DeepSeek was founded in 2023 in Hangzhou, China, and released its first AI large language model later that year, so why has the startup roiled markets from India to the US and set in motion panic, bewilderment, wonder -- all at once -- in Silicon Valley?
Is America's Big Tech about to lose its cutting edge?
Although China's DeepSeek was founded in 2023, two recent developments have powered it to the top of conversation across the tech world.
DeepSeek CEO Liang Wenfeng is no stranger to the AI world, having co-founded one of China's top hedge funds with an eye on AI-driven quantitative trading.
On Monday DeepSeek's AI assistant was the number 1 downloaded free app on Apple Store.
Before that, on January 20 the company released its latest AI model R1, an Open Source one, and a research paper on it claimed that it demonstrated advance reasoning skills -- and came at a much lower cost than OpenAI's o1.
What it meant was that R1 could not only rethink how to tackle a math problem, but it also did so at a much lower cost than Big Tech have been touting all along.
One estimate put the development spend on DeepSeek at $6 million compared to the $1 trillion that OpenAI, Google are expected to spend over the years.
In plain English DeepSeek is an open-source large language model that does 'inference-time computing' meaning each query is addressed by only the most relevant portions of their model, which saves money and computation power.
In short, DeepSeek uses up less energy in computing, which is proving to be more cost-effective than the US models available.
Cheaper ripoffs was how China powered its way as an economic superpower but hitherto the tech realm had remained out of bounds.
With DeepSeek's AI model the West, it seemed, was about to lose its competitive edge in cutting edge generative AI technology. And this is what is causing deep furrows in the forehead of tech titans in Silicon Valley.
Last month DeepSeek demonstrated its chops when its new AI model claimed to be as good as, if not better than ChatGPT's OpenAI, hitherto the benchmark. And its claims slowly started getting traction, as it debuted on Apple and Google app stores.
One of its biggest endorsers was Marc Andreessen who, over the weekend, called DeepSeek 'AI's Sputnik moment' -- a reference to the 1957 satellite launch that set off the space war between the US and then USSR (now Russia).
What Andreessen's remark underscores is that if the US didn't pull up its socks the battle over AI could well be lost as it seemed 70 years ago before the US pulled back with its Apollo missions that put man on the moon.
What the US needs to do, goes an opinion, is revisit its regulation of the AI industry to make it more competitive.
DeepSeek's initial success -- that is what it seems like at the moment -- also calls into question US foreign policy's key strategy to control and curtail the sale US-designed AI semiconductors to China.
On Monday, Nvidia, the advanced chip manufacturer, shed $600 billion in value, raising questions whether investors are relooking at tech stocks.
Will the US roll back its punitive measure? In an important move, President Donald J Trump on his first day in office vowed that his administration will eliminate loopholes in the present export controls -- meaning a tougher approach than his predecessor.
Ominously, over Monday-Tuesday DeepSeek reported that it was hit by a cyber attack and users could not register on its site. What is clear is that there's a battle for supremacy on hand.
© 2025