DAVOS, Switzerland, January 21, 2026, 18:34 (CET)
- Nadella warns AI could turn into a bubble if profits remain locked within Big Tech
- He connects the AI race to energy expenses and the “tokens” required to operate models
- Shares of Microsoft dropped roughly 2.3% in U.S. trading
Microsoft CEO Satya Nadella cautioned Wednesday that the AI frenzy could resemble a bubble if its advantages fail to reach beyond the largest tech companies and wealthier nations, even as firms invest heavily in data centers and chips.
At the World Economic Forum in Davos, Nadella told BlackRock CEO Larry Fink that “a telltale sign” of a bubble would be if “all we’re talking about are the tech firms.” He argued the real test lies in whether AI boosts productivity and revenue across the broader economy—not just in capital expenditures by cloud companies expanding capacity. Meta, Alphabet, and Amazon together planned roughly $342 billion in capex for 2025, while Microsoft reported $88 billion in AI-related investments last fiscal year. J.P. Morgan estimated these AI-related investments added 1.1% to U.S. GDP growth in the first half of 2025. 1
Nadella linked the AI race directly to electricity, saying “GDP growth in any place will be directly correlated” to the energy costs required to run AI. He called “tokens” — the processing units powering AI models — the commodity economies must convert into growth. But he cautioned that if results don’t improve, the sector risks losing “social permission” to consume scarce power. Microsoft plans to pour $80 billion into data center development, with roughly half of that budget going to projects outside the U.S., Inc reported. 2
Microsoft shares slipped roughly 2.3%, closing at $443.89 in U.S. trading.
TipRanks noted that Nadella encouraged job seekers to approach AI like previous generations did with spreadsheet tools like Microsoft Excel, suggesting that broader everyday adoption could ease concerns about an AI bubble. 3
Microsoft has steered clear of putting all its chips on one AI provider, collaborating with multiple firms like OpenAI, Anthropic, and xAI, according to PYMNTS. 4
At Davos on Wednesday, Fink repeated the diffusion argument, stating, “I think there will be big failures, but I don’t think we are in a bubble.” He cautioned that the AI boom might fall short if it remains “just the domain” of a handful of hyperscalers — the cloud giants managing massive data centers. 5
The math doesn’t lie. Larger models and increased usage demand more electricity, yet grid improvements and additional power sources often take years to come online. If energy prices remain elevated or supply tightens, the move from investment to widespread adoption could stall — and talk of a bubble will only grow louder.
For Nadella, the bet is that AI will shift from a competition over spending to a challenge of applying it effectively beyond the tech sector.
