FX.co ★ Oracle scales Nvidia partnership to multi‑gigawatt build‑out while buying chips from rivals

Oracle has announced a major expansion of its strategic partnership with Nvidia to deploy next‑generation AI cloud infrastructure. Reuters reports the companies agreed to scale Oracle Cloud Infrastructure’s (OCI) supercluster capacity to “multi‑gigawatt” levels using Nvidia’s latest server architecture, Vera Rubin. At the same time, Oracle is diversifying its hardware mix by integrating chips from Nvidia rivals, including Google, Cerebras Systems, and SambaNova Systems.

Financial terms of the agreement were not disclosed. Nvidia CEO Jensen Huang has previously suggested that Vera Rubin‑generation systems could generate as much as $40 billion in revenue for the chipmaker per gigawatt of data center power deployed. Oracle’s move to add third‑party suppliers is meant to support the inference layer — the part that runs trained models and answers user requests. Unlike the training market, this layer is much more price‑sensitive and uses a wider variety of specialized accelerators, while Nvidia still dominates training.

The scale of the planned capacity expansion has sparked investor interest in how it will be financed. Oracle, which is due to host an investor day this week, has not yet detailed the financing instruments or reserves that will pay for the costly hardware purchases. Analysts expect management to provide clarity on the capital‑expenditure structure and liquidity plans during the upcoming investor meeting.