The Microchip Era Is About to End (www.wsj.com)

At Nvidia’s recent AI conference in Washington, CEO Jensen Huang delivered a keynote that doubled as a status report on the “microchip era”: a period in which specialized AI accelerators have enabled an industrial revolution that could embed machine learning into nearly every human activity. Huang used the stage to highlight Nvidia’s chip-driven advances and to thank President Trump for energy and industrial policies that, he said, have brought chip fabrication back to the U.S. from Asia. The article frames Nvidia—now roughly a $5 trillion company—as the exemplar of this era. For the AI/ML community the announcement matters because onshoring fabrication and favorable energy policy directly affect the supply of high-performance accelerators used for training and inference. Domestic fabs can shorten lead times, reduce geopolitical supply-chain risk, and support the massive, consistent power demands of datacenter-scale ML. At the same time Nvidia’s dominance underscores risks—concentration of compute, pricing pressure, and platform lock-in—that could shape research directions and deployment models. The piece suggests we’re at a turning point: hardware scale-up is still central, but future advances will hinge on ecosystem-level choices about where chips are built, who controls them, and how energy and industrial policy enable next-generation AI infrastructure.