Hyperscaler cloud companies including Alphabet, Microsoft, Amazon, Meta and Oracle are on track to spend nearly $1.1 trillion on AI data-centre buildout through 2027.
An MIT Technology Review account, drawing on a Wharton paper by Jessica Wachter, models how fast earnings must grow to justify that spend. It concludes the firms would need to lift productivity by a factor of 2.7 to break even by 2030.
The group is already spending about $750 billion this year on data centres, and some are showing signs of free-cash-flow strain. Alphabet recorded its first deficit in decades, of some $5.9 billion.
"That's a lot of growth compressed into a few years," said Wachter.
Why the sums may not add up
The piece warns that chips turn over fast. Compute electronics account for roughly 60% of costs, and performance is roughly doubling every two years.
Rising borrowing costs add a further risk, potentially leaving new facilities stranded.
That leaves profitability and free-cash-flow paths through 2030 as the critical tests of whether the buildout pays off.
