US chip embargo backfired, made China's most valuable company

When Washington tightened export controls on advanced semiconductors to China, the intent was clear: starve Beijing's AI ambitions by cutting off the chips that power AI systems. But the announcement of restrictions targeting Nvidia, AMD, and their successors may have backfired spectacularly.

CXMT Corp, a memory chipmaker once dependent on foreign technology, made its debut on Shanghai's STAR Market last month with a $9.8 billion IPO raise—one of China's biggest in years. On its first trading day, the stock surged 466% in a single session, eclipsing the Industrial and Commercial Bank of China to become the country's most valuable listed company. Regulators and state funds intervened within days when tech stocks threatened to derail the debut in July, demonstrating Beijing's commitment to the listing.

The speed of CXMT's path to market was extraordinary. The company was fast-tracked through a preliminary review process reserved for strategically vital companies and went from filing to trading in under eight months—a timeline that normally takes years. The capital raised is earmarked for the chip capacity Beijing believes it needs to compete in the global AI race. The IPO is not a singular event; it is the opening salvo of a broader wave. DeepSeek, a Chinese AI company, is weighing a listing at a $71 billion valuation. Moonshot AI, whose Kimi K3 model attracted international attention, is filing in Hong Kong. Z.AI and MiniMax are preparing to follow.

Beyond manufacturing, China has fundamentally restructured how it finances its AI ambitions. Historically, Beijing backed strategic industries through subsidies and state lending. Now it is tapping $26 trillion in household savings accounts—the world's largest pool of private capital. This capital is flowing into an ecosystem with stunning cost advantages. Chinese tech firms borrowed at an average coupon of 1.9% this year, more than 300 basis points below the 5.25% US peers are paying, the widest gap since at least 2015. According to UBS, China's leading AI models cost less than 10% of what OpenAI and Anthropic spend to train. API prices for major Chinese models run below 20% of comparable global offerings. Chris Miller, author of Chip War and a professor at Tufts University, told Bloomberg that while US firms have had greater access to capital overall, financing costs in America are rising, and the gap in chip quality between Chinese domestic and foreign semiconductors is closing. Washington's embargo, designed to slow Beijing, may have forced it to build better alternatives.