NVIDIA Vera Rubin Enters Full Production, Liquid Cooling Supply Chain Poised for Realization

On the morning of June 1, 2026, NVIDIA Founder and CEO Jensen Huang confirmed during the GTC Taipei 2026 keynote that the Vera Rubin platform, centered on the Vera Rubin NVL72 rack-scale system, has officially entered full production. Deliveries to system integrators and cloud partners are expected to begin this fall.

 

Huang also revealed that the Vera Rubin supply chain is twice the size of the previous-generation Grace Blackwell, with over 350 partner factories across more than 30 countries. This is not just a “next-generation chip PPT”—it is a production-ready industrial cadence.

 

From a platform engineering perspective, the Rubin generation continues—and aggressively expands—the roadmap established by Blackwell. The NVL72 adopts single-phase direct liquid cooling (DLC), designed for supply water temperatures up to 45°C, and emphasizes fanless, highly modular rack designs. For end customers, this signals that liquid cooling is no longer an optional discussion. Installing a Rubin cluster now requires liquid cooling infrastructure to be integrated into project timelines; without pre-installed secondary-side and primary-side modifications, the racks simply cannot be deployed.

 

High-Value Components and Engineering Precision

Third-party teardown reports of the Nvidia GB200 NVL72 and related broker analyses indicate that core liquid cooling hardware—cold plates, CDUs, manifolds, and UQDs—can account for tens of thousands of dollars per rack. For the Nvidia GB300, which uses independent cold plates and more quick-connect fittings, component value may reach up to $100,000 per rack.

 

Rubin’s evolution is not only about higher thermal density; structural and precision upgrades are central. Cold plate microchannel technology, tighter assembly tolerances, integrated manifolds, and more reliable CDU-to-piping interfaces shift value from “copper and aluminum area” to precision components, sealing lifespan, and verifiable delivery records. For the supply chain, H2 2026 marks the transition from POC to full production delivery and engineering deployment.

 

 

 

Supply Chain Structure: Certification Over Capacity

NVIDIA has implemented stricter control over liquid cooling interfaces and assembly tolerances in the Rubin generation. Cold plate design must conform closely to official thermal design specifications. Entry into NVIDIA’s AVL/RVL system, combined with a track record of large-scale, incident-free delivery, defines supplier tiers.

 

CDU Systems Capture the Highest Margins
A CDU failure can threaten the entire rack’s GPU assets. Reference designs rely heavily on top-tier vendors with full-chain engineering capabilities.

Vertiv is most visible in GB200/Rubin reference designs.

nVent differentiates through quick-connect fittings and manifold processes.

Delta integrates liquid cooling with 800V-class power and energy storage for rack-level systems.

 

Cold Plates, Manifolds, and UQDs: Certification Gatekeepers
Achieving near-zero leakage under high pressure and flow sets a high certification bar.

Taiwanese leaders AVC and Auras are recurrent cold plate suppliers.

Boyd and CoolIT hold long-term positions in high-end cold plates and piping patents.

In China, Readore (a subsidiary of Lingyi Intelligent) holds dual AVL/RVL certifications covering cold plates, manifolds, and UQD fittings, providing a verifiable path into GB200/300 supply.

 

System Integrators
Domestic demand is largely driven by operators and major internet companies’ self-built data centers.

InWin provides full-chain liquid cooling integration.

Gaolan operates dual lines for cold plates and immersion cooling.

Sugon Digital Innovation specializes in immersion and HPC/AI environments with extreme PUE requirements.

Lian Li Liquid Cooling provides end-to-end liquid cooling solutions for data centers—ranging from cold plates, liquid-cooled server cabinets, and CDUs to containerized liquid-cooled data centers. These solutions reduce the product PUE to below 1.15 and achieve a comprehensive energy savings rate exceeding 30%, while the products themselves have obtained authoritative international certifications such as RoHS, CE, and UL.

 

 

Market Spillover and Infrastructure Opportunities

Looking beyond individual racks, Rubin’s mass production will ripple outward. At the primary-side and facility level, 45°C supply water favors dry coolers and natural cooling architectures, reducing the priority of traditional large chiller installations in new AIDC deployments. This benefits vendors capable of total-package design and system optimization.

 

Key components—including precision valves, isolation valves, high-performance sealing materials, and DI water loops with monitoring—become critical to operational cost, as leak-free, reliable, and controlled water quality directly impacts uptime. Distributed monitoring, flow, pressure, and temperature sensors, along with DCIM and flow assurance logic, are moving from “nice-to-have” to mandatory acceptance criteria. Coupled power-thermal integration (HVDC, energy storage) allows vendors delivering combined “power + cooling + control” solutions to command higher negotiation leverage.

 

 

Capital Market Implications

For investors, Vera Rubin’s full production signals a shift from speculative expectations to performance verification in the liquid cooling industry. Only a small group of suppliers with the right certifications and engineering capabilities will capture the growth dividends, while most others remain in highly competitive, low-margin manufacturing roles.

 

Conclusion: Vera Rubin marks a turning point where liquid cooling moves from optional technology to a critical enabler for next-generation AI infrastructure, and the industrial supply chain enters a period of measurable delivery and operational impact.

 

 

 

FAQ:

1. What is the Vera Rubin NVL72 platform?
The Vera Rubin NVL72 is NVIDIA’s next-generation rack-scale system, featuring single-phase direct liquid cooling (DLC) and modular, fanless rack design. It supports supply water temperatures up to 45°C and is designed for high-density AI and HPC workloads.

 

2. Why is liquid cooling critical for Vera Rubin deployments?
Liquid cooling is no longer optional for Vera Rubin clusters. The platform’s thermal density and design require pre-installed secondary and primary liquid cooling infrastructure to ensure timely deployment and stable operation of GPU racks.

 

3. Who are the main suppliers in the Vera Rubin liquid cooling ecosystem?
Key suppliers include leading global and regional vendors for CDUs, cold plates, manifolds, and UQDs. Examples include Vertiv, Delta, nVent, AVC, Auras, Boyd, CoolIT, and Readore, all of which meet NVIDIA’s strict certification and quality requirements.