NVIDIA's issues with its new Blackwell AI GPUs in August caused an issue with its fab partner TSMC, seeing NVIDIA considering using Samsung to fab its upcoming GeForce RTX 50 series GPUs.

In a new report from The Information picked up by Business Korea, we're learning that "during testing of the Blackwell chips in the weeks after Huang's announcement, NVIDIA engineers discovered that the chips failed in the types of high-voltage environments that are common in data centers" according to "two people with direct knowledge of the problem".
Both sides are blaming each other: TSMC saying that NVIDIA rushing production of Blackwell, and NVIDIA saying the defects are from TSMC's advanced packaging technology. The Information reports that we could see NVIDIA moving its Blackwell-based GeForce RTX 50 series gaming-focused GPUs to Samsung, away from TSMC and enjoy a 20-30% discount moving to the South Korean fab.
- Read more: NVIDIA's next-gen Blackwell Ultra 'B300' AI GPU for GB300 AI servers: socketed design rumored
I don't see NVIDIA moving its next-gen GPU family to Samsung after what we saw happen with previous-generation GeForce RTX series GPUs made at Samsung (GeForce RTX 30 series "Ampere"). However, this news can be used to flex a discount with TSMC if NVIDIA is considering moving its RTX 50 series GPU production to Samsung.
Another thing to consider is that NVIDIA needs all of the most advanced silicon wafer TSMC can make for its Blackwell AI GPU family, so we could have some issues with stock in the coming months leading into the RTX 50 series "Blackwell" gaming GPUs depending on what NVIDIA does next.
NVIDIA is far closer to having its GeForce RTX 50 series GPUs on the market than most people realize, with CES 2025 less than 90 days away. I don't see a sudden shift to Samsung from TSMC for the production of the RTX 50 series, but The Information isn't a small outlet.