What is DeepSeek? The Budget-Friendly Chinese AI Company Disrupting the Tech Industry | Science, Climate & Tech News
A previously obscure Chinese tech firm has garnered international attention after its new AI chatbot sent shockwaves through Wall Street.
On Monday, DeepSeek, a small company with no more than 200 employees, led to American chipmaker Nvidia witnessing a loss of almost $600bn in market value—marking the most significant drop in US stock market history.
The emergence of this Chinese AI chatbot poses a threat to the billions invested in AI, with US tech stocks plummeting by over $1 trillion (£802bn) in value, according to market experts.
DeepSeek has developed two models, V3 and R1, which have now become the top free applications on Apple’s App Store in both the US and UK.
What is DeepSeek?
This startup was established in 2023 in Hangzhou, China, by Liang Wenfeng, who had previously co-founded one of China’s leading hedge funds, High-Flyer.
By 2022, the fund had accumulated a stockpile of 10,000 high-performance A100 graphics processor chips from California-based Nvidia, used for building and running AI systems, as noted in a post on the Chinese social media platform WeChat that summer.
Former US President Joe Biden’s administration restricted the sales of those chips to China shortly thereafter, a policy likely to be continued by his successor, Donald Trump, who was recently sworn in for a second term in the White House.
Please use Chrome browser for a more accessible video player
Why the sudden interest?
The excitement—and market volatility—surrounding DeepSeek stems from a research paper released last week detailing the R1 model’s impressive “reasoning” capabilities.
These include its ability to reassess its approach to mathematical problems while being 20 to 50 times more cost-effective than OpenAI’s model, as stated in a post on DeepSeek’s official WeChat account.
Concerns arise that it may undercut the approximately $500bn AI investments by OpenAI, Oracle, and SoftBank endorsed by Mr. Trump.
What makes it remarkable?
As Morgan Brown, Dropbox’s vice president of product and growth in artificial intelligence, remarked, current top AI model training is “insanely expensive.”
“DeepSeek showed up and said, ‘LOL, what if we did this for $5M instead?'”
“And they didn’t just talk—they actually accomplished it. Their models either match or exceed GPT-4 and Claude in numerous tasks. The AI community is, as my teenagers say, shook,” he mentioned in a post on X.
What was the cost?
The company revealed in a recent paper that training DeepSeek-V3 required less than $6m (£5m) worth of computing resources using Nvidia H800 chips.
How did President Trump react?
Addressing House Republicans on Monday, the 78-year-old GOP leader described the development as a “wakeup call for our industries to maintain competitiveness.”
Additionally, Mr. Trump enacted an order on his first official day last week to “identify and close loopholes in existing export controls,” indicating an intent to bolster Mr. Biden’s measures.
Read more from Sky News:
What are the ‘UFO clouds’ over UK skies?
What are the UK’s knife crime laws?
Venture capitalist Marc Andreessen, an advisor to Mr. Trump on technology policy, cautioned that overregulating the AI sector by the US government could stifle American firms and give China an advantage.
“DeepSeek R1 is AI’s Sputnik moment,” remarked Mr. Andreessen in a recent post on X, referencing the 1957 satellite launch that triggered a Cold War rivalry in space exploration between the US and the Soviet Union.
What are the worries?
Several individuals have raised concerns regarding the Chinese company and its user data handling.
Bill Ackman described DeepSeek as “a Trojan Horse,” indicating that TikTok, recently banned in the US for national security reasons, “is merely a toy by comparison.”
Sky News has contacted DeepSeek for feedback.
When questioned about the use of Chinese AI in Whitehall, UK Prime Minister Sir Keir Starmer’s spokesperson said on Tuesday that he would not “preempt specific models.”
However, Dr. Lukasz Olejnik, an independent researcher and consultant affiliated with the King’s College London Institute for AI, asserts that the app design ensures “perfect data privacy.”
He stated to Sky News: “What truly differentiates DeepSeek is its accessibility through open-weight models.
“In contrast to centralized models, its open-source versions can be run locally, providing absolute data privacy.
“Organizations are already deploying complete models in-house, ensuring total control over sensitive information.
“This approach is far superior to AI tools hosted on remote servers.”
Is it too controversial?
Concerns have also been raised about the app’s inability to respond when faced with questions regarding alleged misconduct and human rights violations by the Chinese government.
Controversial subjects include Taiwan, which China asserts as its territory, the 1989 Tiananmen Square protests and massacre, as well as China’s treatment of its Uyghur population, previously characterized by the UN as potentially constituting “crimes against humanity.”
