U.S. export controls on AI chips are being circumvented as Chinese firms like ByteDance access powerful NVIDIA GPUs remotely through data centers in countries like Malaysia. This loophole, combined with complex corporate shell structures, allows them to train frontier models, rendering the current import-focused restrictions largely ineffective.

Faced with restrictions on advanced NVIDIA chips, China is leveraging its electricity advantage to run vast numbers of older-generation GPUs in parallel. This hardware constraint forces a focus on software, with Chinese labs developing sophisticated algorithms and compute methods to leapfrog the hardware deficit.

Before the 2019 US sanctions, Huawei was on a trajectory to dominate the AI hardware space. It had superior talent across the entire stack—from software and networking to AI research—and had already become TSMC's largest customer, indicating it would have likely outcompeted NVIDIA.

The most dangerous policy mistake would be reverting to a 'sliding scale' that allows China to buy chips that are a few generations behind the cutting edge. In the current era of AI, performance is aggregatable. China could simply purchase massive quantities of these slightly older chips to achieve compute power equivalent to frontier systems.

US sanctions intended to cripple China's AI progress have instead forced it to create a robust, independent semiconductor ecosystem. By cutting off access to NVIDIA chips, the policy catalyzed an aggressive domestic mobilization, led by firms like Huawei, significantly reducing China's reliance on American technology and creating a powerful, self-sufficient competitor.

Facing U.S. export controls on NVIDIA chips, Chinese AI lab Zhipu is exploring custom chip design. This move, driven by necessity and surging demand for its powerful, affordable models, shows how geopolitical pressure is inadvertently accelerating China's development of a self-sufficient, vertically integrated AI hardware ecosystem.

China is compensating for its deficit in cutting-edge semiconductors by pursuing an asymmetric strategy. It focuses on massive 'superclusters' of less advanced domestic chips and creating hyper-efficient, open-source AI models. This approach prioritizes widespread, low-cost adoption over chasing the absolute peak of performance like the US.

Sebastian Malabai argues that U.S. chip export bans are ineffective because China circumvents them by renting GPU capacity in other countries and using "distillation" to reverse-engineer and copycat advanced U.S. models. This suggests a need for a new strategy focused on collaborative safety.

China is creating cheaper, 'good enough' AI models by training them on the outputs of US frontier models. This technique, called distillation, undercuts the revenue of US AI companies, threatening their ability to service the massive debt from their infrastructure buildout.