Key Points
- Chinese AI startup DeepSeek has delayed the launch of its new R2 AI model.
- The delay was caused by technical problems encountered during the training of the model on Huawei’s Ascend chips.
- DeepSeek has now switched back to using Nvidia chips for the training process.
- The setback highlights the difficulty Chinese companies face in replacing U.S. tech.
Chinese AI startup DeepSeek has been forced to delay the release of its highly anticipated new AI model after running into persistent technical problems trying to train it on Huawei’s Ascend chips. The setback highlights the major challenges Chinese companies are facing as they try to reduce their reliance on U.S. technology.
According to a report from the Financial Times, DeepSeek’s attempts to train its new R2 model using Huawei’s chips repeatedly fell through. The company has now been forced to switch back to using powerful Nvidia chips for the training process. The problems with the Huawei chips are reportedly the main reason why the R2 model’s launch was delayed from its original May date.
This is a significant blow to Beijing’s push for tech self-sufficiency. The Chinese government has been encouraging its top AI companies to use domestic chips like Huawei’s Ascend, especially as U.S. restrictions on chip exports have become a major source of tension.
While the U.S. recently allowed Nvidia to resume selling its less-advanced H20 chips to China, Beijing has raised its own security concerns about using American-made hardware.
DeepSeek’s previous model, the R1, which made waves in the AI world earlier this year, was developed largely using Nvidia’s H20 chips. It seems that for now, even China’s top AI companies are still struggling to break free from their dependence on Nvidia’s superior technology for the most demanding AI tasks.
