That shift reflects growing recognition that U.S. efforts to slow China’s AI progress through export controls have not produced the expected results. In 2023, many Western analysts predicted that cutting off access to advanced processors would gradually choke China’s generative AI sector. The assumption was that AI progress depended primarily on raw computing power. Three years later, that view has been fundamentally challenged.

Demis Hassabis, founder and CEO of Google DeepMind, said Chinese AI companies are now only about six months behind the most advanced Western labs, a dramatic change from estimates just two years earlier that placed China several years back. Despite hardware limitations, Hassabis said, Chinese models have advanced rapidly and remain close to the cutting edge. The impact of DeepSeek’s 2025 release, which demonstrated competitive performance at far lower cost, still looms large in industry discussions.

The narrowing gap has sharpened concerns over access to computing power, particularly Nvidia’s advanced chips. Dario Amodei, CEO of Anthropic, used unusually stark language to criticize U.S. decisions allowing limited chip sales to China, comparing the move to providing an authoritarian government with capabilities akin to nuclear weapons. Granting access to such computing power, he said, could enable large-scale cyber operations and accelerate weapons development.

Sam Altman, CEO of OpenAI, sought to position himself as a measured but alarmed voice. Speaking at Davos, he warned that the United States is dangerously underestimating China’s progress while focusing too heavily on whether AI investment represents a financial bubble. Altman argued that the vast spending on data centers and infrastructure represents the largest buildout in human history and should be viewed as a security necessity rather than speculative excess. If Western governments and companies hesitate, he said, China will move quickly to fill the gap.

Elon Musk delivered a sharply different message. Once considered a controversial figure at the forum, Musk made a last-minute appearance in Switzerland and emphasized execution over caution. While others debated regulation and risk, Musk highlighted what he called an undeniable engineering milestone: the activation of Colossus 2, a computing cluster operating at the scale of a gigawatt. Musk dismissed calls to slow AI development in the name of safety and mocked rivals who advocate restraint, arguing that only speed and raw computing power can counter China’s advance.

Jeff Bezos offered a cooler, more pragmatic assessment. In remarks delivered outside the main Davos stage, the Amazon founder compared the AI infrastructure race to the early Industrial Revolution, when factories were forced to build their own power plants before centralized electricity grids existed. Bezos warned of what he described as an industrial bubble, not necessarily a financial crash but a period of inefficiency and massive capital waste. Many companies, he said, are likely to fail after burning through cash too quickly, even as the technology itself succeeds. As a major investor in Anthropic, Bezos said such shakeouts are a natural part of economic cycles.

While Western leaders debated strategy, China continued to build a national AI ecosystem largely insulated from Western constraints. The country has invested heavily in nuclear, solar and hydroelectric power to supply the enormous energy demands of data centers. At the same time, Chinese firms have optimized AI models to run on domestically produced chips, particularly Huawei’s Ascend series.

Although those chips still lag behind Nvidia’s most advanced processors, Chinese engineers have compensated by linking tens of thousands together into massive computing clusters using advanced optical networking. Intelligence and media reports have also pointed to a growing gray market in which advanced chips are smuggled into China via third countries and repurposed for high-performance computing in universities and military-linked facilities.

Analysts say China’s most powerful advantage lies not in hardware but in data. Unlike Western companies constrained by privacy laws, copyright litigation and public scrutiny, Chinese AI firms have broad access to data generated by 1.4 billion citizens. Government policy treats data as a national resource, allowing extensive use of information from multipurpose apps such as WeChat, which integrate payments, transportation, health services and communication. That access enables Chinese models to be trained on real-world human behavior at a scale unmatched in the West, fueling concerns about Chinese consumer technology collecting data abroad.

China’s approach also differs in its end goals. While Silicon Valley remains focused on achieving artificial general intelligence, Beijing emphasizes industrial deployment. AI is being embedded into manufacturing, electric vehicles and urban management systems, allowing China to generate immediate economic value. Automakers such as BYD and Xiaomi are integrating language and vision models directly into vehicles, turning them into increasingly autonomous systems.

Another factor reshaping the competition is human capital. For years, China suffered from a brain drain as top researchers trained in the United States remained in Western academia and industry. That trend is now reversing. Rising nationalism, concerns over treatment of Chinese researchers abroad and lucrative compensation offered by Chinese firms have drawn many scientists back to China. These returnees bring Western training and networks into laboratories in Beijing and Shanghai, operating under a system that closely integrates academia, industry and the military.

Participants at Davos warned that the West’s biggest mistake would be continuing to measure China through a purely Western lens. While the United States still dominates access to the most advanced chips, China is constructing an alternative AI ecosystem built on domestic hardware, highly efficient algorithms and vast data access.

The threat, several executives said, is not that China will replicate Western models but that it will develop cheaper, widely deployed AI systems embedded into physical products and exported across emerging markets in Africa, Asia and Latin America.

As Davos 2026 concluded, many executives described the moment as a historic turning point. The era in which technology companies could operate in a bubble of pure innovation is over. Artificial intelligence has become inseparable from geopolitics, and every advance in software or silicon is now part of a global strategic contest.

The six-month gap separating Western and Chinese AI capabilities, they said, is not a period of comfort but a narrowing window that may soon close.