Nvidia resolves flaw with Blackwell AI chips

Blackwell AI chips from Nvidia, initially delayed by a design flaw, will now ship in the fourth quarter.

Nvidia and Vietnam join forces to create AI research and development hubs.

Nvidia’s CEO Jensen Huang announced that a design flaw impacting the company’s Blackwell AI chips has been resolved with assistance from TSMC, its long-term Taiwanese manufacturing partner. The production glitch had delayed chip shipments, initially set for the second quarter, affecting clients such as Google, Microsoft, and Meta.

Huang acknowledged Nvidia was solely responsible for the flaw, which had reduced production yields. He dismissed reports of tensions with TSMC, crediting the manufacturer for helping restore manufacturing efficiency. The chips, which involve the integration of seven different components, are now expected to ship in the fourth quarter.

Blackwell chips, Nvidia’s latest innovation, feature two silicon squares fused into a single unit, delivering speeds 30 times faster than previous models. They are designed for advanced tasks, including AI-driven responses from chatbots. Shares in Nvidia fell by 2% in early trading following news of the delay.

Huang made the announcement during a visit to Denmark, where he introduced Gefion, a new supercomputer featuring 1,528 GPUs. Built in partnership with the Novo Nordisk Foundation and Denmark’s Export and Investment Fund, Gefion is expected to enhance high-performance computing in the region.