Big tech AI spending has hit 1.5 trillion dollars in commitments

Add up what the largest technology companies have committed to AI and the total is approaching 1.5 trillion dollars. Not spent. Committed: chips, data centre capacity, power contracts, multi-year cloud deals signed against demand that has not arrived yet.

Now set that against the figure that keeps turning up in enterprise surveys. Roughly 80 percent of companies deploying AI report no meaningful profit impact so far.

Big tech AI spending is physical, and it is already booked

These are not soft numbers on a slide. SMIC is raising prices because its fabs are running near capacity, and every major foundry on earth is effectively spoken for. Samsung has committed 648 billion dollars to AI chip capacity. Tesla and SpaceX are building Terafab in Texas. The bottleneck has moved past silicon into power and interconnect, which is why optical networking inside data centres suddenly became an investable category.

Concrete gets poured, substations get built, turbines get ordered years ahead. That part of the story is not speculative at all.

The bubble question, without picking a side

Both things are true at once, which is what makes this awkward to write about honestly.

The bear case is arithmetic. Commitments this size assume revenue that has not shown up on anyone's income statement, and depreciation on GPUs starts the day they are racked. If enterprise adoption keeps stalling at proof of concept, someone eventually writes down a very large number.

The bull case is that infrastructure has always run ahead of the application layer. The fibre glut of 2001 bankrupted the companies that laid the cable and then quietly made streaming video possible a decade later. Overbuilt capacity is not the same as wasted capacity; it is capacity whose owner changes cheaply.

What to watch next

Nvidia reports Q2 FY2027 earnings on 26 August. That is the number that stress-tests all of this, because Nvidia's order book is the closest thing to a live reading of whether commitments are converting into deliveries or quietly slipping to the right.

The second thing to watch is the profit-impact surveys. If that 80 percent figure starts moving over the next few quarters, the buildout looks early rather than reckless. If it holds flat into 2027, the argument gets a lot less comfortable.