Image Source: Unsplash

By 2025, Chinese open-source large language models have reached a critical inflection point, positioning themselves as formidable rivals to top AI systems from the US. In an interview with Financial Sense, David Woo—CEO of David Woo Unbound and former head of macro strategy at Bank of America—suggests this development could become a major challenge for US tech giants seeking to recoup their significant recent investments in AI.

Divergent Strategies in the AI Arms Race

Woo framed the contest as a high-stakes struggle, stating that both nations view AI as "the decisive battle that the two countries are going to be fighting, vying for global supremacy and economic dominance." The US strategy, he explained, centers on monopolizing key elements of the AI supply chain to exclude China. This includes controlling access to advanced semiconductors and imposing export restrictions.

"The US strategy is that we’re going to control every choke point of the technology in the supply chain," Woo said. "They want to starve out the Chinese. They are doing everything to make it as expensive as it can be so that the only companies that can afford, for example, the expensive Nvidia chips are going to be deep-pocketed American tech monopolies."

In contrast, China's approach emphasizes commoditization and open access. Woo pointed to initiatives like DeepSeek, an open-source AI model, as evidence of Beijing's intent to make the technology widely available and cost-effective. "China wants to commoditize this entire technology," he noted. "In fact, they want to turn it into a zero-cost technology."

This divergence, Woo argued, could undermine US efforts if AI proves easily replicable. He likened DeepSeek or other Chinese models to pirated software from decades past: "DeepSeek is not proof of China’s technological prowess. It is proof in my mind that AI can be easily copied. Anything that can be easily copied has zero commercial value."

Signs of a Plateau?

Woo expressed skepticism about the rapid returns on America's massive AI investments, citing recent developments that suggest large language models (LLMs) may be approaching their limits. US tech giants such as Microsoft, Amazon, and Google are committing enormous sums to data centers, energy, and infrastructure.

Yet, Woo highlighted underwhelming progress: "The release of ChatGPT 5.0—the reception has been fairly lukewarm. Sure, it’s a bit better than 4.0, but it’s not like a step change." He also referenced Meta's delay in launching Llama 4 due to incremental improvements, concluding that "maybe the whole LLM... is starting to reach a plateau."

Current AI Model Rankings

Source: Lifearchitect.ai

(Click on image to enlarge)

Source: lifearchitect.ai

Supporting this view, Woo cited a recent MIT report indicating limited corporate adoption. "Only 5% of US companies are seeing AI make a positive P&L contribution," he said, warning that this could jeopardize monetization for US firms. Meanwhile, 95% of American workers are already using personal AI tools, reducing the incentive for large-scale enterprise investments.

China's Edge in Practical Applications

Woo emphasized China's focus on integrating AI into manufacturing, where it could yield significant productivity gains. "In terms of applications of AI, I think in manufacturing is where you’re going to get much more bang for the buck than in the service industries," he observed. China, which installs more industrial robots than the rest of the world combined, benefits from AI in "closed systems" like factories, where errors are minimized.

As US models plateau, Woo suggested, China has time to catch up, particularly in less demanding areas like inference chips. US sanctions, he argued, may inadvertently accelerate this: "Now you’re telling the Chinese, ‘I’m sorry, we’re not going to sell you anything at all.’ You think the Chinese have any incentive to respect any IP? No way."

Recent reports, such as Alibaba's development of a chip rivaling Nvidia's, underscore this momentum. For AI servers, Woo noted, advanced nanotechnology is less critical: "For AI chips, you don’t need to make 2-nanometer chips... So for China to play catch-up there, I think it’s going to be much easier."

Risks to US Economic Hegemony

The US stock market's high valuations reflect confidence in an American victory, but Woo cautioned against overoptimism. "The entire US stock market is pricing a US victory," he said. "Right now, the high valuation of US tech stocks suggests that the market is absolutely convinced that... the US will have a permanent lock on this technology."

If investments falter—echoing past tech bubbles—China could gain the upper hand. Woo described it as a "rabbit-versus-turtle kind of thing," with the US potentially "digging its own grave" through overhyped expectations and premature commitments. "The AI industry oversold the technology, and that could potentially be very costly for the US," he concluded.

This analysis comes amid broader geopolitical tensions, but Woo's focus on AI underscores its role as a linchpin for future global power. As benchmarks show Chinese models closing or exceeding the performance gap via open-source releases, the race's outcome remains uncertain, with profound implications for markets and international relations.

More By This Author:

The New Era Of Strategic Investing: Semiconductors, Rare Earths, And Uranium
The Dollar’s Next Act: Why The Greenback’s Best Days May Be Behind Us
Inside The U.S.-China Economic War: Trade, Tech, And The New World Order