AI Bubble on the Brink of Bursting? Economists Warn of Trillion-Dollar Reckoning for Hyperscalers

The five largest US hyperscalers – Amazon, Alphabet (Google), Meta, Microsoft and Oracle – have invested a combined 1.9 trillion US dollars (approximately 1.6 trillion euros) in AI infrastructure since 2020 (as of June 2026). Now leading financial experts are warning of a debt-financed speculation bubble that could shake the entire financial system.

Investment wave reaches historic proportions

For fiscal year 2025, the five hyperscalers together spent 405 billion US dollars on investments, according to Goldman Sachs in August 2026. J.P. Morgan estimates an investment volume of 697 billion US dollars for 2026 alone for the four companies Amazon, Microsoft, Google and Meta in AI infrastructure (as of August 2026). Globally, spending on artificial intelligence amounts to over 2.5 trillion dollars, with expected increases to over 3.3 trillion dollars in the coming year. These sums exceed the annual infrastructure budgets of entire industrialized nations.

Ray Dalio draws parallels to the 1920s

Ray Dalio, founder of hedge fund Bridgewater Associates, compares the current AI boom to the technology euphoria of the late 1920s. At that time, power grids, refrigeration, telephones, radio, airplanes and cars transformed everyday life – and investors eagerly sought participation. The subsequent Great Depression demonstrated how quickly narratives can shift. Dalio's central warning: investors must distinguish between technological development and financial risk when stock price movements are no longer supported by fundamentals.

Bank for International Settlements sees historical patterns

The Bank for International Settlements draws parallels to the railway mania of the 19th century and the dot-com bubble. Both ended with an abrupt reversal of investment flows and economy-wide recessions. The Bank of England pointed in its Financial Stability Report of 7 July 2026 to overextended stock valuations and the fact that AI-related companies now account for about half of market capitalization in the S&P 500 – in 2022 it was about a quarter.

Two-tier structure of the potential bubble

The Rosa Luxemburg Foundation identified two bubble tiers in July 2026: first, a stock market bubble in which OpenAI and Anthropic could each achieve market capitalizations of about one trillion US dollars through initial public offerings. Second, a credit bubble, which is considered significantly riskier since, if it bursts, it could trigger a cascade of defaults throughout the financial system.

Four classic warning signs met

Ruchir Sharma, financial strategist at Rockefeller International, listed four classic warning signs of a speculation bubble in August 2026: overinvestment, overvaluation, excessive capital concentration in few stocks and excessive leverage. According to Sharma, all four warning signs are already present.

Goldman Sachs sounds alarm on debt-financed growth

Goldman Sachs warned in August 2026 about the debt-financed investment wave in Big Tech stocks like Meta, Alphabet and Microsoft. Financial experts see parallels to a 2008-like scenario through debt accumulation. Rising risk factors include rising interest rates, Chinese competition and credit-financed stock purchases.

The business model problem: legend instead of profitability

The Rosa Luxemburg Foundation described a fundamental problem in July 2026: a "perfect legend" has been created – AI is portrayed as an existential threat, making it appear as a tool of "unprecedented power." This justifies massive capital inflows. However: "A legend is not a business model." The spell has held so far, but now the investment must pay off.

OpenAI, Anthropic and the major hyperscalers face the question of whether future demand can justify their enormous obligations. Whether the planned investment sums survive the coming reporting season will be decisive.

Infrastructure bottlenecks as a real constraint

Power and network bottlenecks threaten to become the actual bottleneck of the AI revolution. Hyperscalers are building gigantic data centers, but announced capacities often lie years ahead of actual usage. Utilization rates of the AI giants remain partially opaque.

Concrete stock price losses for individual players

The SpaceX initial public offering on 12 June 2026 brought a very high valuation. After a significant setback in July, the market capitalization at the end of September 2026 was around 2.0 trillion dollars again above the IPO level; SpaceX is thus still valued at more than 80 times annual revenue. The Oracle stock price on 24 September 2026 with a closing price of 139.54 dollars was approximately 57 percent below its twelve-month high of 322.54 dollars. Oracle founder Larry Ellison is spending money "with both hands" on AI data centers.

Differentiation: not all players equally affected

The NZZ determined in July 2026 that while the bubble thesis applies to hyperscalers, it does not apply to all AI boom participants. However: no real "AI bubble" will burst if only a few small companies fail. The risk lies in the concentration among the five large players.

Hyperscalers with their own chips such as Google and Amazon are considered "safe bets." TSMC is seen as the "real winner" of chip diversification. Enterprise software with AI integration such as Microsoft or Salesforce is expected to prevail. NVIDIA emerges as a central supplier and supporter of the investment wave.

Broad consensus among financial experts

A broad consensus among economists and financial experts – from Ray Dalio to the Bank for International Settlements to Goldman Sachs – warns of a potential speculation bubble. The central risk lies not in the AI technology itself, but in the debt-financed overinvestment by hyperscalers whose future profitability is uncertain. Historical parallels suggest that burst bubbles can trigger rapid reversals of investment flows.