Oracle Deepens Nvidia AI Chip Pact, Adds Google and Cerebras to Cloud Offerings

Oracle announced on Tuesday it is significantly expanding its cloud partnership with Nvidia to incorporate next-generation AI servers, while simultaneously diversifying its hardware offerings to include chips from competitors such as Alphabet's Google and Cerebras Systems.

Expanding the Nvidia Partnership

Oracle and Nvidia said the cloud firm will scale up its infrastructure to a "multi-gigawatt scale" using Nvidia’s forthcoming "Vera Rubin" generation of AI servers. While the companies did not disclose the financial terms of the agreement, Nvidia CEO Jensen Huang has previously stated that such systems could generate up to $40 billion in revenue per gigawatt for the chipmaker, according to a Reuters report.

In a move to integrate its own technology, Oracle noted that even these new Nvidia-based systems will utilize Oracle's in-house "Acceleron" networking chips, forgoing some of Nvidia's own networking components.

Diversifying with Competitors

To provide clients with more flexibility, Oracle is broadening its portfolio of available AI accelerators. The company stated it will begin offering Google’s tensor processing unit (TPU) chips for the first time.

Ad

"Different AI workloads have different compute needs," Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, said in a statement. "That choice gives (customers) the flexibility to choose the right infrastructure for the performance, throughput and economics of each AI workload.”

Oracle also plans to offer chips from Intel-backed SambaNova Systems and Cerebras Systems. These companies specialize in hardware for AI inference, the process of running a trained AI model to generate responses, a market segment that is considered more competitive than the AI training market dominated by Nvidia.

Financial and Market Implications

The deal represents a significant win for Nvidia's smaller competitors. "That’s a massive expansion of our customer base of tier-one customers," Cerebras CEO Andrew Feldman told Reuters, highlighting the importance of being deployed at scale in a major cloud provider's data centers.

Oracle did not specify how it plans to finance the extensive chip purchases. However, during an earnings call last month, the company indicated that customers are increasingly helping to fund the hardware costs and maintained its capital spending forecast of $90 billion to $95 billion for its 2027 fiscal year.