The AI Bubble and Trillions of Debt

I graduated with a degree in Architecture in 1999. Coming back to Hong Kong after the Asian Financial Crisis, there were no real jobs for architects. Hong Kong had its SARS epidemic in 2003, and all the friends I knew who had a mortgage fell into negative equity. I remember a flying friend would recall working at the bank, and a colleague coming to him at his desk to tell him his loan was getting called. It would be years later that I would understand this to be deemed the evils of deflation.

Between 1997 and 2003, flats lost maybe half of their value. Consumer prices fell for sixty-eight months in a row. Salaries were cut, or frozen, or disappeared. The mortgages stayed exactly where they were. A loan of a million dollars remained a million dollars, now owed on a flat worth half a million, by someone earning less than the year before. Hong Kong could not print its way out; the peg held the currency fixed, so the whole adjustment fell on prices, and it fell hardest on the people who had borrowed to buy. By the middle of 2003, roughly a hundred thousand households owed more than their homes were worth.

This is the mechanism the economist Irving Fisher described in 1933, watching America sink into the Depression. When prices fall, debts grow heavier. Borrowers sell to repay, the selling pushes prices down further, and the debts grow heavier still. Every central bank since has been built around the fear of it. It is why governments aim for inflation of about two to three percent a year, why interest rates are cut at the first sign of trouble, and why, in 2008, the world’s central banks printed money on a scale nobody had seen before.

Recently I have been reading Jeff Booth, a Canadian technology entrepreneur whose book The Price of Tomorrow argues that we have this the wrong way round. Technology, he says, makes everything cheaper. A phone replaces a camera, a map, a music player and a bank branch. Left to itself, an economy full of innovation would see prices fall year after year, and that fall would be the natural reward of progress: a salary would buy more each year without a raise, and savings would grow simply by being kept. In his telling, deflation is what abundance looks like.

The trouble is that our money is built on debt, and debt cannot survive falling prices. Hong Kong in 2003 demonstrated that. So central banks fight the deflation that comes with technological growth. They hold rates low and expand credit, and each round takes more new debt to buy the same growth. Booth believes this is a war that can be postponed but never won, and that the money printed to fight it does damage of its own.

On the damage, I find him hard to argue with. New money reaches property and shares long before it reaches wages, so those who already own assets grow richer while those who earn a salary fall behind. The gains of productivity flow to capital. Saving in cash becomes a slow way of losing money, so everyone is pushed into the market just to keep up, speculating whether they want to or not. And the widening gap between the two groups feeds the anger we now see in politics almost everywhere. Anyone who has watched the market these past twenty years has seen all of it.

Where Booth loses me is the cure. He wants money with a fixed supply, which for him means Bitcoin, so that prices can fall freely and the gains of technology reach everyone. But I have lived through what falling prices do to a city built on mortgages. In a world where debt and land are this deeply entrenched, there is no version of this where deflation works. It cannot work for mortgages, where every existing loan becomes a heavier burden for the family or companies and negative equity spreads street by street. It cannot work for equities, priced for a generation of cheap money, and yield on the stock is just as important as perceived prospects. And it cannot work for land, where households, banks and governments keep their wealth, and which no government will knowingly let fall by half. Booth’s answer is that the reset will come anyway. That is a prophecy, and it asks the rest of us to live through 1998 on a global scale.

China may be showing us the collision now. Its factories turn out electric cars, solar panels and batteries so cheaply that producer prices have been falling for years, the very technological deflation Booth describes. Its property sector and local governments, meanwhile, carry enormous debts. Beijing faces exactly the choice he describes: stimulate and add more debt, or let the debts be written down. So far it has chosen, cautiously, the first.

So what becomes of all this debt? History offers only a few endings.

Route	How it works	Who pays	Precedent
Grow out of it	The economy grows faster than the debt	Nobody	Rarely enough on its own
Financial repression	Interest rates held below inflation for decades, so the debt shrinks in real terms	Savers and bondholders, quietly	US and UK 1945–1980: US debt fell from ~120% of GDP to ~30%
Restructure or default	Debts written down openly	Creditors, banks, the family in the flat	Greece 2012; Hong Kong’s negative equity, household by household
Tax the gains	Land value tax, wealth tax	Asset owners	Henry George; Piketty.  - China LAT, still the market bubbled
Upheaval	War, revolution or collapse wipes the slate	Everyone	Walter Scheidel, The Great Leveler

The quiet road of financial repression is the one most governments seem to be on today. It needs no vote and makes no headlines.  But we are living the consequences: first, asset inflation, the Cantillon effect.  Then, forced speculation. Saving cash loses value, so everyone is pushed into risk assets just to keep up.  Finally, social fracture. Widening inequality feeds populism, polarisation and conflict between countries.  All this is observable for anyone that may want to watch the news, and yet no one really has a solution for it.

This past week, every headline and every conference stage has been screaming about two things: an AI apocalypse, and sovereign debt about to blow up. What strikes me most is how differently the generations hear it. The young are superbly excited by the explosion of opportunity AI brings; one person with the right tools can now build what used to take a company. The older folk, who remember 1983 or 2008, see only debt that cannot be sustained. Read through Booth, both are right, and they are describing the same story. AI is the most deflationary technology we have ever built, and it is arriving on top of the largest pile of debt in history. The young are looking at the abundance. The old are looking at the bill that doesn't seem to make sense.

Nothing about it is surprising, and so far there is no version of this that ends well for the average Jane caught in between, with a mortgage that falling prices would crush and savings that printing will quietly shrink.

What I have learned from watching forty years of growth and bust still holds: The only solution that is robust over time is to have low leverage, and assets diversified over geographies, asset classes and liquidity. Debt is what turned a falling market into a called loan at my friend’s desk. Spreading across places and kinds of assets is what lets you wait out a market no one can predict, and liquidity is what lets you buy when everyone else is forced to sell.  The reset then becomes a question, not of if but only when.