The artificial intelligence investment boom has developed a familiar pattern: a new threat emerges, technology stocks sell off, and investors soon begin buying the dip.
DeepSeek’s arrival in January 2025 challenged assumptions about the cost of developing advanced AI.
Nvidia fell nearly 17% in a day, losing about $593 billion in market value, before recovering around 8% the next session.
During the tariff turmoil in April 2025, the Nasdaq-100 suffered its fourth-worst two-day loss since 2007, before recording its second-best single-day gain over the same period on April 9, after the White House announced a 90-day pause in reciprocal tariffs.
More recently, the Nasdaq Composite fell 5.5% on November 5 as investors worried that Democratic victories in the US midterm elections could slow data-centre construction.
The index rebounded 2.7% on November 9, although falling oil prices also helped lift broader markets.
The rebounds have not erased all losses.
But the sequence captures the trade’s central tension: investors believe AI could transform businesses while questioning the valuations and financing behind it.
Why does the AI trade keep coming back?
For Louisa Loran, an executive adviser and global business strategist who previously worked at Google, the answer begins with demand.
“The AI investment boom has faced several existential threats, but none of them has touched what investors are actually betting on, which is use,” she said.
DeepSeek’s low-cost models raised concerns that cheaper systems could undermine the economics of US AI developers and reduce demand for high-end chips.
Loran argues that lower costs could instead encourage wider adoption.
“DeepSeek made AI cheaper, and lower access cost has always increased uptake in an industry, so it strengthened the case for adoption rather than weakening it,” she said.
Adoption, however, is not yet translating consistently into corporate profits.
A McKinsey survey published in August 2026 found that 80% of respondents said AI had improved individual productivity, while 37% reported at least some contribution to their organisation’s earnings before interest and taxes.
Loran said she had seen this lag while working with Google customers.
“Infrastructure scales at the speed of capital, while returns arrive at the speed at which organisations change how they work and make decisions,” she told Invezz.
Companies may build infrastructure before redesigning processes and training staff to capture its benefits.
Harry Hamann, founder of investment research firm InvestWealth, offered a financial explanation.
He said the market scares had not broken the revenue and cash-flow engines funding the largest technology companies.
The biggest spenders, he argued, have financed much of the build-out from existing businesses, allowing infrastructure orders to continue despite falling share prices.
Hamann separates short-term market movements from longer-term fundamentals.
“Over days, it is mostly positioning: crowded holdings, leverage and options hedging turn a headline into a 5% to 15% move. Over quarters, fundamentals decide, which is why every dip so far has been bought,” he told Invezz.
While Hamann’s framing relies on broad market mechanics rather than a quantitative decomposition of each pullback, it highlights a key divide between short-term trading and long-term capital allocation.
Why are trillion-dollar companies swinging so violently?
The dollar sums involved can make moves appear unprecedented.
A 5% decline in a company valued at $5 trillion erases $250 billion in market value.
But market capitalisation is the share price multiplied by all outstanding shares; a fall in that figure does not mean an equivalent amount of cash has left the stock.
The percentage moves still matter. Franklin Templeton Institute estimated that hyperscalers and AI-adjacent companies together made up nearly 46% of the S&P 500’s market value as of September 30, 2026.
The figure includes companies supplying or adopting AI, not just businesses that sell AI products.
That concentration means a change in sentiment towards AI can affect a significant share of the broader market.
It can also magnify crowded trades, momentum strategies and leveraged positions.
The Bank of England’s July 2026 Financial Stability Report said high concentration, correlated momentum-driven positions and increased leverage could amplify a fall in AI-related share prices if investors reassessed the sector’s earnings prospects.
Options-market data provide a clearer picture of how investor positioning changed as technology stocks came under pressure.
In a November 2 analysis, Cboe Global Markets found that the proportion of stocks among the S&P’s top 100 showing inverted call skew — a sign of unusually bullish options sentiment — had fallen from 44% in early October to 10%.
At the same time, retail put buying in mega-cap technology stocks on Cboe’s exchanges reached a year-to-date high, accounting for 34% of retail opening trades, compared with an average of 24%.
The figures suggested that investors were moving away from bullish bets on further gains and increasingly seeking protection against potential declines.
