Google released two new TPU chips, the TPU 8i Max and TPU 8tx, via its official blog on Nov 9. It is the company's latest move in AI compute hardware and is widely seen as a direct challenge to market leader Nvidia. AMD confirmed the same day that its 5th Gen EPYC processors provide the host infrastructure for Google's TPU 8tx.
Chip positioning and infrastructure partnership
According to Google's official blog, the two chips released are the TPU 8i Max and TPU 8tx. Google did not disclose specific performance specifications or pricing for either chip in the blog summary.
AMD said its 5th Gen EPYC processors provide the host infrastructure for the TPU 8tx, meaning Google has brought AMD CPUs into the server ecosystem built around its in-house AI chips.
TPUs (Tensor Processing Units) are custom chips Google designs for machine learning workloads and have long been offered as compute only through Google Cloud.
The release continues Google's strategy of using in-house chips to reduce its reliance on Nvidia GPUs.
Competitive landscape and market reaction
The new chips arrive as competition in AI hardware keeps heating up. Media reports the same day said Nvidia CEO Jensen Huang could be ousted, with the company yet to name a successor; the report has not been confirmed by Nvidia. Separately, Nvidia announced investment plans in several Korean AI startups.
In the capital markets, hedge funds are dumping global tech stocks at one of the fastest paces on record, signaling investor caution over tech valuations and prospects. Whether Google's new TPUs can shake Nvidia's position in the compute market remains to be seen.
What to watch
- Whether Google discloses detailed performance metrics for the two new TPUs and their cloud service launch dates;
- The scale of AMD 5th Gen EPYC processor deployment in Google's TPU infrastructure;
- Nvidia's response to Google's new chips and to rumors about its own management.