Nvidia’s Next-Gen Blackwell GPUs Hit Cooling Snag, Delaying AI Chip Rollout to 2025
Staff Editor Nvidia Faces Setbacks as Unreleased AI Chips Overheat, Delaying Shipments
Nvidia, the leading manufacturer of graphics processing units (GPUs), is reportedly grappling with overheating issues in its unreleased Blackwell GPUs, causing significant delays in shipment. The company has informed major clients, including Microsoft, that deliveries may be postponed until early 2025.
The overheating problem becomes particularly acute when these high-performance chips are installed in server racks designed to accommodate up to 72 units. In response, Nvidia has requested suppliers to redesign server racks to mitigate the issue, further contributing to the delay.
This setback adds to a series of challenges Nvidia has faced since announcing the Blackwell GPUs in March. The delay poses significant obstacles for AI companies that have invested heavily in server farms for AI product development.
Despite the setbacks, demand for Nvidia’s Blackwell GPUs remains high. The chips are touted as extremely powerful, with performance claims of being 30 times faster than previous models for AI applications. Industry insiders have described pre-order interest as “insane,” even with the GPUs’ high cost.
However, rumors of design flaws have circulated in recent months. Nvidia CEO Jensen Huang acknowledged a design flaw affecting production yield, taking responsibility for the issue and its contribution to shipment delays. The company maintains that the overheating problem is manageable and part of normal engineering processes.
The cooling challenges extend beyond the largest server configurations. Even smaller racks with 36 GPUs have experienced overheating, underscoring the severity of the issue. Unlike the air-cooling systems commonly used in AI data centers, the server racks for Blackwell GPUs require water cooling, adding another layer of complexity to the problem.
As anticipation for new AI products grows and customers face continued delays, pressure is mounting on Nvidia to resolve these issues promptly. The ongoing problems with Blackwell GPUs could potentially impact the rollout of new AI technologies, and customer dissatisfaction with repeated delays adds to the challenges faced by Nvidia in meeting market demand.
The tech industry will be closely watching Nvidia’s efforts to overcome these obstacles and deliver on the promise of its next-generation AI chips.
