Nvidia CEO Jensen Huang announced on Wednesday that a design flaw affecting the company's latest Blackwell AI chips has been resolved with the assistance of its long-term manufacturing partner, Taiwan Semiconductor Manufacturing Company (TSMC). The issue, which had impacted production and delayed the chip's release, is now fixed, Huang confirmed during an event in Denmark.

Nvidia unveiled its Blackwell chips in March, originally planning to ship them in the second quarter of the year. However, production delays pushed the release to the fourth quarter, potentially affecting major customers such as Meta Platforms, Google, and Microsoft. Huang explained that the chips had a functional design flaw that led to low manufacturing yields, admitting that "it was 100% Nvidia's fault."

Despite reports suggesting tension between Nvidia and TSMC over the production delays, Huang dismissed such claims as "fake news," praising TSMC's role in addressing the issue. "What TSMC did was to help us recover from that yield difficulty and resume the manufacturing of Blackwell at an incredible pace," he said.

Nvidia's Blackwell chips are designed for high-performance AI tasks, featuring two squares of silicon that are fused into a single component. According to the company, this new architecture makes the chips 30 times faster than previous models at tasks such as chatbot responses. The chips are now expected to ship in the fourth quarter of 2024, following the delay caused by the design flaw.

Huang made the announcement in Denmark, where he was launching a new supercomputer named Gefion. Developed in collaboration with the Novo Nordisk Foundation, Denmark's Export and Investment Fund, and Nvidia, Gefion is equipped with 1,528 graphic processing units (GPUs) and aims to support Denmark's growing ambitions in AI and computational research.