A little-known AI lab out of China has ignited fresh panic throughout Silicon Valley after releasing new AI models that appear to be able to outperform the best ones in the U.S. despite being built more cheaply and with less powerful chips.

This is how CNBC introduced DeepSeek, an AI startup that almost every tech and AI enthusiast must have heard about in recent days.

While media reports provide less clarity on DeepSeek, the newly released model, DeepSeek-R1, appeared to rival OpenAI’s o1 on several performance benchmarks.

This raised certain concerns and widespread talks in tech circles, but not as much as for the model itself but for the fact it was built despite U.S. curbs on technology and advanced chips to China and much cheaper than most of leading Western models.

“DeepSeek, as the lab is called, unveiled a free, open-source large-language model in late December that it says took only two months and less than $6 million to build, using reduced-capability chips from Nvidia called H800s,” the report from CNBC said.

“The new developments have raised alarms on whether America’s global lead in artificial intelligence is shrinking and called into question big tech’s massive spend on building AI models and data centers,” it added.

This and similar reports followed widespread debate on social media platform X and it came only days after new U.S. President Donald Trump touted the “Stargate Project,” led by OpenAI, Oracle and Softbank, to invest up to half a trillion dollars in AI infrastructure and data centers.

Chinese models

DeepSeek drew widespread attention in global AI circles last month after tests showed its V3 large language model outperformed those of OpenAI and Meta despite a smaller development budget and plans to charge users a lot less, Reuters reported earlier this week.

It also cited that the developments in AI reasoning by ByteDance, the owner of TikTok, DeepSeek and others, are likely to challenge the market share of OpenAI and other large language models in terms of both performance metrics and fees charged to users.

Other Chinese firms that have unveiled their own reasoning models in the past weeks include Moonshot AI, Minimax and iFlyTek, it also said.

“To see the DeepSeek new model, it’s super impressive in terms of both how they have really effectively done an open-source model that does this inference-time compute and is super-compute efficient,” Microsoft CEO Satya Nadella said at the World Economic Forum (WEF) in Davos, Switzerland, on Wednesday. “We should take the developments out of China very seriously.”

The startup itself says on its website: “DeepSeek-R1 is now live and open source, rivaling OpenAI’s Model o1.”

OpenAI triggered the race in AI development after it launched ChatGPT in November 2022 and its “Strawberry” series of AI reasoning models in September last year. The latter are capable of reasoning through complex tasks and solving more challenging problems than previous models in science, coding and math.

Last week, OpenAI CEO Sam Altman said they had finalized a version of its new reasoning AI model, o3 mini, and would launch it in a couple of weeks.

The company also unveiled on Thursday an artificial intelligence program called “Operator” that can tend to online tasks such as ordering items or filling out forms.

Yet, some critics also pointed out that apparent success and concerns stemming from the rising popularity of DeepSeek come from the fact it has an open-sourced model.

“Unlike many Chinese AI firms that rely heavily on access to advanced hardware, DeepSeek has focused on maximizing software-driven resource optimization,” Marina Zhang, an associate professor at the University of Technology Sydney, who studies Chinese innovations, told Wired.

“DeepSeek has embraced open source methods, pooling collective expertise and fostering collaborative innovation. This approach not only mitigates resource constraints but also accelerates the development of cutting-edge technologies, setting DeepSeek apart from more insular competitors,” she said.