Nvidia applauds DeepSeek model – despite record-breaking share drop

Nvidia praises DeepSeek’s R1 model as “an excellent AI advancement,” despite the Chinese company’s success sparking a record-breaking share drop for the chipmaker.

In a statement to CNBC, Nvidia tipped its hat to DeepSeek, just hours after the company’s stock plunged by 17% following the impact of DeepSeek’s own AI model. The open-source large language model outperformed those in the US, many of which use Nvidia chips.

“DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling,” an Nvidia spokesperson told CNBC on Monday, January 28. “DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant.”

DeepSeek′s self-reported training cost was less than $6 million, a fraction of the multi-billions that major tech companies like OpenAI (the company behind ChatGPT) have sunk into their models. However, Nvidia appears to not see DeepSeek as a threat to the industry but rather creating more opportunities for the America-based chipmaker – although that could also be an attempt to recoup some faith in the tech company after the stocks plummeted.

Notably, Nvidia touched upon the GPUs used by DeepSeek as being fully compliant in terms of exports, going against earlier comments from some believing that DeepSeek used Nvidia GPUs that were banned in mainland China.

“Inference requires significant numbers of NVIDIA GPUs and high-performance networking,” the spokesperson continued. “We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.”

While some may be wondering whether the huge investments into companies like Microsoft, Google, and Apple are going to waste, with DeepSeek proving strong results are possible on a budget, Nvidia’s statement poises them as looking to a future with wider – cheaper – access to AI and large language models.

Featured image: Midjourney

The post Nvidia applauds DeepSeek model – despite record-breaking share drop appeared first on ReadWrite.