Nvidia's New AI Chips Are Reportedly Overheating in Server Farms

Hello!

Graphic Pushback Unit

Nvidia's unreleased AI chips are reportedly overheating, raising concerns among customers that their already-delayed shipments could face further setbacks.

As The Information reports, the company’s powerful Blackwell graphics processing units (GPUs) are overheating when installed in server racks that can accommodate up to 72 units.

According to Nvidia employees involved in the release, as well as customers and vendors familiar with the situation, the company has repeatedly requested that suppliers redesign the racks to resolve the overheating issue.

The problem has become so significant that Nvidia informed Microsoft this week of at least a three-month delay in shipments. This marks the latest in a series of postponements that have affected the Blackwell chips since their unveiling in March.

The delays come at a challenging time, as AI companies continue to invest heavily in expanding server infrastructure to support the training and deployment of next-generation models.

Design and Demand

Nvidia positions its next-generation GPUs as extremely powerful, delivering up to 30 times the performance of previous models in AI applications. CEO Jensen Huang told CNBC last month that demand for Blackwell chips has been “insane,” with customers rushing to pre-order units priced at tens of thousands of dollars each.

Despite the strong interest, rumors of design flaws have circulated for months. Huang partially confirmed one issue during an October 23 press conference, as reported by Reuters:

“We had a design flaw in Blackwell, it was functional, but the design flaw caused the yield to be low. It was 100 percent Nvidia’s fault.”

While that statement referred to a separate production challenge, it has contributed to ongoing shipment delays. A Nvidia spokesperson described the latest overheating reports as routine, noting that “the engineering iterations are normal and expected.”

The 72-GPU rack weighs approximately 3,000 pounds and requires liquid cooling, unlike the air-cooling systems used in many existing AI data centers. The Information noted that Nvidia had already encountered overheating problems with a smaller 36-GPU configuration.

As excitement around new AI products continues to build, Nvidia faces mounting pressure to deliver on its commitments. Customers have already experienced multiple delays, and the latest issue is unlikely to improve sentiment.

Also read:

• Perplexity AI Launches Comet Browser: A New Era of AI-Powered Browsing
• Top 6 Reasons to Keep Track of Business Expenditures
• VPNs the New Cybersecurity Trend in 2026

Thank you!
Join us on social networks!
See you!