OpenAI’s Altman calls DeepSeek-R1 'impressive', promises ‘better models’ as US tech stocks reel

OpenAI CEO Sam Altman on Tuesday lauded Chinese startup DeepSeek’s latest AI model, DeepSeek-R1, as “impressive,” even as the low-cost rival disrupts global markets and sends US tech stocks tumbling.

Altman took to X to comment on the rise of DeepSeek’s models, which are challenging industry leaders like OpenAI.

“DeepSeek's R1 is an impressive model, particularly around what they're able to deliver for the price,” Altman wrote. “We will obviously deliver much better models … but mostly we are excited to continue to execute on our research roadmap.”

The remarks come as DeepSeek’s new AI assistant, powered by its V3 and R1 models, has quickly become the top-rated free app on Apple’s US App Store, overtaking OpenAI’s ChatGPT.

US tech stocks nosedive

DeepSeek’s success is rattling markets, with global investors dumping tech stocks on Monday amid fears of a shift in the AI scene.

Nvidia Corp, whose advanced chips power most state-of-the-art AI models, saw its shares plummet nearly 17%, erasing $593 billion in market capitalisation—the largest single-day loss for any Wall Street company, according to LSEG data.

The Nasdaq Composite dropped 3.1%, with other major players in the AI ecosystem also taking hits. Broadcom Inc fell 17.4%, Microsoft Corp dipped 2.1%, and Alphabet Inc’s shares slid 4.2%.

Trump reacts

US President Donald Trump, who recently talked up a $500 billion joint venture to build AI infrastructure in America, said that DeepSeek was a "wake up call" for US firms.

"Hopefully, the release of DeepSeek AI from a Chinese company should be a wake up call for our industries that we need to be laser-focused on competing to win," Trump told a Republican congressional retreat in Miami.

"I would say that could be a positive," Trump said. "So instead of spending billions and billions, you'll spend less, and you'll come up with hopefully the same solution."

Hangzhou-based DeepSeek claims its R1 and V3 models match or exceed the capabilities of top-tier US offerings at a fraction of the cost. According to a company paper, training DeepSeek-V3 required less than $6 million in computing power, far below the industry norm.

DeepSeek has also touted the cost-efficiency of its models, stating that R1 is 20 to 50 times cheaper than OpenAI’s GPT-4 model, depending on the task.