Nvidia calls China's DeepSeek R1 model 'an excellent AI advancement'

Nvidia has praised China's DeepSeek R1 model as "an excellent AI advancement," despite the model's release causing a 17% drop in Nvidia's stock price. DeepSeek's R1 is an open-source reasoning model that reportedly outperforms leading models from U.S. companies, including OpenAI. Nvidia's spokesperson highlighted that DeepSeek's work exemplifies "Test Time Scaling," a technique that allows for the creation of new models using existing resources. The training cost for R1 was reported to be under $6 million, significantly lower than the billions spent by Silicon Valley firms on AI development. Nvidia views this development as an opportunity for increased demand for its graphics processing units (GPUs). The company clarified that DeepSeek utilized Nvidia GPUs compliant with export controls, countering claims that banned models were used. Analysts are now questioning whether the substantial investments by companies like Microsoft and Meta in Nvidia-based AI infrastructure are justified, given the lower costs associated with DeepSeek's model. Nvidia's comments reflect a broader discussion among industry leaders about evolving scaling laws in AI development, particularly the concept of "test-time scaling," which suggests that longer processing times can yield better results.

- Nvidia commended DeepSeek's R1 model despite a significant stock drop.

- DeepSeek's R1 model reportedly outperforms U.S. AI models at a fraction of the cost.

- Nvidia anticipates increased demand for its GPUs due to DeepSeek's advancements.

- Analysts are questioning the value of large investments in AI infrastructure by major tech companies.

- The discussion around AI scaling laws is evolving, focusing on "test-time scaling."