Nvidia stock price falls by more than 4% as overheating issues in its Blackwell AI servers lead to order delays

Table of Contents
Nvidia’s Blackwell AI servers are reportedly facing delays in orders as the company struggles to resolve overheating and architectural flaws.
For those who aren’t aware, Nvidia’s Blackwell AI servers were expected to enter volume production in Q4 2024. However, reports reveal that the Blackwell GPU servers for AI are experiencing significant overheating issues, to the point where Nvidia might have to redesign them. While Nvidia initially claimed that the overheating issues and architectural flaws had been resolved, this good news was short-lived. A new report from The Information suggests that these problems persist, with the Blackwell AI servers still “glitching out.” This seems to contradict the official word from Nvidia CEO Jensen Huang who said that “production is progressing smoothly“.
🚀 Save Up to $1,200 on the Samsung Galaxy S25!
Pre-order now and save big with trade-in and Samsung credit. Limited time only!
*Includes trade-in value + $300 Samsung credit.
The report goes further, claiming that the first significant shipment of GB200 NVL72 servers – rack versions with 72 GPUs per unit – has been plagued by overheating and glitching issues. While the exact cause remains unclear, the report suggests that the main problem lies in the “way chips connect.” This aligns with previous claims that the issue is related to TSMC’s advanced packaging technology, CoWoS, which refers to the “chip connection.”

Nvidia’s AI server sales slow down
The problems with Nvidia’s Blackwell AI servers have ultimately frustrated major customers like Microsoft, Amazon, Google, and Meta, leading them to cut orders that were initially estimated at over $10 billion. Nvidia has taken a hit, and following this news, its stock price has dropped by more than 4%, as reported by Reuters. The situation will be a concern for Nvidia and its AI business, as supply chain issues in these products can severely impact the company's finances.
According to the report, companies are now opting to either wait for improved versions or use older-generation chips, such as Hopper, until Nvidia resolves Blackwell’s issues. For instance, OpenAI requested that Microsoft use Hopper due to delays in deploying the new models.
While the exact sales impact of the Blackwell design flaw remains unclear, Nvidia must address these problems soon, as the success of Blackwell is now at risk. On the other hand, Jensen Huang stated in November that projected revenue for the fourth fiscal quarter would exceed earlier expectations. He also denied rumors of problems with liquid-cooled servers using Blackwell GPUs.