Apocalypse Tomorrow

Or: Debt, AI and the Future We Have Already Spent.

This week the United States passed $40 trillion in gross federal debt.

The number is almost useless. It is too large to picture, too large to compare with a household, too large even to function properly as a political metaphor.

More useful is perhaps this: around $32.3 trillion is debt held by the public – Treasury securities owned by investors, pension funds, banks, central banks, the Federal Reserve and foreign governments. That is now approximately the size of the American economy. The Congressional Budget Office expects publicly held debt to reach 120% of GDP by 2036 and 175% by 2056 if current policies broadly continue. Net federal interest payments are already around $1 trillion a year and are projected to double over the next decade.

That begins to look like a debt problem.

But first we have to decide what debt is.

But debt is not what we think it is.

A Treasury bond is simultaneously two completely different things.

To the United States government it is a liability.

To whoever owns it, it is an asset.

More than that, US government debt is one of the principal pieces of infrastructure of global capitalism. Treasuries are reserves, collateral, benchmarks for pricing other assets and somewhere for enormous quantities of capital to hide when everything else looks frightening.

The American government’s debt is, in other words, everybody else’s safe asset.

That makes the usual household analogy almost completely useless.

Stephanie Kelton and the Modern Monetary Theory school make the strongest version of this argument: a state issuing debt in a currency it itself issues cannot run out of that currency in the way Greece, a company or a household can run out of money. Its ultimate constraint is not dollars but real resources and inflation. Olivier Blanchard made a different but related mainstream argument: where the safe interest rate remains below economic growth, much larger stocks of debt may be sustainable than traditional fiscal orthodoxy assumed.

Both points are important.

Neither means debt does not matter.

America can create dollars.

But it cannot create confidence.

The loop

The arithmetic is unpleasantly simple.

The US government runs a deficit.

It issues more debt.

Investors demand interest.

If investors begin demanding higher interest because they are worried about inflation, deficits or future supply, the government’s interest bill rises.

That increases the deficit.

The government issues more debt.

Investors demand more interest.

And around we go.

The CBO expects the 2026 deficit alone to be about $1.9 trillion, despite unemployment remaining relatively low. Interest costs rise from 3.3% of GDP now to 4.6% in 2036. By then they are projected to consume almost a fifth of federal spending.

The bond market has begun noticing.

Long-dated Treasury yields recently climbed to levels not seen since 2007. Treasury Secretary Scott Bessent has responded by expanding government bond buybacks, trying to improve market functioning and push down pressure at the long end of the curve. But liability management cannot make the underlying deficit disappear. Reuters reported this week an increasingly open disagreement between Bessent and Federal Reserve chair Kevin Warsh about how much policymakers should interfere with the price of long-term money.

This is where debt stops being accounting and becomes politics.

There are only a few ways out

A heavily indebted state has several basic choices.

It can tax more or spend less.

It can grow rapidly enough that the debt becomes smaller relative to the economy.

It can explicitly default or restructure.

Or it can allow inflation, currency depreciation and sufficiently low real interest rates gradually to reduce the value of what creditors are owed.

Carmen Reinhart and Belen Sbrancia call the last mechanism financial repression.

After the Second World War, governments including the US and Britain held interest rates below market-clearing levels while inflation eroded the real value of enormous wartime debts. Creditors got their dollars and pounds back.

They were simply worth less.

Reinhart and Sbrancia estimate that negative real interest rates and related policies produced very substantial reductions in government debt burdens in the decades after 1945.

One might call that a soft default.

Technically it isn’t.

Politically the distinction can become rather academic to the person whose savings have lost purchasing power.

What would an actual US default do?

Probably not, initially, what people imagine.

A genuine failure to pay interest or principal on Treasuries would not automatically unleash inflation.

Its first effect could be violently deflationary.

Treasuries sit underneath repo markets, bank balance sheets, money-market funds and collateral systems across the world. Question their safety and credit contracts. Equity markets fall. Businesses stop investing. Banks become cautious. Consumers retreat.

Federal Reserve research on previous debt-ceiling confrontations found that even the possibility of default raises Treasury yields and reduces the special safety premium those securities enjoy. An actual default could severely disrupt credit throughout the economy.

Inflation becomes the issue afterwards.

If Washington responds to rising yields by leaning increasingly hard on monetary institutions to keep government financing cheap, then fiscal policy begins dominating monetary policy.

The central bank has an unpleasant choice.

Protect the currency.

Or protect the sovereign bond market.

That is where inflationary debt liquidation becomes conceivable.

Then AI arrives

There is a second debt machine operating alongside the first.

Artificial intelligence.

US companies have issued around $220 billion of AI-related debt in 2026, compared with just $12.5 billion last year, according to Reuters. Amazon, Alphabet and others are borrowing heavily to build the physical infrastructure underlying AI.

And debt understates the commitment.

Microsoft, Meta, Oracle, Amazon and Alphabet have together accumulated around $1.09 trillion in future lease commitments, much of it connected to data centres. Many of those obligations do not yet sit conventionally on their balance sheets because facilities have not begun operating.

This is not American government debt.

But the two markets meet.

There is only so much capital looking for bonds.

The US Treasury wants trillions.

Amazon wants billions.

Alphabet wants billions.

Data-centre developers want billions.

Power companies want billions.

