. Bankruptcy or acquisition within 11 weeks is still remote. https://admiralmarkets.com/analytics/traders-blog/openai-revenue-ai-selloff-october-2026 - [2026-10-16] **Most likely path to YES.** (1) SMCI counts via the all-time-high reading, already satisfied. (2) The rate and credit shock continues, with a Fed hike and a failed OpenAI or neocloud financing. (3) SOXX falls a further ~20% from $493 to about $393, below the 40% line. Memory and optical names are already down 55–65% and are leading. (4) One of AVGO, ANET or TSM falls 50%, or NVDA falls from ~$214 to about $120. The window runs to Dec 31, and 90-day lookback language means moves before Oct 17 may also count. The SOXX split on Nov 5 may complicate how closing prices are read. https://www.ishares.com/us/products/239705/ishares-phlx-semiconductor-etf BOTTOM LINE: The strongest YES argument is that SMCI's all-time-high reading means one of three triggers may already be satisfied, while a rates-and-credit shock is hitting the sector, with SOXX in a record losing streak. Two more large-cap collapses (SOXX −40% plus NVDA or another supplier −50%) are still needed within about 11 weeks. NOT FOUND: AVGO, ANET and TSM all-time-high closes and drawdowns, and ASML's drawdown in USD. NVDA's next earnings date. Any resolver statement on SMCI's all-time-high handling. SOXX's exact all-time-high close (sources say $655.95 or $655.22).

YES means: the AI industry meets the contract's "downturn" test by the deadline (Dec 31, 2026 / trading close 2027-01-01). At least three of the six listed triggers must occur within the 90-day window. The triggers are NVDA −50%, SOXX −40%, OpenAI or Anthropic bankruptcy, an OpenAI acquisition, H100 at or below $1.00 for five straight days, or one of TSM, ASML, AVGO, ANET or SMCI down 50% from its all-time high. - **[2026-10-16] SMCI looks like trigger 6, and it has been met for a long time.** SMCI closed at $31.63, per the stockanalysis.com history. Macrotrends (search snippet only; the page returned 403) gives an all-time closing high of $118.81 on 2024-03-13. That puts SMCI about 73% below its high, so the "supplier down 50%" trigger is probably already true. This matters for the NO case in two ways: - The supplier trigger is not a new signal of an AI bust. SMCI's drop dates to 2024 and reflects company-specific problems: an indictment, dilution and an export-diversion case. - Any later collapse in TSM, ASML, AVGO or ANET would add nothing, because this trigger can only count once. The other four named suppliers are far from −50%. - https://stockanalysis.com/stocks/smci/history/ - **[2026-10-17] AVGO is down 35% from its high, which is not enough.** Benzinga (via KuCoin) puts AVGO at about $328, 35% off its high, with market cap down from $2.29T to $1.57T. A −50% close would need roughly $245, a further drop of about 25%. The same piece says NVDA is down 12% from its peak. It also says the memory ETF DRAM is down 43% from its peak, but DRAM is not a listed trigger. https://www.kucoin.com/news/flash/broadcom-stock-falls-35-amid-ai-sector-concerns-analysts-remain-optimistic - **[2026-10-17] The main supplier-trigger candidates are far from −50%.** TSMC guided Q4 revenue up as much as 61% year on year, and its CEO called demand "extremely robust". Yahoo notes the stock fell only 2.28% on Oct 16. I found no source for ASML, TSM or ANET drawdowns from all-time highs. ASML and TSMC both cite strong order and revenue trends, and Bernstein calls the memory and chip selloff a buying opportunity. https://ca.finance.yahoo.com/news/robust-tsmc-earnings-help-settle-101010790.html - **[2026-10-17] The fundamentals do not match a bubble burst.** Nvidia's last report (2026-08-26) showed revenue of $96.2B, up 106% year on year, with data-center revenue up 117% and gross margin at 75%. The company holds $99B of remaining buyback authorization. A 50% price fall would need a collapse in expected earnings, not just multiple compression. https://www.sec.gov/Archives/edgar/data/0001045810/000104581026000073/q2fy27pr.htm - **[2026-10-17] NVDA's next earnings date is Nov 17 or Nov 25 (sources conflict).