Chinese firm DeepSeek encountered technical difficulties in developing its new AI model R2.
Technical Issues with Huawei Chips
DeepSeek failed to successfully train the R2 model using Huawei’s Ascend chips, which led to the delay in its launch. This situation arose from recommendations by the Chinese government, which urged the company to utilize Huawei chips to reduce reliance on US technology.
Chip Switching and Future Plans
Due to ongoing challenges, DeepSeek switched to using Nvidia chips for the training process, while employing Huawei chips for inference tasks. The anticipated release of R2, originally set for May, is now uncertain. Additionally, the resource allocation for data labeling took longer than expected.
Prospects for R2 Release and Market Reactions
Despite the delay, Chinese media suggests that DeepSeek may release the R2 model in the upcoming weeks to catch up with competitors that have already launched new AI models. Experts point out the inefficiency of Huawei chips compared to Nvidia’s products, which may affect DeepSeek’s competitiveness.
It is expected that DeepSeek will continue working to enhance its R2 AI model, despite current limitations, and may be able to successfully integrate Chinese technology in the future.
