Blackwell AI Chip Overheating Allegations? Dell and Partners Respond

Recent media coverage, notably from The Information, reported that NVIDIA's Blackwell architecture AI chips, designed for AI and high-performance computing (HPC), are facing overheating issues when used in servers equipped with 72 processors. These high-density servers, with power consumption reaching up to 120 kW per rack, allegedly experienced thermal management challenges that led to performance limitations and potential hardware risks. As a result, concerns were raised regarding possible delays in server deployment schedules for data centers.

This report, which gained traction across international media, caused notable market unease, negatively affecting related stock prices. In response, Dell Technologies, a leading server manufacturer, took to social media platform X to state that "the world's first NVIDIA GB200 NVL72 server has officially shipped," directly countering claims that shipments were delayed due to overheating.

Additionally, market sources from key NVIDIA partners Foxconn and Quanta affirmed that shipment schedules remain on track. Foxconn's internal reports indicate no shipment delays related to overheating, and Quanta reiterated during a financial briefing that limited shipments of the GB200 series servers would commence by year-end, with full-scale production slated for Q1 2025. Both companies emphasized that these timelines are unchanged.

Foxconn also announced its collaboration with NVIDIA to develop next-generation AI factories using the Omniverse platform and digital twin technology. This partnership aims to redefine manufacturing processes and supply chain management, opening up new avenues for deploying advanced equipment globally and enhancing enterprise resilience. These developments indicate that the shipment of related servers has not been impacted by the alleged overheating concerns, and Foxconn will proceed as planned with their deployment initiatives.