Big‑tech custom‑chip programs spark fresh doubts about Nvidia's long‑term dominance

Nvidia is still the giant in AI chips. That part has not changed. What changed is the mood around that dominance — and the trigger was a run of earnings and product moves showing that the biggest AI buyers are trying harder to become less dependent on Nvidia. Alphabet’s April 29 results were the clearest example, because they paired booming cloud growth with a new willingness to sell Google’s own AI chips more directly. (money.usnews.com) ### Why are people suddenly nervous? Because Nvidia’s customers are also its richest rivals. Amazon, Google, Microsoft, and Meta spend enormous sums on AI infrastructure, and every dollar they can shift from buying general-purpose Nvidia GPUs to using their own chips can improv(money.usnews.com)ler share of the total AI stack. (latimes.com) ### What exactly did Alphabet do? Alphabet reported first-quarter revenue of $109.9 billion on April 29, with Google Cloud delivering its fastest reported growth quarter yet. On the same earnings call, Google said it would sell its Tensor Processing Units to select customers for use in their(latimes.com) as more than just a good quarter. It looked like a platform move. (money.usnews.com) ### Why do custom chips matter so much? Because AI workloads are splitting in two. Training giant frontier models still rewards flexibility, huge software ecosystems, and top-end performance — all places where Nvidia is very strong. But inference, the day-to-day work of serving(money.usnews.com)other ASICs are the factory tool built for one job. (finance.yahoo.com) ### So is Nvidia actually weakening? Operationally, no. Nvidia reported fiscal 2026 revenue of $215.9 billion, up 65% year over year, and fourth-quarter data center revenue of $62.3 billion, up 75%. Those are absurd numbers. But stocks trade on the next problem, not the last tr(finance.yahoo.com)h. Investors were not arguing that Nvidia is weak. They were arguing that perfection is harder from here. (investor.nvidia.com) ### Who else is pushing this shift? Meta said in April that it is partnering with Broadcom on multiple generations of custom silicon. Amazon has kept pushing Trainium as a lower-cost AI option inside AWS. Microsoft has its own Maia effort. The pattern matters more than any one c(investor.nvidia.com)everse. (about.fb.com) ### Does this kill the Nvidia story? Probably not. Nvidia still has CUDA, developer mindshare, systems expertise, and a full-stack offering that rivals have not matched cleanly. Many customers will keep buying Nvidia because switching costs are real and because frontier training still favors the best all-around platform. (about.fb.com)providers, custom-chip designers, and networking vendors that help stitch giant AI clusters together. (investor.nvidia.com) ### What is the real market takeaway? The AI trade is getting wider. For the last two years, the simplest bet was “buy the company selling the shovels.” Now the market is asking a more interesting question — who captures value when the shovel buyers start making some of their ow(investor.nvidia.com)the AI stack keeps the pricing power. (bloomberg.com)