NVIDIA's highly anticipated Blackwell series of AI chips is reportedly facing delays of three months or more due to late-stage design flaws, potentially impacting the AI development plans of major tech companies like Microsoft, Google, and Meta.
The delay, first reported by The Information, affects the most advanced chips in the Blackwell line, including the B100, B200, and GB200 models. These chips were expected to drive significant revenue growth for NVIDIA and power the next generation of AI applications.
Sources familiar with the production process say the design issues were discovered unusually late, as Taiwan Semiconductor Manufacturing Company (TSMC) prepared for mass production. The problems reportedly involve a processor die connecting two Blackwell GPUs in the high-end GB200 chip, resulting in decreased production yield.
NVIDIA has reportedly informed key customers about the setback. Microsoft, which recently increased its Blackwell order by 20%, may need to push back plans to make Blackwell-powered servers available to OpenAI from January to March or later in 2025.
The delay could have far-reaching consequences for the AI industry. Tech giants have placed massive orders for Blackwell chips, with Google reportedly ordering over 400,000 GB200 chips and Meta investing at least $10 billion. These companies are banking on Blackwell's enhanced performance to drive significant improvements in large language models and other AI applications.
NVIDIA declined to comment specifically on the delay reports but stated that "production is on track to ramp" later this year. The company is now conducting new test production runs with TSMC to resolve the issues.
While chip delays aren't uncommon, the timing and scale of this setback are notable given the current AI boom and NVIDIA's dominant market position. The company's stock has soared on expectations for Blackwell, with some analysts projecting it could drive NVIDIA's data center revenue to over $200 billion in 2025.