The AI landscape is shifting rapidly, and China is leading the charge. In January 2025, Chinese AI firm DeepSeek introduced DeepSeek-R1, an open-source inference model that rivaled OpenAI’s top-tier performance at just 1/30th of the API cost. The sheer efficiency of the model, combined with its open accessibility, sent ripples through the global AI community. Suddenly, a question loomed large: What if AI’s future isn’t shaped in Silicon Valley after all?
DeepSeek’s success defies conventional expectations. With just $6 million in funding, it has developed a model that competes with the billion-dollar efforts of Meta, Google, and Microsoft. As a result, global users—particularly small businesses and individual developers—are flocking to DeepSeek-R1, making it their foundational AI model. The company’s emergence marks a new era where China isn’t just playing catch-up but setting the pace for AI innovation.This shift wasn’t accidental. DeepSeek’s founder, Liang Wenfeng, has long envisioned a future where China moves beyond application development and asserts itself as a leader in fundamental AI research. In a recent interview, he discussed how DeepSeek managed to disrupt an industry dominated by tech giants.
“We never intended to be a disruptor; it just happened by accident.”
One of the core reasons behind DeepSeek’s success is its focus on foundational research rather than rapid commercialization. While many Chinese AI firms have based their models on Meta’s Llama structure, DeepSeek chose to innovate from the ground up. “If the goal is just to develop applications, then adopting Llama’s structure is fine. But we’re aiming for AGI, which requires us to push beyond existing models,” Liang explained. He emphasized that China’s AI research has often lagged in training efficiency, requiring twice the computing power and data to match top global models. DeepSeek’s objective is to close this gap.
Despite its scientific ambitions, DeepSeek isn’t naive about the business side of AI. Liang understands that long-term survival in the AI race requires a balance between openness and proprietary advancements. But he remains firm in his belief that DeepSeek will stay open-source, unlike OpenAI and Mistral, which initially embraced openness but later pivoted to closed models.
“In disruptive tech, closed-source moats are fleeting,Even OpenAI’s closed-source model can’t prevent others from catching up.”
DeepSeek’s commitment to research over rapid monetization contrasts sharply with the prevailing business mindset in China. Many investors and entrepreneurs believe that AI innovation should be immediately commercialized, prioritizing short-term profitability over foundational progress. Industry figures like Zhu Xiaohu have dismissed AGI as impractical, arguing that companies should focus on making money first. But Liang disagrees, pointing out that the most successful American tech giants—Google, Apple, and Microsoft—were built on long-term R&D investments, not short-term gains. As DeepSeek scales its research efforts, its hiring strategy stands out. Unlike Western AI firms that aggressively recruit global talent, DeepSeek has built its models entirely with domestic researchers.
“There are no ‘inscrutable wizards’ here—just fresh graduates, PhD candidates, and young researchers with a passion for AI.”
DeepSeek’s organizational structure also plays a key role in its innovation. It operates with minimal hierarchy, allowing researchers to pursue their ideas freely. This approach has led to breakthroughs like the MLA architecture, which originated from a junior researcher’s personal curiosity.
“We don’t have rigid approvals for accessing compute resources or forming teams, If someone has an idea, they can start working on it immediately.”
China’s AI industry has often been characterized as a follower rather than a leader, but DeepSeek’s emergence challenges that perception. Liang believes that the biggest gap between Chinese and Western AI research isn’t time—it’s originality.
“We’ve grown used to adopting Western technology rather than driving our own breakthroughs, but if we keep following, we’ll always be behind.”
Despite the rapid progress, significant challenges remain. U.S. export restrictions on high-end chips have forced DeepSeek and other Chinese AI firms to find alternative solutions. While funding isn’t a problem, access to cutting-edge hardware remains a major hurdle. Yet, Liang remains unfazed.
“More investment doesn’t necessarily result in more innovation, If that were the case, big tech companies would have monopolized all innovation by now.”
DeepSeek’s rise signals a fundamental shift in the AI industry. No longer confined to Silicon Valley, AI’s future is increasingly being shaped in China. The implications are profound: If China continues to drive foundational AI research, the balance of power in global technology may shift dramatically in the coming years. As DeepSeek pushes forward with its AGI ambitions, the world will be watching closely.
