A semiconductor boom and an AI valuation scare can coexist
The clearest contradiction in the October 8 market was not that AI spending had stopped. It was that a supplier at the center of the advanced-chip ecosystem reported exceptional sales while investors sold semiconductor stocks aggressively. Taiwan Semiconductor Manufacturing Company announced September consolidated revenue of approximately NT$511.86 billion, up 54.6% from September 2025. Yet the U.S. semiconductor index fell 3.4% on Thursday, according to Reuters, with Micron down 4.8%, Broadcom down about 4.4% and Nvidia also under pressure.
The apparent inconsistency is useful. Foundry revenue describes orders being manufactured today. The prices investors pay for AI equities depend on future spending, customer profits, financing availability and assumptions about how long extraordinary growth persists. Both can diverge sharply on the same day. Investors who treat a record revenue number as automatic proof that all chip valuations are cheap miss this distinction; investors who treat one selloff as proof that AI deployment is collapsing make the opposite error.
What TSMC officially disclosed on October 8
TSMC's primary announcement was a monthly revenue release, not a complete quarterly earnings report. September revenue was NT$511.857 billion, 0.6% below August and 54.6% above the prior-year month. August revenue had been NT$514.806 billion and July NT$467.580 billion. Adding the three reported months produces roughly NT$1.494 trillion of third-quarter revenue. Reuters described that total as a record and approximately 51% above the year-earlier quarter.
For the first nine months of 2026, TSMC reported NT$3.899 trillion in revenue, up 41.1% year over year. These amounts are in New Taiwan dollars, not U.S. dollars. The quarterly sum is a calculation from released monthly data; investors should not confuse it with a full GAAP earnings statement containing gross margin, operating margin, net profit and cash flow.
That distinction is especially relevant because the company previously guided third-quarter U.S.-dollar revenue to $44.6 billion–$45.8 billion, using an assumed exchange rate. Reuters estimated the newly revealed quarterly revenue at about $46.7 billion, above that guidance range. Exchange rates complicate direct comparisons between NT-dollar monthly sales and U.S.-dollar guidance, so the authoritative profitability test is still the formal earnings presentation.
Why the selloff happened despite the good number
Reuters linked the broader chip decline to a Financial Times report questioning previously communicated annualized-revenue figures for OpenAI. That is a media-reported interpretation of a private company's financial information, not a public, audited OpenAI filing. It should not be presented as an independently verified $20 billion loss of sales, nor as proof that orders to Nvidia or TSMC have been canceled.
Markets reacted because the report renewed a larger debate: the difference between building enormous AI computing capacity and earning an adequate economic return from the software and services running on it. The spending chain is complex. Model developers purchase computing services; cloud companies finance data centers; chip vendors ship accelerators and networking systems; foundries fabricate the chips. Revenue can be strong near the hardware end of the chain even while investors question the eventual profitability of the final applications.
Reports that technology companies were considering substantial debt financing for AI infrastructure added to those concerns. Debt funding is not inherently a problem, but it increases sensitivity to utilization rates, customer commitments and interest rates. That matters more when long Treasury yields remain elevated. A business may have real demand and still disappoint shareholders if its cost of capital rises faster than expected project returns.
Why the monthly mix matters for Nvidia, AMD and Broadcom
TSMC manufactures chips used across smartphones, computing, high-performance accelerators and other applications. Its monthly revenue therefore cannot be read as a pure Nvidia order book. Still, sustained gains of more than 50% year over year are consistent with very strong demand for advanced manufacturing capacity. The company is an important manufacturing partner to multiple firms, so the data provide a useful cross-check against the most pessimistic interpretations of a single news cycle.
The next questions are profitability and bottlenecks. Advanced-node wafers may command attractive pricing, but leading-edge capacity, advanced packaging, overseas fabrication and the move toward newer process technologies also require heavy capital spending. Quarterly gross margin can tell investors whether revenue growth is translating into attractive economics after these costs. Packaging capacity and customer concentration can also limit how much of the semiconductor market's final demand reaches each participant.
For Nvidia, the issue is whether customers keep buying and deploying accelerators at high rates. For AMD and Broadcom, it is how fast alternative accelerators and custom silicon ramp. For Micron, it is how AI memory demand translates into supply, prices and margins. TSMC is connected to these stories, but it is not a complete proxy for any one of them.
The debate is now about returns, not only demand
The bullish case argues that the record revenue and October 15 earnings event should redirect attention to actual manufacturing output and customer demand. The bearish case argues that the industry could face a period when hardware orders stay high but investor multiples fall because customers begin demanding better returns on committed capital. Both scenarios can happen at once.
One risk is overcapacity if hyperscalers eventually slow their expansion. Another is pricing pressure as customers seek efficiency, especially in inference. Geopolitical and geographic concentration are material for a foundry operating in Taiwan, while rapid overseas expansion adds cost. The short-term market risk is that a financial-reporting controversy at one private AI company gets generalized across an entire supply chain without enough evidence about individual order books.
The next dated test is October 15
TSMC's confirmed third-quarter earnings conference is October 15, 2026, at 2:00 p.m. Taiwan time, corresponding to 2:00 a.m. ET. Investors should watch gross margin, operating margin, the outlook for advanced packaging, the progress of leading-edge nodes and how management describes customer capacity commitments. A revenue beat accompanied by falling profitability would have different implications from a beat with stable margins and constructive guidance.
In the meantime, the most useful market comparison is not one social-media sentiment score. It is the gap between verified manufacturing revenue and the evolving financing economics of the companies purchasing AI infrastructure.
Conclusion
TSMC's September sales release supplied hard evidence of continued expansion: NT$511.86 billion in one month, up 54.6% year over year. The same day, chip equities fell sharply amid questions about AI customer monetization. There is no contradiction once hardware deliveries and the discounted value of future returns are separated. October 15 will test whether TSMC can turn record volume into earnings and cash flow. Until then, a market-wide AI selloff should not be mistaken for an officially documented collapse in foundry orders.