Oracle adds Google's AI chips to its Nvidia buildout so it relies on no single supplier
Oracle's cloud arm will buy Nvidia's next-generation systems and rent out Google's TPU chips too, a sign that big buyers are spreading their chip bets.
The short version
- Oracle's cloud will deploy Nvidia's Vera Rubin systems at a scale of several gigawatts.
- It is the first outside large cloud provider to offer Google's TPU processors.
- Some equipment will use Oracle's own Acceleron networking chips instead of Nvidia's.
- SambaNova and Cerebras chips are also on the purchase list.
- Oracle kept its capital spending range, whose top end is $95 billion.
Why it matters
The chips behind AI services are in short supply, and one supplier's problems can slow everyone who depends on it. A major cloud provider mixing Nvidia, Google and smaller makers gives customers more choice. It does not promise faster or cheaper service for them soon.
What Oracle announced
As Yahoo Finance reported, Oracle laid out a cloud strategy that keeps heavy spending on Nvidia's AI servers while widening its list of chip suppliers. Oracle Cloud Infrastructure, its cloud computing arm, plans a large rollout of Nvidia's new Vera Rubin systems. The scale is described as several gigawatts, a measure of electricity capacity.
Oracle also becomes the first outside large cloud provider to offer Google's TPU processors. TPUs are chips Google designed for AI work. Oracle's stated aim is to avoid leaning too hard on one hardware maker.
Reading the $95 billion figure
The headline links $95 billion to Rubin purchases, but the report is more careful. Oracle did not disclose exact deal amounts. The $95 billion appears as the upper end of the capital spending range Oracle restated for its next fiscal year, which covers all its infrastructure building.
For scale, the report recalls an earlier estimate from Nvidia chief Jensen Huang that each gigawatt of buildout can bring Nvidia up to $40 billion in revenue. That is a past rule of thumb, not a figure for this deal.
A broader menu of chips
Oracle will put its own Acceleron networking chips into some equipment instead of Nvidia's networking parts. Google's eighth-generation TPU is due to be offered on Oracle's cloud at a later date. Chips from SambaNova, which Intel backs, and Cerebras, which focuses on inference (running a finished AI model to answer requests), are on the list as well.
Oracle's cloud president, Mahesh Thiagarajan, said more options let companies balance performance against budget. Cerebras' chief executive said the tie-up should greatly widen its base of top-tier customers.
Who pays, and what is unclear
Oracle expects more end customers to cover chip purchase costs themselves as competition in AI inference grows. That would ease pressure on Oracle's cash flow.
Missing from the report are delivery timing, how much goes to each chip type, and what customers will pay. Treat those as open questions.
What to watch
- Will Oracle publish deal terms, delivery timing or pricing for the Rubin and TPU capacity?
- Will other large cloud providers also start offering Google's TPUs?
Try it
If your company buys AI through a cloud provider, ask which chips run your workloads and whether you can move between providers if capacity gets tight.
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by Yahoo Finance · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.