NVIDIA’s First Blackwell GB200 NVL72 Server Racks Shipped Amidst Reports of Thermal Issues

NVIDIA’s Blackwell NVL72 server racks are reportedly facing thermal issues, at a time when the firm has started mainstream supply to markets.

After Architectural Flaws, NVIDIA’s Blackwell Now Faces Server Rack Design Issues, But First NVL72 AI Server Racks With Liquid-Cooling  Shipping Out

NVIDIA’s Blackwell architecture, despite its high anticipation, has become a victim of development roadblocks since the very start. Initially, it was disclosed that the Blackwell chips were facing an issue with their interconnect technology onboard, which ultimately prompted supply chain companies like TSMC to revise their production methods. This issue did not only limit Blackwell’s supply, but now, according to a report by The Information, it looks like NVIDIA is yet again being faced with another hurdle.

The report claims that Blackwell’s server racks are a platform where multiple Blackwell interfaces are stacked against each other, combining to form the “GB200” AI servers. It is said that the problem is explicitly rising with the higher-end NVL72 configuration of Blackwell servers, which is said to be the more “important” ones for NVIDIA in terms of revenue generation and clientele interest. If the problem were to persist, we might look at a bearish outlook in terms of NVIDIA’s server revenue, given that the NVL72 servers are in significant demand.

Nvidia is working with leading cloud service providers as an integral part of our engineering team and process. The engineering iterations are normal and expected.

– NVIDIA via Reuters

It isn’t specified what aspect is causing the problem, but it is claimed that the issue could stem from the cooling methodology being utilized, and NVIDIA has urged its suppliers to cater to the issue immediately, given that it has disrupted the supply chain massively.

Given Team Green’s vast resources along with an extensive supply chain structure, we doubt that the problem will stay for much longer, as manufacturers would likely be in the process of server rack design revisions.

Interestingly, Dell, NVIDIA’s core partner, has already started to ship out Blackwell server racks, specifically the PowerEdge XE9712 servers, which are the industry’s first enterprise-ready NVIDIA GB200 NVL72 servers. So, the most likely move here by NVIDIA would be to limit the supply initially, unless the server rack problem is catered to.

Incredibly excited to ship the industry’s 1st enterprise ready @nvidia GB200 NVL72 server racks! These fully-integrated, liquid-cooled PowerEdge XE9712 racks are a huge milestone, the foundation for Dell AI Factory with NVIDIA to advance HPC & #AI innovation @CoreWeave pic.twitter.com/IJx3fCtJAJ

— Jeff Clarke (@JClarkeatDell) November 18, 2024

We’re proud to bring up the first @NVIDIA GB200 NVL72 from @Dell with NVIDIA Quantum InfiniBand, setting a new bar for AI infrastructure. This wouldn’t have been possible without the support of our valued partners at @Dell and @Switch. pic.twitter.com/Y8uYvccNCD

— CoreWeave (@CoreWeave) November 18, 2024

Despite temporary barriers, NVIDIA’s Blackwell lineup is still slated to be the “most successful” product in the company’s history, slated to generate billions in revenue. With the markets inclining towards building “AI clusters,” this has driven significant demand for the architecture, ultimately translating into higher turnouts for Team Green.