NVIDIA (NVDA) Resolves Design Flaw in Blackwell AI Chips with TSMC's Help

Author's Avatar
2 days ago
Article's Main Image

NVIDIA's (NVDA, Financial) CEO Jensen Huang announced that the design flaw in the company's latest Blackwell AI chips has been rectified with the assistance of TSMC (TSM). This flaw had previously hindered production, affecting major clients such as Meta (META), Alphabet (GOOGL), and Microsoft (MSFT). Huang attributed the issue to NVIDIA's design, stating it was entirely their responsibility.

The Blackwell chips, unveiled in March, were initially slated for a second-quarter release but experienced delays. Despite this setback, Huang reported an overwhelming demand for the new series. Recent developments include Dell's announcement of devices featuring the Blackwell AI accelerator and Google's deployment of servers powered by these chips. OpenAI and Microsoft have also publicized collaborative efforts involving Blackwell.

Rumors of strained relations between NVIDIA and TSMC due to production delays were dismissed by Huang as "fake news." He praised TSMC for helping restore Blackwell production, emphasizing the intricate design and simultaneous production of seven different chip types.

The Blackwell chips are a significant advancement, combining dual silicon wafers to form a single component, enhancing task execution speed by up to 30 times. Huang confirmed that these chips are set to ship in the fourth quarter.

Recently, Huang introduced the Gefion supercomputer in Denmark, featuring 1,528 GPUs, a project developed in collaboration with the Novo Nordisk Foundation, Denmark's Export and Investment Fund, and NVIDIA.

Disclosures

I/We may personally own shares in some of the companies mentioned above. However, those positions are not material to either the company or to my/our portfolios.