Oracle Adds Google and Cerebras AI Chips While Expanding Nvidia Deal — What Is Available and What Is Still Planned

October 27, 2026 — Oracle’s latest artificial-intelligence infrastructure announcements point to a strategic change in cloud computing: the company wants to sell customers a choice of AI processors, even as it deepens its relationship with market leader Nvidia. The important distinction is between capacity being announced, hardware already available and future services that still have to be deployed.

Nvidia remains central, but no longer the only route

Oracle and Nvidia said they plan to expand Oracle Cloud Infrastructure (OCI) Superclusters using Nvidia’s Vera Rubin NVL72 systems to what they describe as multi-gigawatt scale. Their October 27 announcement also says Vera Rubin NVL72 and Vera CPUs are generally available on OCI, while other configurations are planned. Those statements should not be confused with a claim that every planned multi-gigawatt installation has already been built.

The companies describe a division of labor: GPUs accelerate demanding AI training and inference, while CPUs handle tasks such as code execution, retrieval and tool calls in AI-agent workflows. Oracle’s Acceleron networking is also part of the proposed architecture. The arrangement reflects a practical challenge for cloud providers: powerful chips deliver little economic value if networking, electricity or customer demand becomes a bottleneck.

Google TPUs and Cerebras broaden the menu

Separately, Oracle and Google announced plans to bring Google’s Tensor Processing Units to OCI, making Oracle the first hyperscale cloud provider outside Google to offer them, according to the companies. Oracle and Cerebras also announced plans to make Cerebras’ fast AI-inference services available through OCI. These are partnership announcements, not evidence that every customer can already provision the new services in every region.

Why does this matter? Training a frontier model and serving millions of individual user requests place different demands on processors. For customers, access to several architectures could improve bargaining power and allow more specialized choices. For Oracle, however, a broader catalog creates integration, support and capital-allocation challenges as well as opportunity.

Three questions investors should ask

First, timing: Which announced systems are generally available today, and which remain future commitments? Vendor announcements often combine both. Second, utilization: Can Oracle fill expensive computing capacity with paying workloads rather than simply reserve more chips? Third, financing: How much of the infrastructure investment is supported by customers, and how much remains Oracle’s own obligation?

Reuters reported that the value of the expanded Nvidia arrangement was not disclosed and that Oracle had not specified how it would finance the newly discussed purchases. The report also noted Oracle’s previously stated fiscal 2027 capital-spending target of roughly $90 billion to $95 billion. That figure is a company-wide spending target, not a disclosed price for the Nvidia, Google or Cerebras announcements.

The bottom line: Oracle is not abandoning Nvidia. It is trying to make OCI a multi-architecture AI platform. Whether that strategy improves customer choice and financial returns will depend on actual deployments, pricing, utilization and delivery—not headline capacity alone.

Sources and verification: Nvidia and Oracle joint announcement, October 27, 2026; Oracle–Google TPU announcement, October 27, 2026; Oracle–Cerebras announcement, October 27, 2026; Reuters independent reporting, October 27, 2026.

Featured image: illustrative data-center photograph by Tyler via Unsplash; it does not depict a specific Oracle facility. No event-specific video was embedded because rights and exact-event suitability were not verified.

Please follow and like us: