From inventory dealer to AI titan: How Liang Wenfeng constructed DeepSeek
First got here ChatGPT, then Gemini, adopted by Grok however when DeepSeek got here – it despatched shockwaves throughout the tech business, made traders anxious and wiped off $1 trillion from US tech giants in a day. Touted as ‘AI’s Sputnik second’, the AI assistant shot to the highest rank on Apple’s App Retailer and surpassed even ChatGPT in downloads in US.
“OpenAI just isn’t a god and can’t at all times be on the forefront,” 40-year-old founder Liang instructed Chinese language media outlet 36Kr in July 2024 and simply months later, he did it!
DeepSeek is only a 20-month-old startup nevertheless it has not solely captured the world’s consideration with its groundbreaking AI assistant, nevertheless it’s additionally disrupted the worldwide market with a brand new strategy to AI. Silicon Valley is feeling the warmth already. The worldwide media has dubbed DeepSeek a Chilly Warfare-era “Sputnik” second—foretelling a brand new period in AI.
On the coronary heart of this revolution is DeepSeek’s founder, Liang Wenfeng, whose humble beginnings in China defy the chances. As a self-taught engineering prodigy, Liang has turned a stock-picking AI startup into essentially the most formidable competitor to OpenAI. And DeepSeek’s newest transfer—an economical, open-source AI chatbot that’s already outperforming the competitors—has everybody asking: might this be the start of a brand new tech period?
WHAT IS DEEPSEEK?
DeepSeek’s chatbot is like ChatGPT, however with a twist: it delivers the identical high-level AI expertise at a fraction of the associated fee. Launched in January 2025, it shot to the highest of the U.S. iOS App Retailer, overtaking ChatGPT in downloads and triggering an enormous drop in Nvidia’s inventory worth. And the very best half? It’s open-source, that means anybody can tinker with it, modify it, and construct on it.
HOW DOES DEEPSEEK WORK?
On the core of DeepSeek is an AI-powered search engine that goes past easy key phrase matching. It incorporates machine studying, pure language processing, and semantic search capabilities to grasp the intent behind a question.
What units it other than conventional search instruments is that DeepSeek’s algorithms evolve over time, getting much more correct and environment friendly as you work together with it.
TREMORS IN TECH WORLD
The impression of DeepSeek’s AI fashions, together with R1, is plain. Not solely do they ship cutting-edge efficiency, however additionally they function at a fraction of the price of OpenAI’s fashions—20 to 50 instances cheaper, in truth. This cost-efficiency has traders anxious that cheaper AI might undermine demand for the costly chips driving firms like Nvidia. The outcome? A $592.7 billion loss in Nvidia’s market worth in a single day, sparking a tech sell-off.
THE BRAIN BEHIND DEEPSEEK
The success of the Chinese language-made synthetic intelligence (AI) mannequin introduced its founder Liang Wenfeng into highlight. Whereas DeepSeek’s rise has threatened US tech giants, what additionally drew shock and awe is how an engineering graduate who has by no means studied or labored outdoors of mainland China, pulled off such a feat.
Liang Wenfeng grew up within the Nineteen Eighties in “fifth-tier metropolis” Guangdong, China, as he instructed a 2024 interview. His father was a main college trainer. He obtained his undergraduate and graduate levels from Zhejiang College, certainly one of China’s oldest and best-ranked universities.
Fast rewind: Liang and two former college classmates started buying and selling home shares in 2008, missing abroad or institutional expertise. After experimenting with numerous methods, they adopted a scientific strategy in 2015 and based Excessive-Flyer. Initially utilizing price-and-volume fashions, they included machine studying in 2016, permitting them to uncover new components and determine non-linear relationships. By 2018, machine studying was absolutely built-in into their merchandise.
In April 2023, Excessive-Flyer began a man-made basic intelligence lab devoted to analysis growing AI instruments separate from Excessive-Flyer’s monetary enterprise. In Might 2023, with Excessive-Flyer as one of many traders, the lab grew to become its personal firm, DeepSeek. On November 2, 2023, DeepSeek launched its first sequence of mannequin, DeepSeek-Coder, which is obtainable without spending a dime to each researchers and industrial customers.
After releasing DeepSeek-V2 in Might 2024, DeepSeek grew to become referred to as the catalyst for China’s AI mannequin worth struggle. It was shortly dubbed the “Pinduoduo of AI”, and different main tech giants resembling ByteDance, Tencent, Baidu, and Alibaba started to chop the value of their AI fashions to compete with the corporate. Regardless of the low worth charged by DeepSeek, it was worthwhile in comparison with its rivals that had been dropping cash.
By 27 January 2025, DeepSeek’s AI Assistant, which makes use of the V3 mannequin, surpassed ChatGPT because the highest-rated free app on the iOS App Retailer in the US; its chatbot reportedly solutions questions, solved logic issues and wrote its personal pc applications on parity with something already in the marketplace, in line with benchmark checks utilized by American A.I. firms.
On January 27, DeepSeek restricted its new person registration to Chinese language mainland telephone numbers after reported cyberattack.
In a uncommon interview final yr, he mentioned China’s AI sector “can’t stay a follower ceaselessly”.
He went on: “Typically, we are saying there is a one or two-year hole between Chinese language and American AI, however the actual hole is between originality and imitation. If this does not change, China will at all times be a follower.”
Requested why DeepSeek’s mannequin shocked so many in Silicon Valley, he mentioned: “Their shock stems from seeing a Chinese language firm be a part of their sport as an innovator, not only a follower – which is what most Chinese language companies are accustomed to.”
HOW WAS DEEPSEEK BORN?
The creation of DeepSeek wasn’t a mere accident—it was born out of necessity. In October 2022, US export controls restricted China’s entry to cutting-edge chips like Nvidia’s H100. As an alternative of backing down, Liang and his group leveraged their ingenuity, optimising their mannequin structure with modern methods that made DeepSeek a powerhouse—on a shoestring funds.
The agency had began out with a stockpile of 10,000 H100’s, nevertheless it wanted extra to compete with companies like OpenAI and Meta.
“The issue we face has by no means been funding, however the export management on superior chips,” Liang instructed 36Kr in a second interview in 2024. DeepSeek needed to give you extra environment friendly strategies to coach its fashions.
Liang instructed the Chinese language tech publication 36Kr that the choice was pushed by scientific curiosity slightly than a want to show a revenue.
“I wouldn’t be capable of discover a industrial purpose [for founding DeepSeek] even should you ask me to,” he defined. “As a result of it’s not price it commercially. Primary science analysis has a really low return-on-investment ratio. When OpenAI’s early traders gave it cash, they certain weren’t occupied with how a lot return they might get. Somewhat, it was that they actually wished to do that factor.”
THE MISSION
Liang runs DeepSeek as a bottom-up firm with none preassigned roles or inflexible hierarchy. Talking on the identical, he mentioned, “I imagine innovation is, initially, a matter of perception. Why is Silicon Valley so modern? As a result of they dare to strive. When ChatGPT debuted, China lacked confidence in frontier analysis. From traders to main tech companies, many felt the hole was too large and centered as an alternative on functions. However innovation requires confidence, and younger folks are inclined to have extra of it.”
With DeepSeek, Liang desires to push China forward. He mentioned, “We see that China’s AI can’t stay in a follower place ceaselessly. We regularly say there’s a one- or two-year hole between China and the US, however the actual hole is between originality and imitation. If this doesn’t change, China will at all times be a follower. That’s why some exploration is inevitable.”
