Nvidia says it is donating its Vera Rubin NVL144 server rack architecture to the Open Compute Project and working with 70+ partners on “gigawatt AI factories”
Nvidia Corp. took to the stage at the 2025 OCP Global Summit in San Jose today to talk about how it's collaborating …
Context & Ripple Effects
Nvidia’s collaboration with server makers and cloud providers has long paired its AI software and chips with partner-led system delivery. The company’s earlier server-maker and cloud-provider partnerships provide the precedent for broadening a hardware platform through an ecosystem rather than supplying components alone.
The move also follows Nvidia and OpenAI’s commitment to deploy at least 10 GW of Nvidia systems for AI infrastructure, making large-scale deployment a concrete customer requirement rather than a distant design target. OCP provides a venue to turn that requirement into a more repeatable rack-level blueprint.
First-order effects
- OCP members and Nvidia’s more than 70 partners gain access to a common Vera Rubin NVL144 rack architecture, reducing the need for each deployment partner to start with a wholly proprietary rack design.
- Nvidia shifts the immediate conversation from individual accelerators to a validated, partner-supported AI-factory building block, while retaining influence through the platform and ecosystem around it.
Second-order effects
- Server, networking, power and cooling suppliers participating in the ecosystem can align products around a shared rack specification, increasing pressure on rivals to offer similarly integrated and deployable system designs.
- Large AI-infrastructure buyers may gain more choice among compatible deployment partners, while increasingly evaluating vendors on factory-scale integration and support rather than chip performance alone.
Third-order effects
- If adoption follows, open rack architectures could become a key layer of AI-infrastructure standardization: Nvidia can extend its platform reach by making its preferred system design easier for a broad supply chain to build around.
- The industry’s competitive unit would continue shifting toward the AI factory—compute, networking, power, cooling and operations together—though OCP adoption will determine how portable the design becomes across suppliers.
The trend: AI infrastructure is being productized as an interoperable, factory-scale system rather than purchased solely as discrete compute hardware.