China’s Chip Paradox: $54 Billion Valuation, 20% Yields

Moore Threads’ 425% first-day pop tells one story. The company’s own warning of “over-exuberance”—followed by a 19% drop—tells another. The gap between those two narratives is where China’s semiconductor — as explored in the economics of AI compute infrastructure — reality lives.

The numbers are intoxicating. A $54 billion valuation. An 870x price-to-revenue ratio. Beijing’s $70 billion support package. For context, Nvidia trades at 30x revenue. The market is pricing Moore Threads at nearly 29 times Nvidia’s multiple for a company that cannot access the equipment needed to compete with Nvidia.

The chokepoints tell the real story. Cambricon operates at 20% yields—four of every five chips discarded. SMIC remains locked out of EUV lithography. High-bandwidth memory comes from “smuggled” SK Hynix and Samsung supply chain — as explored in how AI is restructuring the traditional value chain — s, not domestic production. TSMC’s advanced nodes remain blocked by export controls.

This isn’t a technological gap. It’s a manufacturing physics problem that money alone cannot solve.

The scale gap quantifies the challenge. Through 2026, Nvidia will ship an estimated 1 million H20 chips to China (the export-compliant version). Huawei’s Ascend 910C targets 600,000 units. Cambricon projects 500,000. Combined domestic production barely matches what a single American company can ship under restricted conditions.

The Morningstar analyst’s assessment is blunt: replicating the chip-plus-software ecosystem requires more than a decade, given user inertia, fabrication gaps, and equipment restrictions.

What the euphoria actually represents: not confidence in technological parity, but a bet that Beijing’s strategic priority will sustain valuations regardless of competitive reality. The $70 billion support package isn’t investment—it’s a signal that these companies won’t be allowed to fail.

National champions can survive on policy. They cannot compute on it.