Google unveils TPU v5p ‘Ironwood’ chip in direct challenge to Nvidia

Google is taking direct aim at Nvidia with the release of its most powerful chip to date, the TPU v5p Ironwood, which will be available to the public in the next few ￰0￱ announcement was made on Thursday, and it’s part of a much bigger push by Google to claw its way into the top tier of AI infrastructure providers, according to reporting from ￰1￱ chip was originally launched in April for testing and private deployments, but now it’s going live, and the stakes couldn’t be ￰2￱ entirely in-house, Ironwood is designed to power everything from training massive AI models to running real-time chatbots and ￰3￱ says each pod can connect up to 9,216 TPUs, removing what it calls “data bottlenecks for the most demanding models.” The company claims Ironwood gives customers “the ability to run and scale the largest, most data-intensive models in existence.” Google upgrades cloud tech to go after Amazon and Microsoft This launch isn’t happening in a ￰4￱ is locked in a three-way slugfest with Amazon, Microsoft, and Meta for control of the AI hardware space.

Nvidia’s GPUs have powered most large language models so far, but Google’s TPUs fall under custom silicon, which the company says brings better pricing, better performance, and better energy ￰5￱ is the latest product of a ten-year investment in TPU ￰6￱ says it’s four times faster than the last TPU version, and customers are already jumping ￰7￱ startup Anthropic plans to use up to 1 million of the new TPUs to run its Claude model, putting serious weight behind Google’s ￰8￱ the same time, Google is overhauling its cloud platform to stay competitive. It’s rolling out updates that aim to make its cloud services cheaper, faster, and more adaptable, especially as it continues to lag behind Amazon Web Services and Microsoft ￰9￱ Google’s latest earnings report showed $15.15 billion in Q3 cloud revenue, up 34% from a year ago, Azure jumped 40% and AWS came in with 20% ￰10￱ said it has already signed more billion-dollar cloud contracts in 2025 than it did in the last two years ￰11￱ keep up with the rising demand for AI infrastructure, Google increased its capital spending target for the year from $85 billion to $93 ￰12￱ Sundar Pichai said during the company’s earnings call, “We are seeing substantial demand for our AI infrastructure products, including TPU-based and GPU-based ￰13￱ is one of the key drivers of our growth over the past year, and I think on a going-forward basis, we continue to see very strong demand, and we are investing to meet that.” Nvidia CEO backs off after warning on ￰14￱ position As Google moves in, Nvidia’s Jensen Huang finds himself under ￰15￱ Wednesday, Jensen told the Financial Times that “China is going to win the AI race,” blaming lower energy costs and weaker regulations in the ￰16￱ made the comments at the Future of AI Summit, sounding an alarm that the ￰17￱ lose its lead in AI ￰18￱ hours after that quote made headlines, Nvidia posted a new statement from Jensen on his official X account.

“As I have long said, China is nanoseconds behind America in AI. It’s vital that America wins by racing ahead and winning developers worldwide,” he ￰19￱ has stuck with the argument that the ￰20￱ only stay ahead in the AI race if developers stay tied to Nvidia’s ￰21￱ same logic has shaped how he talks to ￰22￱ about keeping exports open, especially to China. Nvidia’s position in the market depends on developers staying dependent on its silicon, and Google’s Ironwood is now threatening that grip. Don’t just read crypto ￰23￱ ￰24￱ to our newsletter.

It's free .