Is the AI Bubble Bursting? October's Sell-Off and the Question of AI Earnings

October prompted a reassessment across financial markets, with investors asking whether the AI bubble is bursting. Technology equities declined sharply worldwide, spanning exchanges from New York to Seoul and Hong Kong.

The October drop

The Nasdaq 100 index declined by 6.7% in October. Prominent AI companies experienced steeper retracements. Alphabet, Meta and Nvidia all recorded substantial losses.

The decline occurred in two phases. By mid-month, equities had slipped. Interest rates remained elevated and markets anticipated a further Federal Reserve rate hike.

The second phase followed corporate earnings reports. These releases detailed the substantial costs required to build AI infrastructure. Although revenues increased, extensive capital expenditure weighed heavily on cash flow. Alphabet reported negative free cash flow—the cash left after paying for operations and equipment—of $9.1 billion for the quarter. Amazon recorded negative free cash flow of $42.4 billion over a 12-month period.

Meta similarly experienced negative cash flow and began selling shares to raise capital. Concurrently, Microsoft and Google reportedly refrained from increasing their future spending projections.

This selling pressure extended globally. In Korea, which hosts major memory-chip manufacturers, the Kospi index fell sharply as local media questioned the financial returns on AI capital expenditure. Hong Kong’s technology index also declined, extending a longer-term downward trend.

Before judging whether the bubble is bursting, it helps to zoom out.

Zoom out: what was already baked into prices

Before assessing the risks, it is worth asking what prices already reflected. By early October, the premium paid by investors was substantial.

The total market capitalisation of US non-financial companies reached 255% of US gross domestic product. This is the highest in Federal Reserve data. At the peak of the dot-com boom in 2000, the equivalent figure was only 163%.

Global savings remain highly concentrated in the American equity market. Foreign investors held a record $22.2 trillion in US equities in mid-2026. The US is 64% of the world stock market index (MSCI ACWI), while US GDP is only about 25% of world GDP.

The US market exhibits similar concentration internally. By the end of September, ten companies accounted for 40.6% of the S&P 500 index’s total value. This is near a record in J.P. Morgan’s data. These companies’ proportionate share of the index’s profits has also risen steeply.

Prices assumed fast growth

Equity valuations reflect forward expectations. Investors are currently paying a multiple of 25.9 times recent earnings for the S&P 500. Against the earnings expected over the next 12 months, the multiple is close to its 10-year average, according to FactSet. This disparity indicates that the market is pricing in substantial profit growth.

Analysts project the index’s aggregate earnings will rise by 32.9% in 2026. Jefferies estimates that AI-related companies globally will roughly double their earnings this year, significantly outpacing other businesses. A chart from Kotak Mutual Fund shows the gap: 96.5% against 16.4% this year.

Given these elevated growth expectations, October’s earnings reports prompted a negative market reaction. This occurred despite rapid growth in the cloud computing divisions of Google, Amazon and Microsoft. Markets penalised Alphabet and Amazon for their cash flow, heavy capital expenditure on physical assets such as servers, and cautious forward guidance. When expectations are this high, good results alone are not enough.

The music is still playing

Profits derived from emerging technologies require time to materialise. In the interim, a few signs show that demand remains strong.

1. Chip rents. According to Silicon Data’s rental index, leasing an Nvidia H100 chip from specialist cloud providers cost $2.94 an hour on 2 November, an increase from $2.78 on 4 October.
2. Order books. Microsoft’s commercial order backlog stands at $677 billion, Google Cloud’s at $561 billion and Amazon Web Services’ at $578 billion. All three corporations state that demand currently exceeds their capacity to build infrastructure.
3. Cloud profits. Google Cloud reported an operating profit of $11.1 billion, more than tripling the figure from a year earlier.
4. Data-centre space. According to CBRE, only 1.4% of capacity in North America’s primary data-centre markets is currently vacant.

How much of AI profit is technology, and how much is accounting?

Current AI profitability derives partly from technology sales and partly from accounting treatments. Four factors influence the reported figures.

1. Profit this year, cost spread out. When a corporation purchases chips from Nvidia, Nvidia recognises the revenue and profit immediately. The purchaser depreciates the cost over five or six years. Goldman Sachs estimates that almost half of this year’s growth in S&P 500 earnings stems from AI expenditure.
2. Longer lives for servers. Microsoft (since 2022), Alphabet (2023) and Meta (2025) have extended the depreciation schedules for their servers. Recognising lower annual costs increases reported profit. At Microsoft, this accounting change added $3.7 billion to operating income in its first year. Conversely, Amazon shortened the useful life of certain servers in 2025.
3. Paper gains. Between April and June, Alphabet recorded $98 billion in pre-tax “other income”. This consisted largely of unrealised gains on its holdings in SpaceX and a private company identified by press reports as Anthropic. For comparison, the company generated $40.8 billion in operating profit from its core business. In the subsequent quarter, this line item resulted in an $8.2 billion loss, attributed primarily to SpaceX.
4. Money moving in circles. Nvidia invested $2 billion in CoreWeave, a firm that leases Nvidia chips. It also contractually agreed to purchase CoreWeave’s unsold capacity for $6.3 billion. In October, Nvidia announced a $3 billion investment in Meta, which remains one of its largest buyers.

The trajectory of emerging technologies is inherently difficult to forecast, and much of the current optimism comes from the companies selling the technology.

The bottom line

Prior to October, three metrics were near historical peaks: US corporate valuations stood at 255% of the domestic economy, foreign investment in US equities reached a record $22.2 trillion, and the ten largest companies accounted for 40.6% of the S&P 500 index.

These records rest on a single factor: earnings. Investors are relying on those profits continuing to grow by roughly 32.9% in 2026.

With substantial expectations embedded in asset prices, the main risk is that reality falls short. While demand is still strong, a portion of currently reported profits stems from accounting treatments. The question is not whether AI is real, but whether its earnings can catch up with what prices already assume.

Three considerations will indicate whether this is achieved:

1. Are AI revenues converting into steady revenue from paying customers?
2. Is free cash flow improving after capital expenditure?
3. Are earnings estimates rising faster than share prices?