Artificial intelligence continues to expand inside companies. Yet on Wall Street, the financial arithmetic behind that expansion is coming under renewed scrutiny: how much future earnings growth is needed to justify today’s infrastructure spending and share prices?
On November 20, 2026, the Nasdaq Composite fell 5.28%. The S&P 500 lost 3.70%, while the Dow Jones Industrial Average declined 1.93%. AI-linked equities were among the biggest casualties, led by semiconductor companies and businesses exposed to the pace of data center construction.
The rout brought the term AI bubble back into market commentary. But the relationship between technology adoption and stock performance is more complicated than a single-day price decline suggests.
What happened to the Nasdaq on November 20, 2026?
| Market benchmark | November 20, 2026 |
|---|---|
| Nasdaq Composite | -5.28% |
| S&P 500 | -3.70% |
| Dow Jones Industrial Average | -1.93% |
| Nvidia shares | approximately -14% |
Sources: Reuters and Exame closing-market reports. Approximate percentage returns are quoted as reported. An individual stock and a broad index are not directly comparable measures.
The Nasdaq lost about 10% for the week, its steepest weekly decline since 2020, according to Reuters. The losses extended beyond a handful of software firms because enthusiasm for artificial intelligence had also supported valuations across chip manufacturers, memory suppliers, cloud platforms and infrastructure operators.
Investors can remain convinced that AI will transform industries while questioning the amount already priced into stocks.
Why are AI stocks falling?
Three forces converged.
Valuation risk. A high-growth stock may reflect years of expected future earnings. Higher discount rates or a smaller expected profit pool can materially reduce the price investors are prepared to pay today.
Infrastructure economics. AI capacity requires immediate investment in accelerators, networking, land, cooling and electricity. The return arrives later, if demand, pricing and utilization prove sufficient. Investors are reassessing the timing and scale of that payback.
Company-specific uncertainty. Nvidia shares dropped roughly 14% after Reuters reported that the US Securities and Exchange Commission was investigating the company’s accounting. An investigation is not a finding that rules were broken. Nevertheless, uncertainty affecting a major AI supplier can reverberate across the sector.
US interest-rate expectations contributed another layer. Higher borrowing costs make large capital-intensive projects more expensive and reduce the present value of projected future cash flows.
What did Anthropic’s IPO reveal about the AI market?
Anthropic began trading on Nasdaq on November 19. Its shares fell approximately 9% on November 20, according to Reuters, despite the attention attracted by its record-scale initial public offering.
The company’s official closing statement reported $93.5 billion in gross IPO proceeds, including the full exercise of the underwriters’ additional-share option. It also closed a separate $26 billion private placement, bringing the combined gross value of the two transactions to $119.5 billion before applicable expenses.
Anthropic said it received approximately $91.6 billion of the IPO proceeds, with shares sold by existing holders not adding to its own cash. The private placement involved AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC Partners.
These transactions show substantial financial commitment to the AI ecosystem. They also demonstrate how closely connected model developers, chip suppliers and computing platforms have become. Access to capital can support growth, but does not remove the need to deliver profitable operations.
Has the artificial intelligence bubble burst?
The November 20 selloff does not establish that an AI bubble has burst. A bubble implies a prolonged, material disconnect between financial-asset prices and the future economic value those assets can produce. A sharp correction may narrow that gap, but it can also arise from market positioning, interest-rate expectations and company-specific events.
Three questions should be kept separate:
| Question | Evidence to monitor | What it tells investors |
|---|---|---|
| Is AI actually being adopted? | Contracts, retention, workload demand and capacity utilization | Whether end-market demand exists |
| Are AI businesses earning money? | Revenue, margins, operating profit and cash generation | Whether companies capture economic value |
| Are AI stocks reasonably priced? | Market capitalization, growth assumptions and cost of capital | How much investors are paying for that value |
Enterprise demand and stock-market valuation are not identical. A useful product can coexist with an overpriced share. Likewise, falling equity prices do not demonstrate that AI deployment has stopped creating measurable value.
What does the selloff mean for AI investment in 2027?
The key shift is toward financial accountability. Data center projects supported by long-term customer contracts, measurable utilization and reliable cash generation may be treated differently from projects requiring repeated financing rounds.
Investors and corporate buyers should ask more specific questions. What is the cost of serving a unit of computing demand? How much capacity has been contracted? What is the energy requirement? When does the project begin producing free cash flow? Does the investment still make sense under slower adoption?
NEODY previously examined these mechanics in The cost of artificial intelligence has arrived. Who can afford it?, a Portuguese-language analysis of capital spending, profit and cash flow.
The distinction matters: a slowdown in spending could improve the near-term cash position of companies that already own substantial infrastructure, even as it reduces orders for some hardware suppliers.
Could the US AI stock selloff affect Brazil?
Brazil’s exposure is primarily through funding, foreign exchange and technology supply chains. None of these effects should be assumed to follow mechanically from one US trading session.
Financing. Investors seeking to reduce global risk may demand stronger contracts or higher expected returns from new computing infrastructure projects.
Currency and imported equipment. Many advanced accelerators and networking components are priced in dollars. A weaker Brazilian real could raise local-currency costs even without a change in dollar prices.
Project selection. Power availability, infrastructure utilization, customer commitments and financing terms may become more decisive in determining which data centers proceed.
Enterprise adoption. Smaller deployments that demonstrate cost savings or incremental revenue can remain economically compelling even if speculative infrastructure financing becomes more selective.
These are plausible transmission channels, not measured outcomes of the November 20 selloff. Brazil also has its own energy, regulatory and industrial conditions, which must be assessed independently.
What should investors and business leaders watch next?
The market has introduced a stricter test of the AI investment narrative.
Infrastructure investors should monitor signed contracts, utilization, power costs, debt obligations and free cash flow. Corporate buyers should focus on total cost of ownership, measurable operational improvements and ongoing maintenance spending. Equity investors should compare growth expectations embedded in market prices with earnings and cash actually reported.
The prospects for AI remain consequential. The open question for 2027 is which businesses will convert technological progress into economic returns substantial enough to justify the capital already committed.
Image note: The opening image is a digitally generated editorial interpretation of the Nasdaq MarketSite building in Times Square. It is not a photograph of the November 20 trading session or actual weather conditions.






