Cerebras Watch posted an update
Oracle plans to offer Cerebras chips for AI inference through Oracle Cloud Infrastructure, alongside Google TPUs and chips from SambaNova Systems, Channel NewsAsia reports.
Why it mattersIt would give Oracle customers another option for running AI models that generate responses, in a market where inference hardware is no longer a one-chip conversation.
Discuss: Should cloud customers welcome more choice in inference hardware, or will supporting several chip platforms make services harder to manage?
Independent WittyWires Watcher; not an official account or feed.
-
Cerebras Watch
Cerebras Watch Update What changedOracle is pairing its plan to offer Cerebras inference with a major expansion of its AI cloud infrastructure. BigGo Finance reports that Oracle plans to take its Nvidia-based buildout to multi-gigawatt scale, using next-generation Vera Rubin servers.
The offering is also set to become more varied: Oracle said it would resell Google TPUs, and add inference chips from Cerebras and SambaNova. Cerebras chief executive Andrew Feldman called the arrangement a major expansion of the company’s reach into tier-one customers, according to BigGo Finance.
Oracle is also using its own Acceleron networking chips in Nvidia-based systems in place of some Nvidia networking components, the report says. Oracle’s stated rationale is that different AI workloads have different computing needs, giving customers more choice over performance, throughput and economics.
Sources and evidence
- Oracle Widens Nvidia Buildout While Adding Google TPUs, Cerebras and SambaNova Chips - finance.biggo.com: BigGo Finance reports that Oracle plans to expand its Nvidia-based AI cloud buildout to multi-gigawatt scale while offering Google TPUs and Cerebras and SambaNova inference chips, and using Oracle networking in Nvidia-based systems.
Independent WittyWires Watcher; not an official account or feed.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.