Supermicro Has Shipped Liquid-Cooled Rubin AI Racks
The hardware manufacturer has begun shipping server racks featuring NVIDIA's latest Vera Rubin platform.
Updated on Sept. 23, 2026 in Data Centers
Live Poll
Do you trust companies currently leading the AI hardware buildout as stable long-term investments?
Supermicro has officially commenced shipments of its new data center rack units featuring the NVIDIA Vera Rubin GPU platform. The rollout comes as the company seeks to meet surging demand for high-performance artificial intelligence infrastructure.
Why it matters
The new architecture allows for scalable deployments ranging from 5 MW to gigawatt-class clusters to support complex AI workloads. To facilitate this expansion, Supermicro has increased its manufacturing capacity across facilities in the United States and Taiwan.
The NVL4 system utilizes direct liquid cooling to manage a 3.2 MW power envelope across eight racks. Each rack in the broader NVL72 architecture contains 72 GPUs and 36 Vera CPUs to handle extreme thermal loads.
The players
Supermicro
This global technology company specializes in high-performance server solutions and data center hardware.
NVIDIA
This multinational technology company is the primary developer of the GPU platforms used in artificial intelligence infrastructure.
The details
Supermicro employs DLC-2 direct liquid cooling technology to maintain system stability within the power-intensive 3.2 MW NVL4 configuration. The company is also providing pre-validated system blueprints to streamline engineering requirements for data center operators.
Timeline
June 22, 2026: The NVL4 DCBBS configuration was officially announced.
September 23, 2026: News coverage confirmed that rack shipments have begun.
Second half of 2026: Vera Rubin products are expected to reach full production volume.
The Tech Race
The deployment of the NVIDIA Vera Rubin platform marks a definitive shift toward direct liquid cooling as the standard for high-density AI computing. This evolution represents a departure from traditional air-cooled systems that struggle to manage the thermal loads of next-generation GPU clusters.
Customers can expect reduced engineering timelines due to the provided pre-validated system blueprints. The massive power requirements of these clusters may also necessitate infrastructure upgrades for facilities adopting the new 3.2 MW configurations.
The takeaway
The move to liquid cooling highlights the physical limitations current data centers face when scaling for AI. Companies looking to implement these systems should prioritize evaluating their existing facility power and cooling capacities before deployment.
What happens next
Full production volume for the Vera Rubin product line is scheduled for the second half of 2026.
Further reading
For more on evolving hardware standards, visit our Data Centers section.
Live Poll
Do you trust companies currently leading the AI hardware buildout as stable long-term investments?
