China's Tech Self-Sufficiency Drive Accelerates Despite US Export Restrictions

China's technology sector is accelerating its push for self-sufficiency in the face of tightened US export controls, with companies now raising $3.1 billion from mainland stock market listings in the first half of 2026. That figure is up more than fivefold from the same period a year earlier, driven by AI and chip companies, according to data from LSEG.

One of the most dramatic examples is the Chinese supercomputer LineShine, which recently ranked as the world's fastest despite using no GPUs. The system, built without access to advanced US graphics processors, achieved the top spot on global performance benchmarks through a novel architecture that relies entirely on domestic components.

AI Access and Open Models

The restrictions have also spawned a thriving underground economy for US-made AI tools. So-called "transfer station" sites buy API tokens abroad and distribute them to users inside China, according to a Wired investigation. Users on Telegram trade fake identities and proxy services to access Anthropic's Claude, bypassing the company's regional blocks.

Meanwhile, Chinese AI firms are releasing increasingly capable open models. Z.ai's GLM-5.2 matches the latest US models at finding security bugs, researchers told the Wall Street Journal. The development has fueled debate among US critics who question why Washington has not more aggressively restricted Chinese open source AI models, which could give Beijing a cyberwarfare advantage.

• $3.1B in tech IPOs on mainland China exchanges YTD 2026, up 5x year-over-year (LSEG/Reuters)
• LineShine supercomputer ranked fastest globally, built with zero GPUs (WIRED)
• Claude access in China maintained through proxy services and fake identities (WIRED)
• GLM-5.2 matches top US models on security bug detection (Wall Street Journal)

Implications for US Policy

The rapid pace of Chinese innovation challenges the effectiveness of US export controls. A separate Nikkei analysis asks whether China can build its own ASML, the Dutch company that dominates the lithography machines needed for advanced chip manufacturing. While full replication remains years away, Chinese firms have made notable progress on domestic tools.

What comes next is likely a period of increased scrutiny and potential policy adjustments. US lawmakers are already examining whether open source AI models represent a loophole in existing restrictions. For China, the goal is clear: reduce dependence on foreign technology across chips, supercomputing, and AI. The data suggests that goal is being achieved, albeit not without creating gray markets for banned services.