The AI revolution has begun partly as an enormous competition for capital, electricity, land and imagined future revenues.

And suddenly Hyman Minsky becomes useful.

Minsky’s essential insight was that long periods of apparent stability encourage financial structures to become progressively more dependent upon optimistic expectations of future income. Eventually current cash flows no longer justify accumulated obligations.

Stability produces fragility.

AI may generate extraordinary productivity.

It may transform medicine, science, manufacturing and almost everything else.

But financially, we are already spending some of that productivity before it exists.

We have borrowed against the future.

Growth will save us

This is where the public and private stories converge.

Almost everybody’s preferred solution to debt is growth.

Growth is politically miraculous.

If the economy becomes sufficiently larger, nobody has to admit openly who should pay.

No wealth taxes.

No major reductions in military spending.

No serious argument about redistribution.

No cutting pensions.

No telling billionaires they are less rich.

Just make the pie bigger.

The Trump administration’s 2025 reconciliation legislation is an almost perfect illustration of the limits of that argument. CBO does expect its tax and spending provisions to increase economic activity somewhat. Yet after accounting for economic effects and additional interest costs, CBO estimates the legislation still adds around $4.7 trillion to deficits between 2026 and 2035.

Growth helps.

But there is something almost religious about the larger idea.

Capitalism creates ecological stress, inequality, political instability and enormous concentrations of wealth.

The proposed solution is more capitalism.

More technology.

More productivity.

More investment.

More growth.

AI will discover the next productivity frontier.

AI will produce the growth that services the debt incurred building AI.

The future will pay for the future.

It is beautifully circular.

Blow up the state

There is another political current running alongside this.

Parts of the MAGA and libertarian right do not merely want government to spend less.

They want the administrative state itself weakened.

Project 2025 explicitly described its objective as dismantling the administrative state. Elon Musk went further during his period leading DOGE, publicly arguing that entire federal agencies should be deleted rather than merely reduced. Peter Thiel wrote much earlier, and unusually candidly, that he no longer believed freedom and democracy were compatible.

It would be too easy to say these people are deliberately engineering a sovereign debt crisis.

There isn’t evidence for that.

The more interesting relationship is structural.

Reduce taxes.

Increase deficits.

Attack the bureaucracy.

Then point towards the fiscal crisis as evidence that government cannot afford to perform its existing functions.

The debt becomes both problem and argument.

The weakened state is then asked to retreat further precisely as extraordinarily powerful private corporations expand into communications, space, defence, intelligence, currency, infrastructure and AI.

The market does not replace the state.

Certain corporations begin acquiring functions previously associated with it.

The dollar is also a weapon

And now Iran enters the story.

This week Scott Bessent announced Operation Economic Outcast, an intensified campaign designed to sever Iran from financial relationships across the world. Treasury’s threat is not merely that American companies will stop trading with Iran.

It is that companies and countries elsewhere can be punished for doing so.

That power ultimately rests on something extraordinary: access to the dollar system.

The United States can weaponise its currency because large parts of world trade, finance, banking and reserves still run through dollars and American financial institutions.

That is an astonishing imperial asset.

It is also something that can be overused.

Russia and China now conduct almost all of their bilateral trade in yuan and roubles. India has just loosened regulations to encourage rupee-denominated trade. BRICS countries are discussing greater use of local currencies and interconnected payment systems. There is no BRICS currency waiting to replace the dollar – indeed BRICS officials have repeatedly denied that such a project is currently under way – but alternative plumbing is slowly being built.

Sanctions themselves accelerate some diversification. IMF research has found that countries exposed to financial sanctions have tended to increase gold holdings.

But predictions of the dollar’s imminent death are greatly exaggerated.

In the first quarter of 2026, the dollar still accounted for 57.13% of reported global foreign-exchange reserves, and its share actually rose slightly during the quarter. No other currency has anything resembling the combination of market depth, convertibility, institutional infrastructure and available safe assets supplied by the United States.

The paradox is exquisite.

America’s enormous debt helps make the dollar indispensable because the world needs enormous quantities of safe dollar assets.

But if America issues so much debt that confidence in those assets begins to deteriorate, the mechanism supporting its privilege begins undermining itself.

Debt is time

Perhaps this is what debt finally is.

A relationship with the future.

Not simply borrowing money from our grandchildren. That cliché is too crude.

Government borrowing moves resources in the present and creates claims stretching forward through time.

Corporate borrowing does the same.

A bond is an agreement between two imagined futures.

I give you something now because I believe that later you will still exist, your currency will still mean something, your institutions will still function and you will still recognise my claim.

Debt therefore requires something more fundamental than money.

It requires belief in continuity.

And that may be the real American debt problem.

Not $40 trillion.

Not some mystical numerical threshold beyond which the Republic automatically collapses.

The question is whether the political institutions guaranteeing those promises remain more credible than the promises themselves.

The United States is borrowing enormous sums.

Its technology companies are borrowing enormous sums.

It is wagering heavily on AI-generated future growth.

It is using the dollar system aggressively against geopolitical enemies.

Parts of its governing coalition are simultaneously trying to dismantle parts of the state whose credibility ultimately stands behind that system.

And everybody is hoping growth arrives before the bill.

Perhaps it will.

But there is something strange about a civilisation responding to uncertainty by writing ever larger claims upon a future it cannot describe.