Out of nowhere, a small Chinese AI lab called DeepSeek stunned the world last month by releasing a cutting-edge model that outperformed its Western rivals – all while supposedly costing a fraction to train.  

It topped app stores worldwide, fuelling (probably premature) claims that U.S. tech is finished. 

Leading Chinese scientist Rao Yi said it was the “biggest shock to come out of China in 185 years.” Silicon Valley venture capitalist Marc Andreessen dubbed it China’s “Sputnik moment.”

The hyperbole was predictable; but this time it might be justified.  Is China’s tech sector overtaking the West? 

China’s New AI Star 

Well, the first thing to understand: DeepSeek isn’t one of China’s established tech giants. In fact, most Chinese people hadn’t even heard of them until recently. Chinese giants like Baidu and Alibaba seem just as blindsided as their Western rivals. They’ve been churning out large language models since ChatGPT’s debut, but it took a startup to actually move the needle. 

Now, they’re scrambling to react. 

Last week, Alibaba Cloud’s Qwen2.5-Max cracked the global top 10 on Chatbot Arena’s leaderboard, outperforming several DeepSeek models. $BABA shares jumped on the news. They also poached top AI researcher Steven Hoi Chu-hong, managing director of Salesforce Research Asia. Clearly, all this signals that they’re now playing catch-up with DeepSeek. 

But what exactly is DeepSeek doing differently? How did this small firm of 150 employees jump to the front of the pack? 

Export Controls 

U.S. politicians have a simple explanation: blame the chips. 

In a rare show of bipartisan unity, Senators Elizabeth Warren and Josh Hawley introduced legislation targeting “loopholes” in export controls that allegedly let DeepSeek access Nvidia’s silicon. 

Anthropic CEO Dario Amodei echoed this view, telling the ChinaTalk podcast that DeepSeek’s future depends entirely on “how many chips they can get access to.” 

There’s some truth here. The H20 – Nvidia’s export-compliant chip designed specifically to bypass U.S. restrictions on China sales – outperformed its sanctioned H800 cousin in certain scenarios. In fact, a DeepSeek employee even posted on X that “nothing can stop us on the path to AGI, except for computational resources.” The company’s leadership have echoed similar sentiments. Not surprisingly, Trump allies are reportedly preparing plans to slam that door shut. 

But there’s something missing here. 

If it was just about the chips, why couldn’t bigger Chinese tech firms, with deeper pockets and far more resources, do what DeepSeek did first? The story is clearly more complex than “export controls failed.” 

But if it’s not just about the chips, what’s really going on? 

A New Work Culture 

MIT researcher JS Tan points to a different explanation for DeepSeek’s success: its radical approach to talent. 

“While many of China’s tech giants have focused on squeezing maximum output from overworked employees, DeepSeek has demonstrated the transformative potential of a supportive and empowering workplace culture,” writes Tan. 

Breaking from China’s notorious 996 culture, founder Liang Wenfeng apparently built DeepSeek more like a laid-back Stanford lab, ditching the rigid hierarchies that he says suffocate innovation at its bigger rivals. 

“We don’t pre-define roles,” Liang told Chinese outlet An Yong in July last year. “Everyone has their own growth trajectory – there’s no need to push them.” 

Instead, employees are encouraged to explore their creative ideas freely: “If someone has an idea, they can freely access the training cluster without approval. Since we have no hierarchical structure or departmental boundaries, people can mobilize anyone as long as the other party is also interested.” 

DeepSeek also binned the common Chinese tech practice of “horse racing,” where multiple teams compete on the same project. 

Normally we’d file such claims under “startup BS,” but DeepSeek’s engineers regularly pop up on X to confirm this easy-going approach. 

More tellingly, DeepSeek is actually innovating. The most obvious result is that their models achieve more with limited chip resources while outperforming rivals. But they’ve also made genuine interesting user-facing innovations too – like letting users see how the AI reasons through problems step-by-step. 

As Platformer’s Casey Newton put it, the company’s “decision to expose chain of thought is a surprise hit,” suggesting it “might prove to be as influential on other companies as anything else that DeepSeek has done.” 

AI Talent Wars 

DeepSeek’s emphasis on workplace culture comes as China’s AI sector faces a talent shortage. Despite a surge in job openings, a recent report from Maimai, a Chinese professional networking platform, indicates that the supply of qualified candidates is not keeping pace. 

So, in a market where both advanced chips and top-tier AI talent are scarce, DeepSeek’s ability to attract and retain skilled engineers might be even more crucial than access to hardware. 

“When you don’t have resources, all you have is your brain power,” Yu Zhou, a professor at Vassar College, recently told U.S. tech outlet Rest of World. 

What’s Next? 

To solidify its competitive edge, DeepSeek is seeking to expand its team, advertising over 50 open positions with salaries almost twice as high as rival firms on Chinese recruitment platform BOSS Zhipin. 

But whether DeepSeek can maintain its unique culture as it morphs from plucky startup to bloated tech firm will be one of the more interesting beats to follow in China over the next few years. 

For China’s broader AI sector, the challenge is whether those lumbering giants can actually learn a thing or two from DeepSeek’s work culture. 

For the rest of us with teams in China, DeepSeek’s success is a useful reminder that employee empowerment isn’t just some HR buzzword – it can actually drive results.

The post Is DeepSeek’s Workplace Culture the Secret to Its AI Success? appeared first on Kinyu.