Liang Wenfeng is not your typical AI entrepreneur. Unlike Sam Altman, Demis Hassabis, or Jensen Huang, Wenfeng’s origins lie in finance. He graduated from Zhejiang University and later co-founded High-Flyer, a quantitative hedge fund, in 2015.

At just 40, he has built DeepSeek, an AI research lab that has stunned Silicon Valley, not just by rivaling industry giants like OpenAI and Google but by doing so on a shoestring budget compared to its American counterparts.

His latest creation, DeepSeek-R1, has ignited a firestorm in the AI world. How did a former finance mogul manage to outmaneuver trillion-dollar tech giants and push China into the AI spotlight?

For years, he focused on predicting market trends using AI models to gain an edge in trading. This obsession with AI-driven insights led him to quietly start experimenting with machine learning.

Liang Wenfeng at a meeting hosted by Chinese Premier Li Qiang after his start-up DeepSeek released its AI chatbot © CCTV/YouTube

By 2021, long before AI was a global battleground, Wenfeng was secretly buying thousands of Nvidia GPUs—a move that seemed irrational at the time but now looks like a masterstroke.

"When we first met him, he was this very nerdy guy with a terrible hairstyle talking about building a 10,000-chip cluster to train his own models. We didn’t take him seriously," a business partner told the Financial Times.

Nobody imagined that by 2025, this same "nerdy guy" would create an AI model that could go head-to-head with OpenAI’s GPT-4.

The Birth of DeepSeek: AI Innovation on a Budget

DeepSeek was launched in 2023 as a research lab aimed at pushing AI forward without burning billions of dollars.

Instead of poaching talent from Google and Meta, Wenfeng focused on hiring brilliant young minds from Chinese universities, offering salaries on par with top tech companies.

DeepSeek’s first AI model, DeepSeek Coder, launched in November 2023, an open-source AI specialized in coding.

Then came DeepSeek LLM (67B parameters) and DeepSeek-V2 (May 2024), both of which delivered outstanding performance at a fraction of the usual cost.

But it was DeepSeek-R1, released in January 2025, that truly made Silicon Valley sit up.

DeepSeek-R1 vs. GPT-4: How a $5.6M Model Shocked the World

Most major AI models are trained with budgets in the hundreds of millions—sometimes even billions—to handle the massive computing power required.

DeepSeek-R1 was built with just $5.6 million.

To put that into perspective:

• OpenAI’s GPT-4 reportedly cost over $100M in compute alone.
• Google and Meta are investing billions into their AI research.
• Wenfeng did it with a team of just 200 people.

And the kicker? DeepSeek-R1 actually competes with GPT-4 on reasoning tasks.

"What’s most shocking isn’t just that DeepSeek’s AI is good—it’s that they did it with fewer resources," said an industry analyst.

This lean approach is redefining AI development. DeepSeek’s focus on reinforcement learning and model distillation (compressing large models into smaller, efficient ones) is proving that you don’t need infinite cash to build world-class AI.

How DeepSeek’s Success Is Changing the AI Game

One of the most radical aspects of DeepSeek’s strategy is its commitment to open-source AI.

Unlike OpenAI and Google, which closely guard their models, DeepSeek allows developers to build on top of its technology without paying hefty licensing fees.

Why This Matters:

• Lower AI costs: Smaller companies can now afford cutting-edge AI.
• Faster AI adoption: More businesses can integrate AI without starting from scratch.
• Increased transparency: Open-source AI builds trust in a world worried about black-box algorithms.

DeepSeek’s API pricing is also a game changer:

• $0.14 per million input tokens
• $0.28 per million output tokens

For comparison, OpenAI’s GPT-4 API can cost up to 5x more.

The result? AI is becoming more accessible than ever.

How Wenfeng Outsmarted Silicon Valley

Wenfeng’s financial background has given him an edge few tech founders possess:

• He understands capital efficiency better than anyone in AI.
• He’s not afraid to think outside the traditional AI playbook.
• He plays the long game, focusing on strategic investments over hype cycles.

By betting on efficient AI rather than brute-force computing, Wenfeng has forced the world’s biggest tech companies to rethink their approach.

What’s Next for DeepSeek?

DeepSeek is still young, but Wenfeng’s vision is clear—make AI accessible, affordable, and powerful.

With DeepSeek-R1 already shaking up the AI industry, insiders believe the company is working on an even larger model—one that could take on OpenAI’s rumored GPT-5.

Whether DeepSeek can maintain its momentum remains to be seen, but one thing is certain: Liang Wenfeng has changed the AI landscape forever.