Central Banks and Analysts Warn of Overvaluation in Leading AI Companies

Capital markets show a contradictory situation in October 2026: On one hand, 88 percent of surveyed companies worldwide are using artificial intelligence, 70 percent are deploying generative AI in at least one function, according to data from Stanford University. On the other hand, central banks and analysts are warning of overvaluations in parts of the industry. The Bank for International Settlements (BIS) found in its 2026 Annual Economic Report that the implied long-term earnings growth of leading AI companies is significantly above historical comparison values, while risk premiums have simultaneously declined.

Concentration Reaches Record Level

The ten largest stocks in the S&P 500 now represent 35 percent of the total index – more than during the peak of the dot-com bubble in 2000, when the figure was 25 percent. This concentration increases the risk of a sector-specific correction that could heavily burden the overall market. Global investments in artificial intelligence are expected to exceed $2.5 trillion in 2026 – an increase of 44 percent compared to the previous year. Around half of this sum flows into data centers and infrastructure.

NVIDIA shares have increased by more than 880 percent over the past three years. Leading AI stocks such as NVIDIA and Broadcom have price-to-sales ratios of over 30 – a level that historically preceded sharp market corrections. AI and semiconductor stocks reached valuations in 2024 and 2025 that were last observed during the dot-com era.

Private Valuations and Debt Markets Under Pressure

OpenAI and Anthropic achieved exceptionally high valuations in the private capital market in 2026, which imply high expectations for future revenue growth and margins. At the same time, IPOs of AI companies are being postponed because investors are responding cautiously to the proposed valuations.

Credit markets are sending warning signals: Banks are showing increasing discomfort with financing AI infrastructure. At least one bank that had arranged loans for data centers sold portions of the debt at a price below 90 cents per dollar. Other banks faced difficulties placing billion-dollar loans for data centers in Texas and Wisconsin that were to be rented to major technology companies. Rising refinancing risk is increasingly drawing the focus of risk assessors to the financing of so-called neocloud providers.

Differences from Established Technology Corporations

Microsoft, Alphabet, Meta and Amazon finance their data centers from operating cashflow and not through debt or capital increases – an important difference from the dot-com era. Smaller providers, by contrast, are increasingly relying on debt-financed expansion, which concentrates bubble risks.

Three Risk Zones in Focus

The main risks of an AI bubble are concentrated in three areas: First, extreme valuations of private companies, second, debt-financed data center expansion and third, long-term computing power contracts that assume very high future cashflows. This structure differs from the broad speculation of the dot-com era, when many companies had neither revenue nor viable business models.

Analysts recommend paying attention to observable indicators: revenue growth among customers, credit spreads, graphics processing unit utilization, free cashflow and contract cancellations can provide early warning signs of overheating or cooling. There is no reliable way to predict the exact timing of a correction.

Systemic Effects of a Correction

The International Monetary Fund (IMF) warned in October 2026 that a bursting of the AI bubble could shake global financial markets. A loss of investor confidence would lead to delays and cuts in AI investments and burden gross domestic product – in the worst case, a pronounced recession looms. Financial markets are increasingly exposed to an AI-driven correction, especially if the billions flowing into technology fail to deliver the expected profits.

Action Options for Investors

Investors should conduct thorough due diligence on individual companies, diversify across the AI value chain and adjust position sizes to inherent volatility. Exposure to high-quality AI beneficiaries can be combined with strict risk management. Corporate customers are advised to finance measurable workflows rather than speculative implementations, maintain model portability and avoid long-term infrastructure commitments that are not supported by validated demand.

The central analytical distinction is: A bubble can form around useful technology – overinvestment and lasting technological value are not mutually exclusive. The closest historical analogy is not the disappearance of AI like failed dot-com companies, but the possibility that too much capital flows into valuable infrastructure at unsustainable prices.