Editorial illustration for Oracle Adds Google TPUs, Cerebras to AI Compute Options
Oracle Expands AI Chips Beyond Nvidia with TPUs
Oracle Adds Google TPUs, Cerebras to AI Compute Options
Oracle used its AI World conference in Las Vegas this week to announce it's widening the menu of chips available on Oracle Cloud Infrastructure, moving beyond its heavy bet on Nvidia. The company confirmed plans to scale Nvidia's Vera Rubin systems to multi-gigawatt capacity, with Vera Rubin NVL72 systems and Vera CPUs already live on OCI. But the bigger news is who else is joining the lineup.
Google's Tensor Processing Units are coming to OCI starting in 2027, letting Oracle customers train and run models on Google's custom silicon without migrating to Google Cloud. Cerebras Systems is also joining as an inference partner, bringing its wafer-scale chip design, which merges computing and memory on a single piece of silicon, to OCI that same year.
The moves signal Oracle's attempt to position itself as a neutral ground for AI compute, rather than a single-vendor shop. CPUs in this expanded setup aren't just along for the ride either. They're being tasked with running code, fetching data and coordinating tool calls between model requests, the plumbing work that keeps AI agents and applications functioning between the heavy lifting done by GPUs and TPUs.
Why this matters
Oracle is betting that chip neutrality sells better than chip exclusivity. By putting NVIDIA's Vera Rubin systems, Google's TPUs, and Cerebras' inference hardware under one roof at OCI, Oracle is telling developers they don't have to pick a cloud based on which silicon they want. That's a real shift from the last few years, when choosing a chip often meant choosing a cloud vendor by default.
For teams building and training models, the practical upside is flexibility: you can tap Google's TPUs for training without signing up for Google Cloud, and Cerebras inference arrives on OCI in 2027, giving a multi-year runway to plan around. But 2027 is also the catch. This isn't live infrastructure you can rent tomorrow. It's a roadmap, and Oracle's roadmaps have to compete with NVIDIA's own release cycle and whatever Google and Cerebras do independently in the meantime.
Worth watching: whether Oracle can actually deliver multi-gigawatt scale on schedule, and whether pricing across three different chip architectures ends up simpler or just more confusing for customers trying to compare costs.
Related Reading
Browse all Google news →
Research & Benchmarks Microsoft AI Chief: "If It Isn't Safe We Shouldn't Build It
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic Google to End Free Access to Gemini AI for Non-Subscribers
LLMs & Generative AI
Related Topic NVIDIA and Google Cloud let developers scale AI from prototype to production
Market Trends
From Archives Maya Research Releases Open-Weight Speech Model With 12 Languages and 20 Voices
From Archives Newest AI Quest Maps the Human Brain, Builds on Fruit Fly Brain Map
From Archives DeepMind Genie 4 Modernizes Pre-LLM Agents Using Gemini Omni
LLMs & Generative AI
From Archives Google's Genie 4 world model adds audio, realistic worlds, real-time steering
Open Source
Common Questions Answered
What new AI chips is Oracle adding to its cloud infrastructure beyond Nvidia?
Oracle is adding Google's Tensor Processing Units (TPUs) starting in 2027 and Cerebras' inference technology to Oracle Cloud Infrastructure. This expansion allows Oracle customers to access multiple chip options on a single cloud platform rather than being locked into one vendor's hardware.
When will Google TPUs become available on Oracle Cloud Infrastructure?
Google's Tensor Processing Units are scheduled to come to OCI starting in 2027, according to Oracle's announcement at its AI World conference in Las Vegas. This timeline allows Oracle customers to plan their AI infrastructure investments with access to Google's specialized hardware for training and inference workloads.
What is Oracle's strategy with chip neutrality compared to traditional cloud vendor approaches?
Oracle is betting that offering chip neutrality—providing Nvidia's Vera Rubin systems, Google's TPUs, and Cerebras' inference hardware all on one platform—is more attractive than chip exclusivity. This represents a significant shift from previous years when choosing a specific chip often forced developers to select a particular cloud vendor by default, now giving teams flexibility to choose their preferred silicon without changing cloud providers.
Which Nvidia GPU systems are currently live on Oracle Cloud Infrastructure?
Oracle has confirmed that Nvidia's Vera Rubin NVL72 systems and Vera CPUs are already live on OCI, with plans to scale these systems to multi-gigawatt capacity. These latest-generation Nvidia GPUs provide Oracle customers with access to cutting-edge AI compute capabilities for training and inference tasks.
What practical benefit does Oracle's multi-chip approach provide to AI development teams?
By offering multiple chip options under one cloud platform, Oracle provides developers with flexibility to choose the hardware that best suits their specific AI workloads without being forced to switch cloud vendors. This chip-agnostic approach eliminates the previous constraint where selecting a particular processor often meant committing to a specific cloud provider's ecosystem.
Further Reading
- Oracle integrates ChatGPT, Codex into apps and expands cloud chip lineup - CNBC
- Oracle is Expanding the Accelerator Options in OCI for Enterprises - The Next Platform
- Oracle Expands Cloud Infrastructure Arsenal With Google, Cerebras Offerings - Yahoo Finance
- Oracle Brings Google TPUs to Oracle Cloud Infrastructure - Oracle
- Cerebras and Oracle Team Up to Bring Cerebras AI Inference to Oracle Cloud Infrastructure - Oracle
