At Taiwan’s biggest computing show this month, Jensen Huang, the chief executive of Nvidia, was mobbed as if he were a rock star, trailed by a crowd from booth to booth as he signed autographs and posed for selfies.
At the stand of SK Hynix, one of his most important South Korean suppliers, Mr. Huang picked up a marker and left a message. Atop a reflective wafer of memory — a component of Nvidia’s A.I. supercomputers now in short supply — he wrote: “Please make more :).”
It was only half a joke. As tech giants pour hundreds of billions of dollars into data centers across the world, demand for the chips, wiring and power systems that run artificial intelligence is racing ahead of supply.
Nvidia became the world’s most valuable publicly traded company because of this spending boom. Now the bonanza is lifting a constellation of less famous semiconductor companies — among them many Asian firms few Americans have heard of — that make the essential, unglamorous parts of the data centers being built across the globe.
“This was a boring industry that no one cared about, but now it’s become the most critical infrastructure for the world,” said Timothy Arcuri, a semiconductor analyst at UBS. “They’re basically putting down the tracks, and there will be all this commerce on those tracks for years.”
The gold rush is redrawing the map of technological power. Chips are the brains of A.I. systems that process huge amounts of information. They also need memory technology that holds the information the systems are thinking about. In recent years, A.I. models have grown so vast that holding the information has become as valuable as processing it.
