A Chinese researcher told a colleague the trick for big research budgets at home: say the United States is already far ahead, and China will never catch up without more money. The American version is the mirror image. Say China is already ahead, and we are falling behind. The money flows on both sides. Whether the science is better for it is a separate question.
They raise a national-security budget by warning that the other side is already ahead. The money would pay to build a machine that nobody, including the winner, can control.
That exchange, reported in mid-2026, belongs in the oldest chapter of the national-security playbook, dusted off for a new technology. In the Cold War the line was "missile gap." Now it is something closer to a "shoggoth gap": the claim that the other side is about to build an uncontrollable machine mind, so we must build one first, and faster, and with fewer brakes. The budget logic is the same as the missile-gap years, except the money would buy a mind nobody can control rather than a missile count.
How a gap becomes a blank check
The original missile gap was a political instrument. In the late 1950s American officials and candidates warned that the Soviet Union was racing past the United States in intercontinental ballistic missiles. The claim moved money, elections, and force structure. When better intelligence came in, the gap mostly dissolved. The Soviets had fewer missiles than the speeches required. The spending did not reverse as cleanly as the estimate did.
That pattern is useful because it is boring: two rival systems, each with labs and ministries that grow when the threat narrative grows, and each with leaders who would rather overfund a scary story than get caught underfunding a real one. Nobody has to conspire. The safe bureaucratic move is always the same: assume the worst about the other side, ask for more, and treat restraint as naivete.
Swap missiles for frontier models and the incentives line up again. A lab that wants chips, talent, and regulatory patience has every reason to say the rival is six months from a decisive lead. A ministry that wants a larger AI budget has every reason to believe it. A politician who does not want to be blamed for "losing" has every reason to fund both. Nobody in that loop is paid to ask whether the finish line is a prize anyone can hold.
A gap is a budget story before it is a fact. The question is always who benefits from you treating it as weather.
What the shoggoth gap smuggles in
The modern version is sharper than the Cold War one, because the object of the race is different. A missile is a weapon a state owns. An artificial superintelligence, if the people building it are honest about the risk, is a system no one has shown they can control. Getting there first means being first to answer to something you cannot steer.
The pitch needs you not to notice. It needs the finish line to feel like a silo key, something a president holds. Then "falling behind" sounds like a national emergency, and "slow down until we know how to control it" sounds like unilateral disarmament, before any evidence arrives.
What actually has to be true for the emergency version to hold. The rival must be both capable of building an uncontrollable superintelligence soon, and eager enough to do it that no shared interest in survival can slow them. It is rarely argued with the care it deserves. More often it is asserted, then used as a permission slip for the next training run.
China has signed international AI risk declarations, published its own governance proposals, and sat in direct talks with the United States on the subject. Nobody has shown they will build an uncontrollable machine no matter what we do, but the money arrives as if they will.
Both sides can be right about the trick and wrong about the cure
The funding trick works because there is real competition underneath it. American and Chinese labs do watch each other. Export controls on advanced chips exist because the industrial contest is real. The politics of those controls and the broader China myth that turns rivalry into destiny are the live questions.
In a normal tech race, more money buys you a better product you still own. In a race to superintelligence, more money buys you a faster approach to a system that, on the builders' own risk estimates, may not stay a product. The same pitch that filled missile silos now fills data centers, and the people writing the checks are rarely asked to sit with the difference.
When both capitals run the same play ("they are ahead, fund us"), both sides accelerate. Each side's recklessness becomes the other side's evidence. That is how you get a race that selects for the least careful lab, and how a budget heuristic becomes an extinction pathway without anyone needing to want extinction.
The move that is not in the briefing
Cold War adversaries eventually noticed that some races have no acceptable finish. Still enemies, they built verification around the weapons that could end them both, because unbounded competition left everyone worse off than a rule that bound them equally. They wanted the paperwork because they distrusted each other.
Superintelligence has the same structure, only cleaner. No major power benefits from a system nobody can control. That shared interest is why a treaty can exist: a binding, verifiable halt on superintelligence, with inspectors, consequences, and the same paperwork that made arms control real. A speech will not stop a lab, and neither will a promise to slow down that dies the moment a rival looks ahead.
That pitch has no room for that option, because it is a fundraising device. Its job is to make the next appropriation feel inevitable. "We are falling behind" stands in for an argument it has not earned. Refuse to let a Cold War budget trick decide whether the species keeps the keys.