Rare! Report: NVIDIA's latest AI chip delayed due to design flaws
NVIDIA's latest AI chip is delayed for three months or longer due to design flaws, which may impact some customers' plans. This is a significant event for companies like Meta Platforms, Google, and Microsoft, as they have ordered chips worth billions of dollars. NVIDIA is working with chip manufacturer TSMC to address the issue and plans to increase chip production later this year. It is very rare to discover design flaws before mass production
NVIDIA's latest AI chip in the Blackwell series may face delays in release.
According to The Information, sources familiar with the matter stated that NVIDIA's upcoming artificial intelligence chip will be delayed by three months or longer due to design flaws.
This could impact customers such as Meta Platforms, Google, and Microsoft, who collectively ordered chips worth billions of dollars.
NVIDIA has declined to comment on the delay, but mentioned that customers are testing samples of the Blackwell chip, with "production expected to ramp up later this year."
Major design flaws discovered before mass production are not common
The Information cited individuals involved in the production of the Blackwell chip, stating that design issues with Blackwell have surfaced in recent weeks, as engineers at TSMC discovered defects while preparing for mass production.
The GB200 chip consists of two interconnected Blackwell GPUs and a Grace central processing unit. The defect involves a processor chip (a silicon wafer used to accommodate chip circuits) that connects the two Blackwell GPUs. This obstacle has lowered TSMC's chip yield for NVIDIA and could even potentially halt production.
Reports indicate that NVIDIA is conducting new trial production runs with its chip manufacturer TSMC to address the issue.
As per the original plan, TSMC was set to begin mass production of Blackwell chips in the third quarter and deliver them to NVIDIA starting in the fourth quarter, with servers expected to ship in subsequent quarters if no further issues arise.
Analysts believe that discovering major design flaws before mass production is highly unusual. This is because multiple production tests and simulations are required in the early stages to ensure product feasibility and a smooth manufacturing process.
If the upcoming AI chips like B100, B200, and GB200 are delayed by three months or longer, some customers may not be able to deploy large chip clusters as planned in the first quarter of 2025.
The design flaws will also impact the production and delivery of NVIDIA NVLink server racks, as companies involved in server operations must wait for new chip samples before finalizing server rack designs.
Stay tuned for further updates