Google adds two TPU variants as Vahdat takes the Nvidia fight wider
Bloomberg reports two new Google TPU variations aimed at confidential computing and broader information access for models, as Vahdat pushes TPUs beyond Google Cloud.
By Ryan Merket · Published
Primary source: Bloomberg Technology
Why it matters
AI buyers are moving from training bragging rights to production questions about agent latency, memory access, privacy, and where capacity can actually be bought. Google's answer is to widen the TPU family and sell it through Oracle as well as Google Cloud, putting pressure on Nvidia at the point where customers choose infrastructure for years of inference spend.
Google is widening the TPU bet that Amin Vahdat, its senior vice president and chief technologist for AI and infrastructure, has been carrying into customer conversations: custom silicon as an alternative to Nvidia for training and running large AI systems. Bloomberg reported on November 9th that Alphabet Inc.'s Google is introducing two new variations of its tensor processing units, with one aimed at confidential computing and another intended to let…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.