Ahead of NVIDIA’s upcoming earnings report tomorrow, rumors have been circulating that the AI giant’s Blackwell processors are encountering overheating problems when installed in high-capacity server racks. However, Dell CEO Michael Dell soon posted a photo on social media, showcasing the first GB200 server shipment globally, effectively dispelling concerns about potential delays.
The 1st in the world @nvidia GB200 NVL72 server racks are now shipping. We are thrilled to deliver our liquid-cooled PowerEdge XE9712 to @CoreWeave. The AI rocket just got a massive boost! 🤖🚀🤝 pic.twitter.com/2QzlxbQE5f
— Michael Dell (@MichaelDell) November 18, 2024
In addition, another report from the Economic Daily News notes that NVIDIA’s two major contract manufacturers headquartered in Taiwan, Foxconn and Quanta, have both stated that the GB200 server cabinet delivery schedule remains unaffected.
The two companies are not only the primary manufacturers for NVIDIA’s GB200, but also key suppliers of the first batch of server racks, notes Economic Daily News.
It is worth noting that a source from a server chassis manufacturer cited by the report indicated that they had never heard of any heat dissipation problems with the server racks.
The source further explained that the chassis itself is not a heat source, as any heating would likely stem from the chips. The chassis’ primary function is to reinforce the server structure and support the liquid cooling system, he said.
According to the report, Foxconn confirmed that its internal shipment remains on schedule with no delays. It is understood that Foxconn plans to make small-scale shipments this quarter and ramp up production in the first quarter of next year.
Citing Foxconn Chairman Young Liu’s previous remarks, the report states that the demand for the GB200 series, as NVIDIA CEO Jensen Huang described, is “insane.”
Quanta agreed with the comments, saying that the shipping schedule for the GB200 series servers remains as projected, with mass production ramping up in the first quarter of next year, according to the Economic Daily News.
According to a report by The Information, NVIDIA’s Blackwell GPUs for AI and high-performance computing (HPC) face overheating issues in servers housing 72 processors, which can demand up to 120kW per rack. The report indicates that these challenges have led to design modifications and delays, sparking concerns from major customers such as Google, Meta, and Microsoft.
Read more
(Photo credit: NVIDIA)