Oracle, in an effort to reduce its dependence on Nvidia, adds AI chips from Google, Cerebras and SambaNova

Oracle (ORCL) is diversifying its AI cloud infrastructure by signing deals to integrate chips from Google (GOOG, GOOGL), Cerebras (CBRS) and SambaNova (SANS.PVT) onto its networks as it seeks to diversify beyond Nvidia’s processors.

The move marks the first time Google has sold its tensor processing units (TPUs) for use in a third-party hyperscaler’s data center rather than simply leasing capacity to enterprises.

Cerebras’ massive chips will provide AI inference capability and AI agent processing power, while SambaNova’s SN50 processor will handle agentic inference.

Oracle said its decision to tap processors from Google, Cerebras and SambaNova provides its customers with the infrastructure tailored to their particular needs.

Companies are increasingly working to reduce their reliance on Nvidia’s artificial intelligence chips to avoid being stuck with a single vendor and help contain costs.

Google, Amazon (AMZN), Microsoft (MSFT), Meta (META), OpenAI (OPAI.PVT), and Anthropic (ANTH.PVT) currently own or are developing their own custom AI chips. Like Google, Amazon is looking to offer its chips to third-party data center companies.

In June, Amazon’s head of AI, Peter DeSantis, told Bloomberg that the company was entering discussions to start selling its chips to external data centers.

Nvidia has previously said it generates about half of its revenue from hyperscalers, making Google’s decision and Amazon’s potential efforts particularly concerning for the company.

But that doesn’t mean Nvidia is in immediate trouble. Even though the company’s customers are clearly moving into its own turf, they’re still spending billions to buy Nvidia’s own chips.

Case in point: In addition to bringing chips from Google, Cerebras, and SambaNova, Oracle announced that it would bring Vera Rubin NVL72 superclusters and Nvidia’s Vera processors to its Oracle Cloud Infrastructure (OCI).

In addition to Oracle’s chip updates, the company announced its partnership with OpenAI on its Oracle AI Workspace. The platform allows users to access Oracle’s tools through ChatGPT, allowing them to describe a task they want to accomplish and let the AI ​​get to work.

The team also includes a new initiative called Oracle AI Security Mesh, which allows companies to access OpenAI’s cyber models to scan their software for flaws and vulnerabilities within OCI, patch them, and deploy them.

Correction: This is the first time Google has sold its chips for use in a third-party hyperscaler’s data center. An earlier version did not indicate that the chips had been sold to other data center operators.

Email Daniel Howley at dhowley@yahoofinance.com. Follow him on Twitter at @DanielHowley.