What is DeepSeek? The low-cost Chinese AI firm that has turned the tech world upside down
The arrival of a previously little-known Chinese tech company has attracted global attention as it sent shockwaves through Wall Street with a new AI chatbot.
On Monday, DeepSeek, a tiny company which reportedly employs no more than 200 people, caused American chipmaker Nvidia to have almost $600bn wiped off its market value - the biggest drop in US stock market history.
The Chinese AI chatbot threatens the billions of dollars invested in AI while causing US tech stocks to lose well over $1trn (£802bn) in value, according to market analysts.
DeepSeek, which has developed two models, V3 and R1, is now the most popular free application on Apple's App Store across the US and UK.
What's DeepSeek?
The startup was founded in 2023 in Hangzhou, China, by Liang Wenfeng, who previously co-founded one of China's top hedge funds, High-Flyer.
The fund had by 2022 amassed a cluster of 10,000 of California-based Nvidia's high-performance A100 graphics processor chips that are used to build and run AI systems, according to a post that summer on Chinese social media platform WeChat.
Why all the attention now?
The hype - and market turmoil - over DeepSeek follows a research paper published last week about the R1 model, which showed advanced "reasoning" skills.
They include the ability to rethink its approach to a math problem while, depending on the task, being 20 to 50 times cheaper to use than OpenAI's o1 model, according to a post on DeepSeek's official WeChat account.
That's why there are fears it could undermine the potentially $500bn AI investment by OpenAI, Oracle and SoftBank that Mr Trump has touted.
What's so cool about it?
As Morgan Brown, vice president of product and growth in artificial intelligence at Dropbox, put it, it is currently "insanely expensive" to train top AI models.
"DeepSeek just showed up and said 'LOL what if we did this for $5M instead?'
"And they didn't just talk - they actually DID it. Their models match or beat GPT-4 and Claude on many tasks. The AI world is (as my teenagers say) shook," he said in a post on X.
So how much did it cost?
The company wrote in a paper last month that the training of DeepSeek-V3 required less than $6m (£5m) worth of computing power from Nvidia H800 chips.
How has President Trump responded?
Speaking to House Republicans on Monday, the 78-year-old Republican called the development a "wakeup call for our industries that we need to be laser-focused on competing to win".
But Mr Trump signed an order on his first day in office last week that said his administration would "identify and eliminate loopholes in existing export controls", signalling that he is likely to strengthen Mr Biden's approach.
Venture capitalist Marc Andreessen, who has advised Mr Trump on tech policy, has warned overregulation of the AI industry by the US government will hinder American companies and enable China to get ahead.
"Deepseek R1 is AI's Sputnik moment," said Mr Andreessen in a recent post on X, referencing the 1957 satellite launch that set off a Cold War space exploration race between the Soviet Union and the US.
What are the concerns?
Some people expressed their reservations about the Chinese company and its handling of users' data.
Bill Ackman described DeepSeek as "a Trojan Horse" and said, TikTok, which was temporarily banned in the US earlier this month over national security concerns, "is just a toy by comparison".
Sky News has reached out to DeepSeek for comment.
UK Prime Minister Sir Keir Starmer's spokesman said on Tuesday he would not "get ahead of specific models" when asked whether he would rule out using Chinese AI in Whitehall.
But Dr Lukasz Olejnik, independent researcher and consultant, affiliated with King's College London Institute for AI, claims the way the model is designed provides for "perfect data privacy".
He told Sky News: "What truly sets DeepSeek apart is its accessibility thanks to open-weight models.
"Unlike centralised models, its open-source versions can be run locally, providing perfect data privacy.
"Organisations are already deploying full models internally, ensuring complete control over sensitive information.
"This is much better than AI tools served on remote servers."
Too controversial?
People have also been flagging how, when it comes to questions about alleged wrongdoing and human rights abuses at the hands of the Chinese government, the app seems unable to respond.
Seemingly controversial topics include Taiwan, which China claims as its own, the 1989 Tiananmen Square protests and massacre as well as China's treatment of its Uyghur population, which the UN has previously said "may constitute crimes against humanity".
-SKY NEWS