Krishna, who has run Biggish Blue since 2020, told the Decoder podcast that even a basic calculation shows “no way” these mega investments make financial sense. He said data centres gobble energy and capital in absurd quantities.

Goldman Sachs reckons global data centre power use sits at about 55 gigawatts, with only 14 per cent devoted to AI. Its analysts say demand could hit 84 gigawatts by 2027.

Krishna said a single gigawatt-scale data centre costs roughly $80 billion to build in today’s money. If one company promised 20 to 30 gigawatts, it would face about $1.5 trillion in capital spend, a sum he compared to Tesla’s market value.

He estimated that all hyperscalers together might dream of adding 100 gigawatts although that still needs about $8 trillion of investment.

He said: “It’s my view that there’s no way you’re going to get a return on that, because $8 trillion of capex means you need roughly $800 billion of profit just to pay for the interest.”

Hardware obsolescence only sharpens the pain. Krishna said: “You’ve got to use it all in five years, because at that point, you’ve got to throw it away and refill it,” he said.

Part of the frenzy comes from the scramble to crack AGI. Krishna reckons there is at most a one per cent chance of doing it with current tech, although he praised AI’s value for enterprise productivity.

He said: “I think it’s incredibly useful for enterprise. I think it’s going to unlock trillions of dollars of productivity in the enterprise, just to be absolutely clear,” he said, adding that AGI will need more than the present LLM approach.

Hyperscalers are still shoving money into AI infrastructure. Spending is tipped to reach $380 billion this year. Alphabet lifted its 2025 capital spend forecast to between $91 billion and $93 billion from $85 billion, with the firm’s chief financial officer promising a “significant increase” next year. Amazon raised its estimate for the third quarter to $125 billion from $118 billion.