When Jensen Huang, the chief executive of Nvidia, announced a $500 billion capital-raising effort on Monday, he did not describe his company primarily as a chipmaker. "Today," he said, "we are helping create a new class of productive, investable infrastructure: AI factories." That single reframing may matter as much as the money itself. The coalition Nvidia has assembled reads like a roll-call of global finance: Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR.

What unites them is a shared conviction that AI compute — the physical hardware needed to run artificial intelligence systems — should be treated as an asset class in its own right, comparable to roads, power grids or telecommunications networks. The funds will be directed both toward Nvidia's own projects and toward those of its partners, financing the construction of new data centres, the cooling systems that keep stacked chips from overheating, and the factories required to manufacture those chips at greater scale. The logic behind the move is not difficult to follow. Essentially every major technology company — from Google and Microsoft in North America to the countless AI start-ups now emerging across Asia and Europe — depends on Nvidia's graphics processing units.

Collectively, these companies have spent more than one trillion dollars on AI infrastructure in just three years, a pace of investment that shows little sign of slowing. That sustained demand has pushed Nvidia's stock market value up fivefold over the same period. Jim Zelter, president of Apollo, which manages more than one trillion dollars in assets, called modern compute "a scarce, mission-critical asset class" positioned to drive long-term economic growth. The deeper question is what it means when the physical layer of artificial intelligence — the silicon, the servers, the power — is treated as infrastructure in the same way a toll road once was.

If compute is revenue, as Huang insists, then whoever finances compute may, in time, shape the direction of AI itself.