Nvidia's State-Stake Odds Tick Up as Black-Market Chips Double in Price

Grey-market Nvidia accelerators in Shenzhen doubled in price between late 2024 and mid-2026, and along the way they did something more interesting than become expensive: they became a proxy for whether Washington would end up owning a slice of the company they run on. The price of an H100 on the Chinese black market, which traded around $20,000 when export controls tightened in late 2022, climbed past the $40,000 mark as buyers routed shipments through intermediaries in Malaysia, Vietnam and the Gulf. That doubling is the cleanest available read on the gap between official semiconductor policy and the actual demand inside China's data-centre build-out. It is also, increasingly, the data point a small but growing corner of the prediction-market industry is using to handicap the probability that the United States government takes an equity stake in Nvidia itself.

The thesis is ungainly but worth taking seriously. For three years the dominant frame on Nvidia has been a bullish one, carried by revenue guidance, hyperscaler capex announcements and the steady drip of sovereign AI deals. The counter-narrative has been industrial-policy shaped: export controls designed to slow Chinese access to frontier compute, and the unresolved question of whether those controls ultimately benefit the company they are meant to constrain. A state equity stake in Nvidia is the logical endpoint of that counter-narrative. It is also the kind of outcome that prediction markets are unusually well-suited to price, because the relevant inputs are public (CHIPS Act disbursements, Treasury statements, congressional committee language) and the relevant question is binary enough to fit a contract.

From Washington think-tanks to a prediction market

The industrial-policy debate around Nvidia has lived, until recently, inside a small set of Washington venues. The Center for a New American Security published an early paper on the national-security case for semiconductor equity stakes. The Information reported on informal discussions inside the Biden administration's Commerce Department about a potential sovereign investment vehicle. Senator proposals to expand CHIPS Act authorities to include equity, rather than grants, surfaced in committee markup in late 2024. None of those conversations produced a concrete instrument, but they established the menu of options: direct equity, a sovereign wealth-style fund, a Treasury backstop tied to strategic industries, or a quiet accumulation through existing federal vehicles.

The prediction-market line is newer and more sober. Kalshi and Polymarket both listed contracts during 2025 on whether the US government would hold a material equity position in a designated set of AI-relevant companies by a specified date. The contract volumes have been modest in dollar terms but the price signals are useful: implied probabilities moved from low single digits in mid-2025 to a meaningfully higher band by mid-2026, with the largest tick coming after Nvidia's own disclosures about supply allocated to restricted end-users and the Treasury's expanding use of the Defense Production Act for chip-related investments. The grey-market Chinese chip price, by contrast, has moved in one direction for three years and shows no sign of reversing. That is the asymmetry the prediction market is gradually pricing in.

What a state stake would actually mean

Equity is a different instrument from a grant, a loan, or a procurement contract. A grant subsidises production without transferring control. A loan attaches covenants. Equity transfers governance rights, however diluted, and creates a permanent seat at the table on capital allocation, export licensing and customer selection. The historical precedent is partial: the United States took equity in General Motors during the 2008-2009 auto industry restructuring, and in AIG via the Treasury's TARP vehicle. Both cases involved acute crisis, both involved temporary holding periods, and both involved companies whose survival was in question. Nvidia is not a distressed asset. Its market capitalisation, its order book and its pricing power all place it firmly in the category of companies the government has never had to rescue.

That is exactly what makes a stake unusual. A state equity position in a functioning, dominant, strategic firm is an instrument of industrial policy on a different scale. It would convert a regulatory relationship into a governance relationship. It would create, for the first time, a direct US fiscal claim on the upside of the AI compute cycle. And it would do so at a moment when the dominant customer base for Nvidia's most advanced chips is a small group of US hyperscalers, several of which are themselves under sustained antitrust scrutiny, all of which are operating under increasingly detailed export-control rulebooks. The geometry is novel enough that the legal authorities are not fully settled. The Congressional Budget Office has flagged the question of fair-value accounting for any direct federal equity acquisition; the Government Accountability Office has noted that prior equity holdings were liquidated at a loss relative to their peak mark.

The grey-market signal

The Chinese grey-market price for restricted Nvidia parts is, in the absence of better data, the cleanest real-time read on a specific question: are the export controls actually constraining supply on the ground? Three years of data suggest the answer is mixed. Volume has fallen relative to the pre-control baseline, but prices have risen sharply, which is what supply restriction looks like in any market. Intermediaries have proliferated. End-users have moved to lower-tier parts (A800, H800, the China-specific SKUs Nvidia itself designed to comply with the original rules) and to domestic alternatives where they exist. The doubling of black-market prices is, in that sense, a confirmation that the policy is biting at the high end and that demand has not collapsed.

For a prediction market trying to handicap US industrial policy, the grey-market signal matters because it tells you whether the strategic stakes are rising or falling. If prices were flat or falling, the case for further intervention would weaken. The fact that they are doubling tells you that the gap between official policy and on-the-ground reality is widening, and that political pressure for a more aggressive response is building. A state equity stake is one of the more aggressive responses available.

Open questions

The cleanest live questions for the rest of 2026 are four. First, whether the Treasury or Commerce will formalise an equity instrument, or continue to express preference for grants and loans. Second, whether congressional appetite exists to amend the CHIPS Act to authorise direct equity acquisitions in companies that did not receive CHIPS funding. Third, whether Nvidia's board, which has historically resisted any dilution of the founder-controlled structure, would accept a negotiated stake under any terms. Fourth, whether the grey-market price itself is the right signal, or whether it is being distorted by end-user stockpiling ahead of expected further restrictions, in which case a future reversal could undercut the industrial-policy case just as decisively as the doubling built it up.

The prediction-market contracts will resolve one way or another. The grey-market price will move on its own cycle. The interesting question is whether the two will tell the same story at the end of it.

*Sources: The Information; Center for a New American Security; Congressional Budget Office; Government Accountability Office; Kalshi market data; Polymarket market data; Reuters; Bloomberg. * Desk note: Monexus frames this as an industrial-policy story whose centre of gravity has moved from Washington think-tanks to a prediction market. The wire line on Nvidia has been bullish by default; the prediction-market line is more sober, and the Chinese grey-market data is the strongest available counter-narrative.