Our Take
Export controls work until they don't—and the price signal shows China's demand for restricted chips hasn't moved, only the distribution channel.
Why it matters
Nvidia's H100 and H200 chips are central to U.S. strategy to slow China's AI capability. When prices double on gray markets instead of demand falling, it suggests the restriction is raising cost and friction, not capability access—a meaningful but incomplete policy outcome.
Do this week
Supply chain leads: audit your chip procurement channels and document any pressure to source through non-official distributors before Q1 audits, so you can flag exposure early.
The price signal from China's gray market
Nvidia's most advanced AI accelerators, banned from export to China under U.S. Commerce Department restrictions, have doubled in price on China's black market (per Financial Times reporting). The H100 and H200 chips remain in high demand for training large language models, but official channels are closed. Buyers in China with access to gray-market suppliers are now paying multiples of retail price for the same hardware.
This is not a supply shortage in the traditional sense. The chips exist. Nvidia continues manufacturing them. The restriction creates an artificial scarcity in one geography while demand remains constant, driving prices upward for any buyer willing to pay premiums to unauthorized resellers.
What doubled prices actually tell us about sanctions
Export controls succeed in two ways: they reduce capability access, or they increase cost so steeply that projects become uneconomical. The black market price doubling suggests the second mechanism is working at the margin, but with a catch.
If the restriction were working optimally, Chinese AI labs would either forgo training runs or shift to permitted alternatives (AMD chips, older Nvidia models, domestically-produced processors). A doubled price that doesn't kill demand instead signals that some buyers have the budget to absorb it. State-backed research institutions or well-funded private companies can outbid the open market.
What the restriction does accomplish: it raises friction and cost for smaller labs and startups in China. It also creates enforcement complexity and legal liability for intermediaries. But it does not appear to have made the chips inaccessible to funded buyers.
The policy outcome is asymmetric. Large, well-capitalized AI efforts in China face higher costs. Smaller competitors face barriers. But the most capable actors may still proceed.
Monitor your supply chain now
If you source advanced Nvidia chips through subsidiaries, authorized resellers, or trading partners in or near China, rising black market prices are a leading indicator of enforcement tightening. Regulators will intensify scrutiny of gray-market flows. Document your procurement sources and timing. If you operate in regulated industries (defense, energy, finance), audit your supply agreements for any indirect routing through restricted geographies. The U.S. government has shown willingness to retroactively hold end-users accountable for chip provenance, not just distributors.