The U.S. export curbs have forced Nvidia to write off $5.5 billion in inventory and walk away from $15 billion in sales. The restrictions have introduced new limits on GPU memory bandwidth, capping it at 1.7-1.8 terabytes per second, compared to the H20's 4 terabytes per second. This bandwidth is crucial for AI workloads requiring extensive data processing. Nvidia's CEO, Jensen Huang, has expressed concerns that continued U.S. export curbs could lead more Chinese customers to switch to Huawei's Ascend 910B chip.
Key takeaways:
- Nvidia plans to launch a new AI chipset for China at a lower price than the restricted H20 model, with mass production starting as early as June.
- The new GPU will be based on Nvidia's RTX Pro 6000D and will use conventional GDDR7 memory, avoiding advanced packaging technology.
- Nvidia's market share in China has dropped from 95% to 50% due to U.S. export restrictions, with Huawei emerging as a main competitor.
- The new export restrictions limit GPU memory bandwidth, impacting AI workloads, with the new Nvidia GPU expected to achieve 1.7 terabytes per second.