Catenaa, Wednesday, April 22, 2026- Google debuted two AI processors during its Google Cloud Next 2026 conference on Wednesday, as Google pushes further into competition with Nvidia and AMD.
The new chips, called the TPU 8t and TPU 8i, could be optimised for training AI models and inference.
Earlier this month, the company announced an expanded deal with Anthropic to provide “multiple gigawatts of next-generation TPU capacity” to the AI lab.
Google is also working to provide Anthropic rival OpenAI with TPU capacity to power that company’s own AI offerings.
And in February, The Information reported that Meta signed its own multiyear, multibillion-dollar deal for access to Google’s TPUs.
The TPU 8t, Google says, is optimized for training AI models and can “reduce the frontier model development cycle from months to weeks.”
Google says the TPU 8t also offers 2.8x better price to performance than its predecessor, an important metric for customers who need access to high-powered chips but don’t want to pay the steep cost to run them.
The TPU 8i, meanwhile, is best suited for inferencing, or running AI models, and handling AI agents. Both will be available later this year.
Hyperscalers like Google and its rivals Amazon and Microsoft, as well as massive companies like Meta, are increasingly encroaching on Nvidia’s and AMD’s territory.
Amazon and Google both produce their own AI chips and sell/provide them to third-party partners.
On Monday, Amazon announced it had entered an expanded chip deal with Anthropic, under which the AI lab will spend more than $100 billion on AWS technologies over the next 10 years.
Meta, meanwhile, is working on multiple generations of its own AI chips called the Meta Inference and Training Accelerator, or MTIA. More powerful versions of the processors, the company says, are meant to compete with some of Nvidia’s top offerings.
The continued incursions from Nvidia’s own customers could prove problematic in the future. In its most recent quarter, the company reported that hyperscalers accounted for slightly more than 50% of its total data center revenue.
Nvidia’s data center segment is its most important business, accounting for $193.7 billion of its $215.9 billion in total sales in fiscal 2026, which ended in January.
The company, however, has repeatedly pushed back against claims that its customers’ chips pose a strategic threat, saying its own processors are reprogrammable for use across a wide range of workloads.
Google stock rose by over 1% on Wednesday morning. The stock is up by over 6% so far this year.