That shift became more pronounced during the November 5 sell-off.
In a follow-up analysis published on November 9, Cboe found that increases in S&P 500 put skew and put convexity accounted for 5.5 points of the 11.5-point increase in the VIX that day.
Yet the rise in hedging demand did not translate into widespread expectations of a broader market sell-off.
Cboe’s COR1M implied correlation index reached just 22% on November 5, compared with a historical average of around 70% on days when the S&P 500 fell 4% in a session over the previous two decades.
Five of the 11 S&P 500 sectors also finished the week higher.
Cboe said the options market was still pricing a rotation out of AI and technology stocks rather than a broader market contagion.
The data point to a sharp change in investor positioning and demand for downside protection, but do not establish that options trading caused the underlying share-price declines.
These mechanics help explain why a headline can produce a larger move than the immediate change in earnings outlook.
Still, explaining the market mechanics behind violent swings does not prove that underlying valuations are safe from fundamental shocks.
On November 12, Nvidia shares tumbled as much as 14.3% intraday following CEO Jensen Huang’s sudden resignation amid an internal conduct and security probe.
Yet the stock recovered to close down just 4.3% after CFO Colette Kress was named interim CEO—while rival chipmakers AMD and Broadcom surged roughly 6% on the day.
That intraday reversal underscored how quickly governance uncertainty can be repriced in a central AI stock without breaking broader appetite for the sector’s hardware.”
Nor are large percentage declines in major indices unprecedented. The Nasdaq Composite fell 12.3% on March 16, 2020, during the Covid-19 market shock.
What makes today’s moves striking is the enormous market value attached to a concentrated group of AI-linked companies, not proof that markets have entered an entirely new category of volatility.
When does volatility become dangerous?
A share-price fall is not automatically a financial-stability event.
The risk grows if an AI repricing undermines debt used to finance data centres and chips, or spreads losses to lenders.
Professor Iwa Salami, professor of financial law and regulation at the University of East London, said both business concerns and investor behaviour contribute to volatility.
Worries about profits can trigger selling, while crowded ownership, borrowing and rapid efforts to reduce risk can intensify declines and affect the wider market.
Salami said a more serious warning would be companies spending heavily on AI for years without generating sufficient returns.
“Falling profits, mounting debt or a lasting drop in AI investment would be more worrying than a short-term fall in share prices. Regulators should also watch whether losses spread to banks and other financial institutions,” the professor told Invezz.
That concern is reflected in official assessments.
A survey of 20 market contacts included in the Federal Reserve’s May 2026 Financial Stability Report identified concerns about AI-related equity valuations and capital expenditure increasingly funded by debt.
The report clarified that these responses reflected the views of survey participants, not those of the Federal Reserve Board or the New York Fed.
The Bank of England said AI-related companies’ use of credit markets had accelerated rapidly, spanning public markets, private credit, leveraged finance and structured finance.
It described the pace of investment as historically unprecedented, although it found little evidence so far that AI financing was crowding out other borrowers.
It warned that declining hyperscaler free cash flow and increasingly complex financing arrangements could create risks if earnings disappoint or refinancing conditions worsen.
The Bank for International Settlements said financial-stability risks from the AI boom appeared moderate, but warned that its sustainability hinged on companies meeting high earnings expectations.
It also noted that equity prices had moved considerably ahead of debt-market valuations.
The AI trade has so far survived shocks to assumptions about model costs, trade policy and political support for infrastructure.
But resilience should not be confused with immunity.
AI may become more widely used even if some companies have overbuilt, borrowed too heavily or paid too much for infrastructure.
The technology can succeed while particular investments fail.
For investors, the question is whether companies financing AI’s expansion can earn returns that justify the capital committed.
For regulators, it is whether the debt links built around it could transmit a correction into the wider economy.
The next scare may pass, as previous ones have. The test will be whether the cash flows and financing behind the build-out remain intact.
AI disclosure: AI tools assisted with research and drafting of this piece. An Invezz editor reviewed the facts, sources, and wording before publication. Any opinions or analysis are Invezz’s, not those of an automated system.
The post Why the AI trade is hard to kill, but easy to frighten appeared first on Invezz