TLDR:
- Nvidia confirmed and fixed a design flaw in their new Blackwell AI chips
- The flaw affected production yields but didn’t impact chip functionality
- TSMC helped Nvidia recover from the yield issues
- The chips will now ship in Q4 instead of Q2 as initially planned
- Nvidia CEO states demand for Blackwell chips is “insane”
Nvidia has successfully resolved a design flaw in its latest Blackwell AI chips, according to CEO Jensen Huang.
The issue, which caused production delays and lower yields, has been fixed with assistance from manufacturing partner Taiwan Semiconductor Manufacturing Company (TSMC).
The semiconductor giant encountered difficulties with its newest AI chip platform, which was originally scheduled to ship in the second quarter of 2024. The design flaw, while not affecting the chip’s functionality, resulted in lower production yields and pushed the shipping timeline to the fourth quarter.
Speaking at an event in Denmark, Huang took full responsibility for the setback.
“We had a design flaw in Blackwell,” he acknowledged. “It was functional, but the design flaw caused the yield to be low. It was 100% Nvidia’s fault.”
The complexity of the Blackwell project contributed significantly to the challenges faced during production. The platform required seven different types of chips to be designed from scratch and manufactured simultaneously, creating a complex production environment that needed careful coordination.
The news of the design flaw first emerged in August 2024, causing Nvidia’s stock to drop approximately 8%. The delay potentially affected major customers including Meta Platforms, Google, and Microsoft, who rely on Nvidia’s advanced AI chips for their operations.
During Nvidia’s second-quarter earnings call, the company disclosed that it had made changes to Blackwell’s GPU mask to improve production yield. However, Huang emphasized that no functional changes were necessary, as the core technology remained sound.
TSMC’s role in resolving the issue was crucial.
“What TSMC did was to help us recover from that yield difficulty and resume the manufacturing of Blackwell at an incredible pace,”
Huang explained. He dismissed media reports of tensions between Nvidia and TSMC as “fake news.”
The Blackwell platform represents a significant advancement in AI chip technology. The new chips combine two squares of silicon, equivalent in size to Nvidia’s previous offerings, into a single component that delivers 30 times faster performance for tasks such as chatbot response generation.
Despite the initial setback, market demand for the Blackwell chips remains strong. In October, Huang reported that the chips were in full production and described customer demand as “insane,” noting that “everybody wants to have the most, and everybody wants to be first.”
The company expects to generate several billion dollars in Blackwell revenue during the fourth quarter, indicating strong market confidence in the platform despite the earlier delays.
Nvidia’s stock performance has reflected investor optimism about the company’s prospects. While shares declined 2.9% following the news of the design flaw resolution, the company’s stock has shown remarkable growth, up approximately 189% year-to-date.
The announcement coincided with Huang’s visit to Denmark, where he launched a new supercomputer named Gefion. The system features 1,528 graphic processing units (GPUs) and was developed in partnership with the Novo Nordisk Foundation and Denmark’s Export and Investment Fund.
The resolution of the design flaw marks a significant milestone for Nvidia as it continues to lead the AI chip market. The company’s ability to quickly address and resolve the production issues demonstrates its technical capabilities and strong partnership with TSMC.
Currently, Nvidia’s shares trade near record levels, having reached an all-time high of $143.71 per share earlier this week, ahead of major technology companies’ earnings reports.
