Anthropic, one of the world's leading AI research labs, has issued a stark warning: the outcome of the US-China AI race will likely be decided by 2028, and the stakes could not be higher. The company outlined two dramatically different futures in its essay "2028: Two Scenarios for Global AI Leadership." In one scenario, democracies maintain control of the world's most powerful AI chips and establish global norms for safe AI development. In the other, authoritarian regimes like China gain the upper hand, potentially ushering in an era of unprecedented digital surveillance and state repression.

The reason for the 2028 deadline is sobering. Many experts, including analyst Matthew Berman, believe this timeframe coincides with the potential arrival of self-improving AI, where artificial intelligence systems can enhance themselves without human intervention. Whichever nation achieves this breakthrough first could secure an insurmountable, permanent advantage in the global AI race.

Why Does Compute Power Matter More Than Talent or Data?

Anthropic identifies one critical bottleneck as the most important factor in AI development: compute, which refers to the specialized computer chips, or GPUs (graphics processing units), required to train and run large AI models. The United States currently maintains a decisive advantage because American companies like Nvidia develop the world's most capable chips. The US government has leveraged this advantage by enforcing strict export controls, severely limiting China's access to these critical components.

According to Anthropic's research, these export controls have been "incredibly successful" at slowing China's AI trajectory. However, China has demonstrated remarkable resilience in circumventing these restrictions through multiple strategies. Chinese AI labs exploit regulatory loopholes, engage in hardware smuggling, and deploy what Anthropic calls "large-scale distillation attacks" to extract innovations from American companies.

Despite these constraints, China possesses significant advantages that go beyond just acquiring chips. Estimates suggest that approximately 50 percent of the world's AI researchers are Chinese, providing a deep talent pool that fuels impressive algorithmic innovations. Chinese AI labs have developed highly efficient models optimized for less powerful hardware, a necessity given the chip restrictions they face.

How Is China Closing the AI Gap Despite US Export Controls?

China's multi-pronged strategy to bridge the AI gap involves several key approaches:

• Distillation Attacks: Chinese AI labs train their own models by feeding them outputs generated by superior US-developed models, effectively harvesting American innovation and allowing smaller, less resource-intensive models to mimic the performance of larger counterparts at a fraction of the cost.
• Algorithmic Innovation: Chinese researchers develop pioneering research and highly efficient algorithms that maximize performance from constrained hardware, demonstrating independent capacity for innovation beyond simply circumventing US restrictions.
• Talent Concentration: With roughly half of the world's AI researchers based in China, the country maintains a significant advantage in developing new techniques and optimizing existing approaches for resource-constrained environments.
• Hardware Smuggling and Regulatory Loopholes: Chinese companies exploit gaps in enforcement and regulatory frameworks to acquire restricted components through indirect channels and alternative suppliers.

The true scale of China's distillation attacks remains contested. Many industry observers argue that reported instances are often relatively small-scale or could be justified as competitive benchmarking. Regardless, Anthropic specifically flags these as "large-scale" threats in its foundational essay.

What Happens If China Wins the AI Race?

Anthropic's second scenario paints a chilling picture of a world where China overtakes the frontier of AI development by 2028. This outcome would hinge on the US failing to tighten export controls and disrupt China's algorithmic innovations and distillation attacks. If this occurs, authoritarian values, not democratic principles, would shape the norms governing artificial intelligence worldwide.

The real-world implications extend far beyond economic competition. Authoritarian powers would deploy advanced AI for unprecedented social control, enabling sophisticated censorship, pervasive population monitoring, and predictive policing on a global scale. This technological leverage would also significantly strengthen military capabilities, fundamentally altering the global balance of power and increasing the risk of international instability. The cost of losing this AI race extends far beyond economic competition, impacting fundamental human rights and global security for decades.

Is the US-China AI Race Actually Being Discussed at the Highest Levels?

Despite the stakes, artificial intelligence remained largely an undercurrent at the recent Trump-Xi summit in Beijing in May 2026, rather than a central focus of negotiations. President Donald Trump arrived with an entourage of tech and business leaders, and Nvidia CEO Jensen Huang joined the trip at the last minute, hoping to secure a deal to sell China Nvidia's H200 chip, which has not been delivered to China despite US approval for its sale.

However, the talks appeared to focus more on limited questions of trade, particularly what observers called the "three Bs": Boeing, beef, and beans. US Trade Representative Jamieson Greer told Bloomberg that US export controls on semiconductor chips were not a major part of the talks. Several observers had hoped the summit would produce a broader framework on AI governance or US-China technological cooperation, but those hopes largely went unfulfilled.

"The talks looked more like a trade show, especially from the US side," said Chong Ja Ian, a professor of international relations at the National University of Singapore.

Chong Ja Ian, Professor of International Relations at the National University of Singapore

While AI remained largely unaddressed in formal discussions, it likely stayed on the leaders' minds as an undercurrent in broader geopolitical talks, which centered on Taiwan and Iran. Taiwan's dominance of semiconductor manufacturing makes it strategically vital to both nations, and the island's independence remains a flashpoint in US-China relations.

What Role Is Quantum Computing Playing in the Geopolitical Race?

Beyond traditional AI, China is also making quantum computing a national priority, investing heavily through state funding, policy support, and coordinated development strategies. Quantum computing represents another frontier where geopolitics and physics have become inseparable, as the nation that cracks fault-tolerant quantum computing first will hold an asymmetric advantage in cryptography, materials science, drug discovery, and other domains that determine economic and military power.

China's 15th Five-Year Plan (2026-2030) names quantum technology first among seven "future industries" designated as new economic growth engines. The National Venture Guidance Fund has allocated 121.8 billion yuan (approximately $17 billion USD) across three regional quantum-focused investment vehicles, covering Beijing-Tianjin-Hebei, the Yangtze River Delta, and the Guangdong-Hong Kong-Macao Greater Bay Area. China's quantum industry reached 11.56 billion yuan in 2025, with annual growth rates exceeding 30 percent.

Notably, US export controls requiring licenses for quantum computers, cryogenic systems, and related control electronics have accelerated Chinese domestic development rather than slowing it, mirroring what happened with semiconductor export controls. Chinese companies like Origin Quantum, China Telecom Quantum, and others are rapidly advancing their capabilities. Origin Quantum, for example, released Origin Wukong-180, its fourth-generation superconducting quantum computer built on a 180-qubit chip in May 2026, with all four core systems fully self-developed.

The convergence of AI and quantum computing advances suggests that the 2028 deadline identified by Anthropic may apply to multiple transformative technologies simultaneously. As both the US and China race to achieve breakthroughs in AI, quantum computing, and semiconductor manufacturing, the window for maintaining technological leadership continues to narrow. The decisions made in the next two years will likely determine which vision of the future prevails: one shaped by democratic values or one shaped by authoritarian control.