The Great Mispricing

Executive briefing

Three quantities that govern how a modern state acts are each measured against a reference that stopped corresponding to anything. Government capacity is quoted against total managed expenditure, which is almost entirely committed. Political influence is priced against the cost of running a campaign, rather than against the value of the decision the campaign settles. Enforcement is resourced against the volume of transgressions, rather than against what a transgression is worth to whoever commits it. Every figure is correct. Every one is denominated wrongly. Nobody chose the moment it happened, which is why nobody is answerable for it.

The discretionary envelope is smaller than a single building. The Office for Budget Responsibility scores every measure at every fiscal event and publishes the lot. Across the nine events of the 2019–2024 parliament, spending measures total £680.8bn over five years. Strip Covid at £290.2bn and energy support at £68.2bn, both shock responses on the Treasury’s own delineation. Strip a further £160.2bn of envelope-setting and accounting reclassification, including the cessation of EU contributions, which the OBR itself describes as the removal of a fiscally neutral holding assumption rather than a decision. What remains — actual choices about what the state should do, across every department — is £162.3bn, or £32.5bn a year. A one-gigawatt AI campus costs about $38bn upfront on Epoch AI’s May 2026 model. That is £29.9bn. One campus is one year.

Britain is the median. Discretionary capacity lands between roughly 0.5 and 1.3 per cent of national output in every state examined: the Netherlands, Italy, Canada, Germany and the UK. Each measures itself on a different published instrument, so the spread is a range rather than a series, but it is tight enough to establish that the constraint is structural rather than national. Italy’s entire budget law, after consuming a full parliamentary autumn and confidence votes in both chambers, mobilises about €22bn and pays for itself.

Regulation has not substituted for spending. The Digital Markets Act is calibrated correctly, at up to ten per cent of worldwide turnover — a $39bn instrument against Apple. The Commission used €500m, then declined to escalate. That sum is about two per cent of Apple’s annual selling, general and administrative expense, which is the line that already carries counsel, compliance and audit. A penalty that lands as a variance in the function built to absorb it is not deterrence failing at the margin; it is an instrument in the wrong category, because the question a fine asks is how much, and how much is the question the counterparty answers better than the regulator.

What survives is permission, and it works only by anticipation. Nvidia abandoned Arm six months before its FTC trial; Adobe abandoned Figma the day before its CMA response fell due, paying a $1bn break fee. Across two transactions and five jurisdictions, no regulator issued a final order. The state’s remaining instrument is an option whose value lies in never being exercised, which is also why it is rarely used.

The state that can allocate sits inside the capital structure. Germany amended its constitution in March 2025 to authorise €500bn over twelve years. State Grid Corporation of China announced ¥4tn for 2026–2030: about €525bn, one company, five years, more than the constitutional amendment bought over twelve. None of this allocates well, and the IMF puts Chinese local-government vehicle debt at ¥58tn. But the shareholder state fails faster and redirects faster, and its losses land on a balance sheet rather than in a chamber.

Influence is cheap because exit is cheaper. Firms whose assets can move do not buy political outcomes; they relocate, and the state finds itself bidding. The Netherlands lost Unilever, lost Shell, then paid €2.51bn to retain ASML on the explicit written assumption that the company would keep its statutory and fiscal seat — and could not then deliver the grid capacity its own letter had named. Firms whose assets cannot move have no such option. A stablecoin has no site; its regulatory treatment is decided in a legislature or nowhere.

And the instruments cannot see what is bought. In Clacton at the 2024 general election, a candidate could spend £20,660.72, a party £54,010, a registered campaigner £17,553 — three limits descending from an anti-bribery statute of 1883, all denominated in pounds spent inside one seat during one window. A private individual gave Nigel Farage £5m, which is 5.3 per cent of everything every party and campaigner in the United Kingdom spent contesting all 650 seats. No instrument in the regime is written in that unit. A police assessment of the constituency limit was time-barred at twelve months; a standards investigation paused when Farage resigned the seat. An unenforced rule does not restrain money. It certifies it.

The consequence. A single national permit is worth little, because refusal costs the refuser more than the applicant: one state loses the investment and keeps the refusal, while the applicant loses a location and keeps everything else. Aggregate enough sovereigns and the proportion inverts — a market of 450 million can refuse on terms that cannot be routed around. That is why the European Union is treated in the corporate account of the world as an adversary rather than an inefficiency, and why the hostility it attracts is the evidence that the instrument works. Coordination between states is now the only thing that makes sovereignty mean anything.