DeepSeek: The AI startup that's saying no to money

DeepSeek's billionaire founder has a peculiar problem: His AI company is making too much money. The Hangzhou-based startup hit profitability last month. Then it did something unusual. It stopped taking new customers.

While Silicon Valley's AI companies sprint to grab market share, DeepSeek's founder Liang Wenfeng shrugs off eager investors. He's turned away China's tech giants. He's dodged venture capitalists. He's even ghosted government-backed funds.

The quest for artificial minds

His reason? He's too busy chasing the holy grail of artificial intelligence.

DeepSeek burst onto the scene in January with its R1 model. The AI system matched its U.S. rivals' performance at a fraction of the cost. Customers from healthcare and finance rushed to buy access. The demand overwhelmed DeepSeek's servers.

Most startups would celebrate this problem. They'd raise more money. They'd expand. Not DeepSeek.

Sorry, we're not interested in your billions

Liang wants to build artificial general intelligence (AGI) – machines that think like humans. He's channeling most resources into research, not revenue.

This stance baffles investors. One Chinese tech fund pulled high-level government connections just to get a meeting. They ended up talking to a finance employee who delivered a simple message: "Sorry, we're not raising."

DeepSeek runs lean with just 160 employees. For comparison, OpenAI employs over 2,000 people. While OpenAI burns through $5 billion yearly and hunts for $40 billion more, DeepSeek operates on its founder's personal fortune.

Building the war chest

Liang's war chest includes 20,000 high-end AI chips from Nvidia – bought before export restrictions kicked in. He's also secured access to state-funded data centers, easing computing constraints.

The company's low-key approach has created unexpected winners. Tech giant Tencent now offers DeepSeek's open-source models to its cloud clients. About half have tried them. Twenty percent want customized versions.

Some major players have noticed DeepSeek's absence in the commercial space. Apple chose Alibaba's Qwen over DeepSeek for AI functions on Chinese iPhones. The decision highlights the risks of prioritizing research over market presence.

A tale of two strategies

But Liang appears unfazed. While competitors chase corporate contracts, his team focuses on pushing technical boundaries. The contrast with Silicon Valley couldn't be starker.

Take OpenAI's approach. The San Francisco company leveraged its early lead to build a massive consumer business around ChatGPT. It charges both corporate customers and individual users. The strategy worked – OpenAI generated about $4 billion in revenue last year.

DeepSeek chose a different path. It didn't even have a consumer product until R1's launch. The company's commitment to research over revenue puzzles industry observers used to the "grow fast" mindset.

Beijing tightens its grip

Success brings scrutiny. China now watches DeepSeek with eagle eyes. The government sees a potential tech champion – and wants to keep it that way.

New restrictions hit the company after January's R1 launch. Employees can't travel freely abroad anymore. Their parent company, High-Flyer, holds their passports. The Chinese government now screens potential investors.

The clampdown fits a pattern. Beijing recently ordered AI researchers to avoid U.S. travel, fearing leaked trade secrets. For DeepSeek, the price of excellence means dancing to Beijing's tune.

Storm clouds and silver linings

Challenges loom on the horizon. Future chip access remains uncertain under U.S. export controls. Some industry insiders suggest DeepSeek may eventually need state-backed funding to maintain political support in China's AI landscape.

Meanwhile, the company's engineers race to release upgraded R2 and V4 models. The launch might come earlier than the planned May date to maintain momentum.

Venture capitalist Yusen Dai sees method in this apparent madness. He compares AI development to education: "A high school student can't make much money, while if you train him to become PhD, he can make a lot more."

The monastic approach to AI

This long-term view shapes everything at DeepSeek. While competitors optimize for quarterly results, Liang pursues his AGI dreams with monastic focus.

One investor compared DeepSeek to a Navy SEAL team: small, elite, and laser-focused on its mission. The company's motto might as well be: "We don't do IPOs, we do AGI."

DeepSeek's emergence has rattled markets. It raises doubts about whether U.S. tech giants can maintain their technical edge. It questions the wisdom of massive AI infrastructure spending by Big Tech groups.

The art of swimming upstream

The startup's unconventional strategy offers a masterclass in swimming against the current. In an industry obsessed with hypergrowth, DeepSeek proves that sometimes less is more.

Why this matters:

• Silicon Valley's "grow at all costs" playbook isn't the only path to AI success. DeepSeek proves you can compete globally while staying small and focused.
• Money can't buy everything in the AI race. Sometimes, having a billionaire founder who doesn't care about making more billions is the ultimate competitive advantage.