Google is reportedly pushing its in-house AI chips into a market Nvidia has dominated, according to a report by The Information that was picked up this week by outlets including TradingView, Stocktwits and GuruFocus.
The report says Google is expanding sales of its custom TPUs — tensor processing units, the specialized processors it designed to train and run AI models. Until now, those chips have mostly been a Google story: built by Google, used inside Google's own data centers and rented out through its cloud.
What's new is who Google is said to be courting. According to The Information's report as summarized by Yahoo Finance, the company is specifically targeting "neocloud" providers — the newer wave of companies that rent out AI computing power — to drive broader adoption of TPUs. One of the names mentioned is Nscale, which is itself backed by Nvidia.
That detail is the whole story in miniature. Nvidia doesn't just sell the chips that power most of the AI boom; it has invested in the companies that buy and resell that computing capacity. Google approaching one of those Nvidia-backed firms with a rival chip is a direct challenge to the arrangement.
The reports landed against a jittery market backdrop. Coverage from TradingView and Stocktwits framed the news alongside Wall Street price target cuts on Big Tech heading into Q2 earnings — a sign that investors are reassessing how much AI spending will pay off, and who captures the profit from it.
Google has not publicly confirmed the sales push described in the report, and the sources here do not include figures on how many chips are involved or what any deals might be worth.
Why it matters: Nvidia's grip on AI chips gives it enormous pricing power over nearly every company building AI, and a credible second supplier — especially one aimed at the fast-growing rental market — could start to loosen that hold and lower the cost of AI for everyone downstream.