Oracle's cloud will offer Google's AI chips from 2027, a rival to Nvidia
Oracle said it will put Google's TPU chips into its cloud from 2027, giving AI customers a second chip supplier next to Nvidia.
The short version
- Oracle announced on October 27, 2026 that Google TPUs will arrive on its cloud from 2027.
- Planned models: TPU 8t for training AI and TPU 8i for running it (inference).
- Industry reports call OCI the first big cloud outside Google to offer these chips.
- Oracle gave no OCI prices for the TPUs; Nvidia stays its primary partner.
- Oracle's stock rose about 5 percent to $132.28 after the news.
Why it matters
Until now, Google's chips were mostly reachable only through Google's own cloud. A rival cloud carrying them gives big buyers a second supplier next to Nvidia, which mostly means more bargaining power. For most readers the effect on AI prices is indirect and slow, and nothing changes on your bill today.
What Oracle announced
As BornCity reported, Oracle said on October 27, 2026, at its Oracle AI World conference in Las Vegas, that Google's TPUs will be offered on Oracle Cloud Infrastructure (OCI) from calendar year 2027. A TPU, or Tensor Processing Unit, is a chip Google designed specifically for AI workloads. The rollout is meant to stretch over several years.
Two models are named. The TPU 8t is meant for training, the heavy process of teaching an AI model. The TPU 8i is meant for inference, which is the everyday work of running a finished model to answer questions. Industry reports cited in the coverage say OCI is the first hyperscale cloud, meaning one of the very large cloud providers, outside Google itself to offer these accelerators.
How the deal is set up
Karan Batta, a senior vice president at OCI, described it as a multi-year partnership at multi-gigawatt scale. A gigawatt is a measure of electrical power, so the phrase signals very large data centers. He said Oracle will receive each new TPU generation on the day Google releases it.
Oracle's job under the agreement is the physical side: power, cooling and networking. Mahesh Thiagarajan, OCI's president, said customers will be able to choose among several leading AI computing designs inside one cloud, picking by purpose, cost and performance.
What is missing is the price. Oracle has not named OCI rates for the TPUs. For scale only, Google Cloud lists $13.80 per hour on demand for older tpu7x configurations, but that is Google's own price for a different setup, not a preview of Oracle's.
What Google says the chips can do
The performance claims come from Google, not from independent tests. Google says the TPU 8t offers up to 2.7 times more performance per dollar than the previous generation, and nearly three times the computing power per pod, which is a large cluster of chips linked together. For the TPU 8i, it cites up to 80 percent more inference performance per dollar.
Google also says the eighth generation reaches up to 12.6 PFLOPS of native FP4 per chip. PFLOPS counts quadrillions of calculations per second, and FP4 is a compact number format that AI chips use to move faster. The chips also use optical circuit switching, which routes data between chips with light-based switches.
Treat these as vendor figures until customers report real-world results.
Nvidia stays, but the field widens
Oracle did not present this as a break with Nvidia. It still calls Nvidia its central partner for general-purpose AI work. Nvidia's Vera Rubin NVL72 systems and Vera processors are already available, the HGX Rubin NVL8 has no date yet, and Oracle plans multi-gigawatt OCI superclusters built with Nvidia technology.
Alongside Google, Oracle plans to bring in accelerators from Cerebras, SambaNova and Positron, and expects systems from AMD, Intel and OpenAI's Jalapeño. Positron Atlas for inference is already generally available. Cerebras systems are slated for 2027, and the SambaNova SN50 has been announced without a date.
The picture is of a cloud that wants to be a menu of chip options rather than a single-supplier shop.
The money and the wider ties
The same day, Oracle and OpenAI announced closer cooperation. Plans include linking business applications to ChatGPT and to Codex, OpenAI's coding tool. US customers are expected to be able to pay for selected OpenAI services through the Oracle Marketplace using existing Oracle Universal Credits. OpenAI's Derrick Choi said that internally, 130,000 Oracle employees use ChatGPT and more than 95,000 use Codex. Oracle and Google are also expanding offerings around Gemini, BigQuery and database services.
On the numbers, OCI posted revenue of $7.4 billion in the first quarter of fiscal 2027, up 121 percent from a year earlier, with a backlog of $664 billion. Oracle's stock rose about 5 percent to $132.28 after the announcement. Media reports also say OpenAI plans to buy Google TPU hardware directly.
On Google's side, finance chief Anat Ashkenazi said on October 28, 2026, in the earnings call, that most TPU revenue will arrive in 2027.
What to watch
- What will Oracle charge for TPU access, and how much capacity will it actually have in 2027?
- Will independent tests confirm Google's performance-per-dollar claims for the TPU 8t and 8i?
- Will OpenAI follow through on buying Google TPU hardware directly, as media reports say?
Try it
If your company buys cloud AI services, ask your provider which chip types it runs and whether it publishes prices per chip type. Today the answer is mostly Nvidia.
Sources and links
- ReportingBornCity: the article we sourced
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 BornCity · 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.