# How Washington Policies Enable China's AI Advancement Despite US export controls, China maintains a competitive AI edge through American infrastructure. - **Source:** [80,000 Hours](https://80000hours.org) - **Published:** 2026-11-10 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/7ec16881-d3ef-4a01-990b-179c1b15353c ## The Six-Month Gap Despite having significantly less computing power than the United States, China remains only six months behind in AI development. This narrow gap persists because Chinese companies leverage American technology and proprietary models. While the US possesses roughly 16 times more computing power, China's ability to innovate within these constraints remains remarkably high. ## The Role of Distillation Chinese labs utilize a technique called distillation to close the performance gap. By querying advanced American models millions of times, they record reasoning patterns to train their own systems. This process effectively mimics the capabilities of frontier models, providing Chinese firms with a massive shortcut that saves significant time and computational resources. ## Cloud Rental Workarounds US export controls prevent the sale of top-tier chips to China, but they do not stop Chinese companies from renting them. Firms like Tencent and ByteDance access NVIDIA hardware through data centers in Southeast Asia. This allows them to train sophisticated models remotely, bypassing physical hardware restrictions imposed by Washington's current trade policies. ## Smuggling and Gaming Chips Beyond cloud rentals, Chinese entities acquire American chips through illicit smuggling networks and by purchasing gaming-grade graphics cards. Smugglers have been known to relabel high-end AI chips to evade detection. These methods ensure that even with strict export limits, a steady supply of American hardware continues to fuel China's domestic AI research. ## The Incentive Conflict American AI companies face a dilemma regarding model security. Blocking Chinese users who copy models results in direct revenue loss, as these entities pay for access like any other customer. Consequently, some firms maintain weaker defenses, allowing Chinese developers to harvest data and replicate the logic of the world's most advanced AI systems. ## Corporate Influence on Policy Industry leaders like NVIDIA CEO Jensen Huang have lobbied against restrictive export controls, citing lost market access. Meanwhile, former officials like David Sacks faced scrutiny for holding stakes in AI companies while shaping government policy. These financial incentives complicate the US effort to maintain a technological lead through strict regulatory measures. ## Hardware Bottlenecks China struggles to produce advanced AI chips domestically due to a lack of specialized manufacturing equipment from ASML and high-speed memory. Even Huawei, a leader in Chinese chip design, relies heavily on foreign components. These structural limitations remain the primary factor preventing China from fully closing the gap with American AI capabilities. ## The Human Capital Factor China benefits from a deep pool of domestic talent, with over half of the world's top AI researchers originating from Chinese universities. Many of these experts are choosing to remain in China, providing the intellectual foundation for companies like DeepSeek and Zhipu. This human capital is critical to their rapid model development. ## Different Strategic Goals While the US focuses on building the most powerful frontier models, China is pursuing a broader strategy. Beijing aims to integrate AI across every factory, hospital, and government office. This approach prioritizes widespread industrial application over the singular pursuit of the most advanced chatbot, potentially creating a different type of competitive advantage. ## Enforcement and Effectiveness Export controls are effective when strictly enforced, as demonstrated by the drop in server imports after Malaysia implemented new permit requirements. However, the current system is porous. Without closing loopholes like international cloud rentals and model distillation, the US policy of restricting hardware access remains insufficient to significantly slow China's progress. ## The Security Dilemma The emergence of highly capable models like Claude Mythos has raised national security concerns regarding hacking and safety. If China continues to replicate these models within six months of their release, the US faces a perpetual race. Policymakers must decide whether to prioritize speed or implement rigorous safety testing before releasing new models. ## Future Outlook The AI race is not just about computing power; it is about policy choices and corporate incentives. As long as American chips and models remain accessible through third-party channels, China will likely maintain its current trajectory. The US must decide if its current regulatory framework is truly designed to lead or merely delay. ## Sources covering this story - https://80000hours.org/podcast/episodes/china-ai-race-chips-slowdown/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._