A Free Market Is Not a Zero-Sum Game — and AI Won't Break That Rule

When AI went mainstream, the loudest question was the most primal one: "What happens to our jobs?" I shared that fear. It only dissolved after I forced myself to look past the headlines — into macroeconomics and, more usefully, into economic history. What I found is that the fear, however natural, was largely unfounded.

Thomas Sowell put the principle plainly: "A free market is not a zero-sum system, where the gains of one party have to come at the expense of losses to another party." Wealth is created, not merely redistributed. New tools don't just take slices of a fixed pie — they bake bigger pies. History proves it repeatedly, and the early AI data is already saying the same thing.

History doesn't repeat, but it rhymes

Consider two moments most people have forgotten.

Before sewers,an unfortunate class of workers made their living emptying London's human-waste receptacles and selling the contents to farmers. One of the Victorian era's great innovations — sewers with pumping stations — put them out of work entirely. It also transformed public health and the general wellbeing of an entire society. (John Elledge tells this and other forgotten stories in 31 Inventions That Built Our World.) The job vanished; the world got better.

Then came the great churn of the twentieth century.In Basic Economics, Sowell notes that more than 17 million American workers lost their jobs between 1990 and 1995 — yet unemployment never came close to 17 million at any point, and in fact fell to its lowest level in years. People were moving from one job to the next, not clinging to one. As he puts it: "Job security policies save the jobs of existing workers, but at the cost of reducing the flexibility and efficiency of the economy as a whole, thereby inhibiting production of the wealth needed for the creation of new jobs for other workers."

That is the pattern. Destruction and creation happen at the same time — and creation usually wins, provided we let the economy move.

The AI data is already telling the same story

We don't have to reason only from history. The evidence from AI's first act is in, and it looks far more like an investment and hiring boom than a mass-unemployment event.

The Economist estimates that AI has so far been responsible for roughly 1 million new American jobs, against about 200,000 layoffs attributed to AI since mid-2023. And the sources of those jobs are the most revealing part — because they show how a new technology creates work in places no one was watching:

• The picks-and-shovels boom. AI runs on physical infrastructure — chips, servers, data centres, power, cooling. Goldman Sachs estimates spending on that equipment is now about $500bn a year higher than before ChatGPT, with data-centre construction running above $75bn annualised, nearly 60% up on the year. That means electricians, HVAC specialists, grid engineers, construction crews and technicians. The Economist finds roughly 320,000 more jobs in data-centre-linked industries than trends would have predicted, data-centre vacancies have more than doubled in two years, and those installation and maintenance roles pay wages around 40% higher than comparable jobs.

• A brand-new white-collar class. AI engineers, data annotators, AI evaluators, "forward-deployed" engineers, heads of AI — occupations that barely existed a few years ago. Postings for these roles have roughly doubled since 2023–24, adding on the order of 640,000 AI-specific jobs.

• Productivity that expands demand, not just cuts headcount.Make a lawyer twice as productive and the naive conclusion is "we need half as many lawyers." But cheaper legal work invites more of it. Tellingly, the very professions everyone marked for death are growing: US paralegal employment rose about 11% and market-research analysts about 6% between 2023 and 2025 — well above the ~2.5% overall

None of this means AI destroys nothing. Customer-service roles are down roughly 10% and secretarial/admin roles down about 15% since early 2023 — repetitive, predictable work is exactly what AI does first. But set that against the fact that the US labour market normally sees around 1.7 million people leave jobs every month: AI-related cuts, at roughly 16,000 a month, are a rounding error inside that churn.

The crusades that miss the point

This is why fighting the change itself is a poor use of energy. Two crusades illustrate it.

The campaign against data centres. A backlash has grown against the buildings that make AI possible — on grounds of water, electricity and noise. Yet The Economist calls much of it a moral panic fuelled by misinformation: viral "noisy data centre" clips that actually show crypto-mining sites; water figures quoted with no regard for the cooling technology involved. Electricity is a genuine issue — but the honest fix is to price the externality and make operators pay for the grid capacity they demand, not to ban the infrastructure outright. The deeper tell is this: much of the anger at data centres is really displaced anger at AI. As the argument goes — if you think AI is dangerous, argue about AI. Blocking a building full of computers doesn't solve it.

The moral high ground on AI-assisted writing. When Stanley Druckenmiller's AI-assisted op-ed drew outrage, speechwriter Ryan Jacobs (15 years writing for others, including former Treasury Secretary Janet Yellen) pointed out the obvious hypocrisy: we have always accepted enormous human assistance in writing — speechwriters, editors, researchers, assistants — without calling the result fraudulent. His sharpest line skewers what he calls the "hidden Hemingway" fantasy: the elitist belief that a great writer is trapped inside everyone, waiting to emerge from a blank page. For most people, writing is a means to an end, not an artistic vocation. What matters is who owns the ideas, judgment and responsibility — not who physically typed each sentence. AI simply forces us to admit that.

Both crusades spend real energy defending a status quo that is already gone.

The honest caveat — and why it strengthens the case

Intellectual honesty demands one qualifier: The Economist's own framing is "so far." Today's numbers reflect an economy still building and deploying AI. As agents get more capable, the arithmetic could shift, and there's a real risk that entry-level rungs — the jobs through which people traditionally gain skills — get pulled up behind us.

But that caveat is an argument for my point, not against it. If the transition is coming either way, the smart move isn't to barricade the door. It's to shape the transition — to be early, skilled and positioned.

The bottom line

Trying to thwart change is a bet against economic history, and history is undefeated. Every wave that eliminated work also created more of it, often in forms no one could have imagined — twenty years ago, "influencer" sounded absurd.

So stop pouring energy into the fights that yield nothing. Pull your time, attention and resources out of the activities the new order is quietly making obsolete, and redirect them toward the one that's arriving whether we approve or not. In a market that isn't zero-sum, the winners aren't the ones who resisted the change. They're the ones who moved first.