Go to work, get paid, eat, buy things, pay the mortgage. This cycle has run for thousands of years. We call it “the economy.”

It has an implicit assumption: people are both producers and consumers.

AI is dismantling that assumption.

The Broken Loop

The operating system of the modern economy is really a closed loop: labor earns income -> income becomes consumption -> consumption creates demand -> demand drives production -> production requires labor. Every link depends on the one before it.

What AI is doing is pulling out the first and last links simultaneously. When machines can replace most human labor, “production requires labor” no longer holds. When labor is no longer a reliable path to income, “labor earns income” collapses too. With both ends severed, consumption and demand in the middle naturally cave in.

This isn’t a problem in any single industry. It’s a structural fracture in the entire loop.

The Fallacy of Composition

Every company that uses AI to replace human labor is being rational. Cut costs, boost efficiency, profits go up, stock price looks good.

But what happens when every company does this at the same time? They eliminate their own customers.

Your employees are someone else’s customers. Someone else’s employees are your customers. When everyone is laying people off, everyone’s customer base is shrinking. Individually rational decisions, summed up, become a collective catastrophe – this is the fallacy of composition.

Can the market self-correct? No. The market optimizes within a framework – price signals guide resource allocation, supply-demand imbalances trigger adjustments. But when the very foundation of the framework is pulled away, the market has nothing to “correct back to.” It can sense that customers are disappearing, but it cannot conjure a new income-distribution mechanism to replace wages out of thin air. This isn’t market failure; it’s the market’s scope being exceeded.

Institutions Can’t Keep Up

Many people will say: governments will step in – UBI, AI taxes, free public services – there’s always a way.

But technology follows an exponential curve; institutions follow a staircase function.

Technology iterates daily. Institutional reform requires consensus, legislation, execution, and error correction – each step carries enormous friction. More critically, institutional change is crisis-driven, not foresight-driven. Nobody votes for UBI while most people still have jobs. By the time UBI is truly needed, the government’s fiscal capacity may already be strained – the tax base is shrinking, income tax and consumption tax revenues are falling, while spending demands are surging.

There’s an even sharper contradiction: the people with the power to drive institutional reform are precisely the beneficiaries of the AI revolution. Tech companies, capital holders, political elites – they lack sufficient incentive to proactively restructure a distribution system that disadvantages them. Every major redistribution in history – the New Deal, the European welfare state – happened only when social pressure became impossible to ignore.

So the question isn’t “can we change?” It’s “can we change in time?”

Four Phases

This won’t happen overnight, but it won’t be slow either.

2025-2027: Displacement Penetration

Already underway. AI doesn’t replace jobs overnight; it first compresses the people-to-output ratio – a team of 10 becomes 6, hiring freezes, natural attrition goes unbackfilled. The first to be hit: content creation, customer service, junior programming, data processing, translation, basic legal and financial analysis – these “cognitive assembly line” jobs.

Hallmarks of this phase: corporate profits rising, employment quality declining, young people finding it harder and harder to get jobs, but the statistics haven’t yet triggered alarms.

And this phase is shorter than most people think. AI Agents will be able to independently complete end-to-end workflows by late 2026 to early 2027 – not a distant vision, but something already taking shape. Once Agents mature, displacement penetration will immediately accelerate into displacement collapse.

2027-2030: Accelerating Collapse

The critical inflection point. Agents are no longer assistive tools but autonomous executors. Companies launch their second wave of AI transformation – not optimizing processes, but eliminating entire departments. Simultaneously, robot costs hit a tipping point, and physical jobs in logistics, manufacturing, and retail begin large-scale replacement.

Unemployment rises rapidly. But what’s more dangerous than the number itself is the structure – displaced workers can’t find new jobs at comparable income, because those new jobs are also being filled by AI. The middle class collapses en masse. Real estate, automotive, education – industries that depend on middle-class purchasing power – feel the chill first.

2030-2036: Crisis and Contestation

A positive feedback spiral takes shape. Consumption drops -> corporate revenue drops -> further layoffs -> consumption drops more. Government finances are under pressure: the tax base is shrinking while social spending demands are exploding.

Pressure for institutional reform reaches a critical point. Social unrest, political polarization, and populist movements force governments worldwide to begin seriously discussing fundamental adjustments. But response speeds vary enormously across countries.

2036-2042: Restructuring

Early-mover nations begin running new models. The core challenge is finding a value-distribution mechanism that doesn’t depend on “labor for income” – public distribution of AI output, ultra-low-cost basic living guarantees, and new economic forms built around uniquely human value.

“End” is not quite the right word. More accurately, it’s entering a new steady state. In this new equilibrium, the basic unit of the economy, the definition of growth, and the content of the social contract will all be fundamentally different from today.

Accelerating Variables

If an energy breakthrough (fusion) materializes, AI deployment costs will plummet further, compressing the entire timeline by 3-5 years. A global financial crisis or geopolitical conflict might slow AI deployment in the short term but would intensify social contradictions. If a small advanced nation (say, a Nordic country) successfully runs a new model first, the demonstration effect would accelerate adoption elsewhere.

What This Means for Individual Investors

If the above analysis is roughly correct, then stock market investment logic needs to change accordingly.

Over the next 3-4 years, profits for AI beneficiaries will surge. The market won’t immediately price in the long-term consequences of demand collapse – markets always chase current-period profits first. During this window, betting on AI infrastructure, compute, and energy – supply-side assets – could deliver very attractive returns.

But once the accelerating collapse phase arrives, the stock market’s foundational assumption – “corporate profits grow indefinitely” – will shake. This is not a “buy and hold for compound returns” era. This is a window with an expiration date. Making money is phase one. Knowing when to stop is phase two. Phase two matters more.

When picking stocks, one dimension is critical: what percentage of a company’s revenue depends on consumer purchasing power? The lower the percentage, the more resilient it is during the collapse phase. And more important than stock selection is building an “exit radar” – continuously monitoring macro signals and getting out or repositioning before the inflection point arrives.

Where to move? When the market peaks and capital flees, there are really only three destinations:

Gold – A hedge against currency-credit risk. When government finances are strained and central banks are forced to print, gold is the parking lot for value with thousands of years of validation. It produces nothing, but during a framework collapse, “not losing” is winning.

AI infrastructure – Compute, chips, cloud platforms, foundation models. This is the bedrock of the new framework. No matter how the old economy collapses, AI’s compute demand will only grow. The key: only touch monopoly-position leaders, not the application layer – application companies’ customers are still people and businesses, and demand collapse will hit them just the same.

Energy infrastructure – Nuclear power, data center electricity, grid upgrades. AI needs power to run. This is one of the few hard-demand assets that doesn’t depend on consumer purchasing power. Not traditional oil and gas – those are tied to the consumption economy and will shrink alongside it.

Three asset classes, three logics: preservation, appreciation, and essential demand. Add cash and short-term government bonds as a liquidity reserve – the collapse phase will produce extreme bargains, and you need ammunition to seize them.

Epilogue

For thousands of years, the engine of the economy has been people – more people, more labor, more consumption, more demand. From agriculture to industry to the information age, technology changed, but this engine never did. AI is making it obsolete.

It’s not that wealth is shrinking – it’s that the pipes for distributing wealth are broken. Production continues, even more efficiently than before. But if all output flows to capital holders while most people lose their entry point to the distribution system, the very word “economy” needs to be redefined.

The time left for each of us to prepare is running short.