The Chinese startup DeepSeek released its flagship AI model R1 on January 20, surprising Silicon Valley with the model’s advanced capabilities. R1 matched or surpassed the functionality of AI released by OpenAI, Google, and Meta — on a much smaller budget and without the latest AI chips.
Over the past week, the DeepSeek app has proven popular with the public. It surged past ChatGPT in popularity, reaching No. 1 on the U.S. Apple app store and within the top free Android apps on the Google Play Store at the time of publication.
DeepSeek’s R1 release has prompted questions about whether the billions of dollars of AI spending in the past few years was worth it — and challenged the notion that the U.S. is the world’s leader in AI, per BBC.
DeepSeek’s AI arrives as the U.S. looks to ramp up spending on AI. Last week, President Donald Trump announced a joint project with OpenAI, Oracle, and Softbank called Stargate that commits up to $500 billion over the next four years to data centers and other AI infrastructure.
What is DeepSeek?
DeepSeek is a Chinese AI startup that creates open AI models—so any developer can access and build on the technology.
DeepSeek is different from ChatGPT because it states its chain-of-thought reasoning before giving a response to a prompt. Apple App Store and Google Play Store reviews praised that level of transparency, per Bloomberg.
It is free to download and use, though it does require users to sign up before they can access the AI.
DeepSeek AI. Photographer: Andrey Rudakov/Bloomberg via Getty Images
How much did DeepSeek cost to develop?
In a paper released last month, DeepSeek researchers stated that they built and trained the AI model for under $6 million in only two months.
Related: I Co-Founded an App That Uses AI — Here’s Why I’m Worried About My Child’s Future
They stated that they used around 2,000 Nvidia H800 chips, which Nvidia tailored exclusively for China with lower data transfer rates, or slowed-down speeds when compared to the H100 chips used by U.S. companies. In October, the U.S. Department of Commerce banned the sale of the H800 chip to China with the goal of preventing access to chips that could fuel AI breakthroughs, especially for military purposes.
In contrast, Dario Amodei, the CEO of U.S AI startup Anthropic, said in July that it takes $100 million to train AI — and there are models today that cost closer to $1 billion to train.
So how did DeepSeek pull ahead of the competition with fewer resources? Meta’s chief AI scientist Yann LeCun stated in a Threads post on Saturday that DeepSeek had “profited from open research and open source.”
Related: Meta’s AI Chief Is in the Middle of a Public Clash With Elon Musk: ‘Secrecy Hampers Progress’
“They came up with new ideas and built them on top of other people’s work,” LeCun stated. “Because their work is published and open source, everyone can profit from it.”
Why is DeepSeek so popular?
DeepSeek is free and offers top-of-the-line performance. Last week, the scientific journal Nature published an article titled, “China’s cheap, open AI model DeepSeek thrills scientists.” The article showed that R1’s performances on certain chemistry, math, and coding tasks were on par with one of OpenAI’s most advanced AI models, the o1 model OpenAI released in September.
In addition to high performance, R1 is open-weight, so researchers can study, reuse, and build on it. It isn’t considered fully open source because DeepSeek hasn’t made its training data public.
The AI model has also received stellar reviews. Investor Marc Andreessen called it “one of the most amazing and impressive breakthroughs” he had “ever seen” in a Friday post on X while Microsoft CEO Satya Nadella called it “super impressive” at last week’s World Economic Forum in Switzerland.
Deepseek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen — and as open source, a profound gift to the world. ??
— Marc Andreessen ?? (@pmarca) January 24, 2025
What are DeepSeek’s effects on U.S. tech stocks?
Nvidia shares fell by 13% after the opening bell on Monday, wiping $465 billion from the AI chipmaker’s market cap. It was the biggest drop in value in U.S. stock market history, per Bloomberg.
According to NBC News, Microsoft and Alphabet fell 4% each while Meta fell nearly 2%.
NBC News also notes that the Nasdaq Composite fell by 3.4% at the opening bell, while the Dow fell 180 points and the S&P 500 fell nearly 2%.
Related: Why One Prominent Investor, ‘Britain’s Warren Buffett,’ Is Staying Away From Nvidia Stock
