California released a revised budget this week showing that its deficit has vanished. In January the state projected a shortfall of nearly three billion dollars this fiscal year and twenty-two billion the next; the new figures erase both, on the strength of sixteen and a half billion dollars in revenue above forecast over three years. The source is not a tax on AI companies. It is the capital-gains tax on the stock portfolios of high-income residents, portfolios that have soared on the promise of artificial intelligence. In the same document, the state’s own nonpartisan analyst observed that the spike in collections suggests the market has entered “bubble territory” and may be heading toward “an eventual bust.” So the most powerful subnational government in the world has balanced its budget on the unrealized euphoria of a market its own forecasters believe is a bubble — and acquired, in the act, the strongest reason a government can have to never call it one.
A Budget Built on Paper Gains
The mechanism deserves precision, because the precision is where the fragility lives. California did not become solvent by taxing the profits of the AI companies; many of them generate little profit to tax. It became solvent by taxing the capital gains of wealthy residents whose stock holdings have appreciated on the expectation of AI profits that have not yet arrived. The state’s revenue is therefore a second derivative of a promise — a tax on the market’s belief about a future that has not happened, collected as though the belief were already realized. When the market is euphoric, the portfolios swell, the capital-gains receipts pour in, and the budget balances. The solvency is real. What it rests on is sentiment.
This makes the state’s fiscal health a high-beta instrument tracking the AI rally, and the analyst who flagged “bubble territory” was not editorializing. He was describing the asset the budget is now collateralized against. Capital-gains revenue is the most volatile line a government can depend on, because it rises and falls not with the economy’s production but with the market’s mood, and a market’s mood, particularly near the top, is the least reliable thing in finance. California has tied its ability to fund schools, hospitals, and roads to the continued willingness of investors to pay rising prices for the promise of artificial intelligence. The budget is balanced for exactly as long as the euphoria holds, and not one quarter longer.
The analyst’s counsel was the only honest one available: treat the windfall as temporary, save against the coming bust, do not build permanent spending on a transient spike. It is also the least likely counsel to be followed, for a reason as old as government. A windfall is not saved by an institution that has discovered it needs the money; it is spent, and then it is depended upon, and then its disappearance is experienced not as the end of a bonus but as a crisis. The surplus that erased this year’s deficit will fund commitments that outlast the rally that produced it, and when the rally turns, the commitments will remain, and the deficit will return larger than it left. The save-for-the-bust advice assumes a discipline that fiscal windfalls have never, in any government, reliably produced.
The Regulator Becomes a Shareholder
The deeper consequence is not fiscal but political, and it is the reason this budget matters beyond California’s accounts. This is the jurisdiction most capable of regulating the AI industry — its legal home, the seat of the firms, the legislature that has repeatedly attempted to impose safety requirements the companies opposed. And it has now made its own solvency contingent on those same companies’ stock continuing to rise. A government that depends on an industry’s ascent for its budget is a government that cannot move against that industry without moving against itself, and the conflict does not announce itself as corruption. It operates as an instinct, a quiet recalculation in which every proposal to slow, tax, or constrain the boom is now also, unavoidably, a proposal to threaten the revenue the state has come to need.
This is the inverse of the claim that a South Korean official floated and his market punished. There, a government proposed to take a share of the AI windfall for its citizens, and the market dropped five percent at the suggestion. Here, a government has quietly become dependent on the windfall instead — not demanding a cut, but needing the boom to continue, which is the same dependence pointed in the opposite direction and far more dangerous, because it is invisible. The official who wanted a dividend announced an adversarial relationship to the windfall that the market could see and price. The state that balanced its budget on the rally announced nothing, and acquired an aligned relationship to the windfall that no one will price until it is too late to.
A regulator that has become a shareholder does not regulate the company downward. It cannot, without writing down its own holdings, and the holding here is the budget itself — the schools and the hospitals and the salaries that the capital-gains revenue now funds. Every official who must choose between a constraint on the industry and a hole in the budget will feel the budget’s weight, not as a bribe but as a responsibility, and will resolve the choice accordingly. The capture is complete precisely because no one had to be captured. The state simply took the easy revenue, and the revenue came with an interest, and the interest is the continuation of the very thing the state exists to oversee.
What This Means
The most effective capture is the kind that requires no capturer. No lobbyist arranged this; no official was bought. The state confronted a deficit and a windfall in the same season, applied the windfall to the deficit because that is what windfalls are for, and in the arithmetic of a single budget cycle bound its own fortunes to the ascent of the industry it is uniquely positioned to restrain. The alignment was not chosen. It was the path of least resistance through a balance sheet, and the path of least resistance is how institutions are captured far more often than through any deliberate corruption, because it leaves no fingerprints and produces no scandal — only an interest, quietly acquired, that bends every subsequent decision.
So the regulator that should be the boom’s most serious check has become, instead, one of its largest stakeholders, and the conversion happened in the language of responsible governance — a balanced budget, a closed deficit, a fiscal house in order. The achievement is real and the dependence it created is real, and the two are the same fact. California will spend the next several years unable to consider any serious constraint on the AI industry without first considering the cost to a budget that the industry’s stock price now underwrites, and it will experience that constraint not as capture but as prudence, which is what capture feels like from the inside.
I have watched governments balance their books on a single rising asset before — the oil state on the barrel, the financial center on the trade, the boomtown on the mine — and the pattern does not vary in its ending. The revenue is treated as permanent because depending on it requires believing it is. The warnings about the bust are filed and not acted on, because acting on them would mean surrendering the surplus while it is still arriving. And when the asset finally turns, the government discovers that it did not merely lose a windfall; it lost the only thing holding its commitments up, and it had spent the years of plenty acquiring the reasons not to see it coming. The analyst said save for the bust. The budget already spent the warning.