Nvidia Blackwell GPU Launch Delayed by Design Flaw
A late-stage design flaw in Nvidia's highly anticipated Blackwell GPUs pushes the launch back by at least three months. The delay disrupts massive orders from major tech companies relying on the chips for next-generation AI data centers.
Nvidia informs major customers like Microsoft, Meta, and Google that the launch of its new Blackwell B200 GPUs experiences a delay of at least three months. A manufacturing partner discovers a design flaw unusually late in the production process, preventing the highly anticipated artificial intelligence chips from shipping on schedule. The Blackwell series serves as the successor to Nvidia's wildly popular H100 chips and promises up to 30 times better performance.
The technical issue stems from the processor die that connects two Blackwell GPUs on a single GB200 Superchip. Taiwan Semiconductor Manufacturing Co. identifies this problem during production, forcing Nvidia to revise the die design. The company needs several months to conduct new production tests with TSMC before mass manufacturing can resume.
This unexpected setback causes significant disruption for tech giants that plan to deploy new AI data centers early next year. Google reportedly orders over 400,000 GB200 chips in a deal worth more than $10 billion, while Meta and Microsoft place similarly massive orders to power their expanding AI infrastructures. As a result of the news, Nvidia's stock drops more than 7% in Monday morning trading.