** Market Chameleon, Earnings Whispers and Wall Street Horizon give Nov 17 after the close. TipRanks gives Nov 25. Either way it falls inside the window. Nvidia has beaten estimates on every report in the history table. A −50% close would need about $119, roughly 45% below the Oct 16 close of about $214. The panic would have to start before the print, because a collapse of that size has no precedent for a $5T company within weeks. https://www.tipranks.com/stocks/nvda/earnings - **[2026-10-17] SOXX's drawdown is measured from a high that is only days old, and the fund is still up strongly this year.** 24/7 Wall St. says SOXX fell 16% in eight sessions from $589.45 on Oct 6, to $493.39. It is about 25% below its 52-week high and still up 64% year to date. A −40% close, about $393, needs another roughly 20% fall. Speculative leveraged ETFs are already hit hard (SOXL −33% on the week). However, a 16% fall in eight sessions has happened before without turning into a 40% fall, and even the 2022 chip bear market took many months. https://finance.yahoo.com/markets/stocks/articles/chip-stocks-just-fell-8-161520543.html - **[2026-10-17] SOXX's pending forward split adds a small resolution risk, and it does not help YES.** The earlier report notes a split with adjusted trading from Nov 5. Providers back-adjust, so the 40% test should still be computed on adjusted closes. A careless reading of unadjusted prices would make the drawdown look smaller, which also points toward NO. - **[2026-10-14] Official and bank commentary calls this a recalibration, not a break.** UBS's CIO office, quoted by Quartz/Bloomberg, called the selloff "recalibration of expectations" and not a fundamental break. The drivers are rates (September CPI 3.7% against 3.6% expected), financing worries and Chinese competition. These are macro and competitive pressures that can reverse. They are not demand collapses like the 2000–02 telecom capex bust. https://qz.com/chip-stocks-trillion-ai-selloff-semiconductor-index-101426 - **[2026-10-15] Rates are a swing factor.** The inflation print has raised Fed hike odds, which is why chips and CoreWeave sold off. A softer print or Fed pause could reverse part of the move. Yahoo describes the pressure as bond yields and cost of capital, not a loss of AI demand. https://finance.yahoo.com/markets/stocks/article/nvidia-stock-sinks-as-chip-sell-off-deepens-on-hot-inflation-data-203235137.html - **[2026-10-17] The remaining triggers are very unlikely to fire.** The H100 index is about $2.8 and rising, against a $1.00 trigger. Anthropic has filed for an IPO. OpenAI's revenue is $50B on the lower definition (Oct 13 report). With those three effectively off the table, YES needs NVDA −50%, SOXX −40% and one more supplier (AVGO, TSM, ASML or ANET) to −50%, on top of the SMCI trigger already counted. That requires the three largest chip names to lose about half their value within about 11 weeks of the Oct 6 peak. Of the 2000–02 collapse, the earlier report says the sector-wide 40–50% drawdowns took a year or more. - **[2026-10-17] Sentiment extremes also cut against a clean YES.** Soft media signals such as "bubble" headlines and the cited ~7% Polymarket figure (stale) show that traders priced a defined downturn as unlikely. The deadline arithmetic is hostile to YES: even a 2008-style crash from here would need the simultaneous trigger drawdowns to complete by Dec 31, with only about 10 weeks of trading left. BOTTOM LINE: Only SMCI (down about 73% since 2024) already meets a trigger. The other five need NVDA near $119, SOXX near $393 or a second large supplier down 50%, and AVGO is only down about 35%. Each of those is a 20–45% further fall inside about 10 weeks, and the H100 price and the OpenAI/Anthropic triggers show no movement toward YES. NOT FOUND: Verified drawdowns from all-time highs for TSM, ASML and ANET. An independent confirmation of SMCI's $118.81 all-time closing high (the Macrotrends page was blocked, so this rests on a search snippet). A confirmed AVGO all-time-high closing price (the 35% figure is Benzinga's).