• Canada is treating a massive expansion of data-centre capacity as evidence of AI sovereignty, even though hosting compute is not the same as controlling the models, talent, intellectual property, or economic value running through it.

• Data centres fit the same land, debt, construction, and long-duration asset machinery that turned real estate into Canada’s dominant economic strategy. Banks, pension funds, provinces, utilities, and developers can all find something familiar and profitable inside the same project.

• Canada should judge AI investments by the capabilities that remain after the hardware cycle: Canadian talent, applied research, public access to compute, domestic companies, and institutions capable of governing the technology. Buildings should support that strategy, not substitute for it.

Before I understood infrastructure, finance, or industrial strategy, I understood the Quebec Nordiques.

I understood what it felt like to love a team at the bottom of the standings. I understood what it meant to watch a small-market city hold on to hope through bad seasons and draft picks and rebuilds. Then came Eric Lindros.

Lindros did not want to play in Quebec. At the time that felt like one more humiliation. But the trade that followed became one of the most important in hockey history. The Nordiques turned a rejected superstar into the foundation of a contender. Peter Forsberg. Mike Ricci. Ron Hextall. Steve Duchesne. Picks, money, depth. After years near the bottom, Quebec suddenly had the bones of something powerful.

And then the team left.

In 1995 the Nordiques moved to Colorado and became the Avalanche. They won the Stanley Cup in their first season there. For Quebec fans it was not only that the team had gone. It was that the future we had suffered for arrived almost immediately, wearing someone else’s jersey.

Then the final insult. Patrick Roy, the legendary goaltender of the Montreal Canadiens, Quebec’s great rival, was traded to Colorado that same season and helped secure the Cup. The team Quebec lost, strengthened by a legend from its arch-enemy, became a champion somewhere else.

Years later Quebec City tried to build its way back into that future. A new NHL-ready arena went up. Quebecor invested. Politicians supported the dream. Fans believed. The container was there.

But the NHL never returned.

I do not feel about artificial intelligence the way I felt about the Nordiques. Nothing in technology has ever done to me what that team did when I was a kid. Hockey and AI have nothing to do with each other, except that Quebec City taught me what it looks like when a place builds its way toward a future it does not control.

Every bit of it was real: the arena, the millions invested, the civic pride, and the hope. The only thing missing was the league itself.

The machine that knows one trick

Canada may end up domesticating AI by turning it into real estate.

Not because data centres are useless. They are not. But because data centres are the version of AI that Canada’s existing economic machine already knows how to finance: land, construction, permits, power contracts, leases, debt, infrastructure funds, pension allocations, insurance-backed long-duration assets, bank lending.

That is very different from building deep AI capability.

Canada has spent decades getting excellent at turning physical assets into balance-sheet products. Residential real estate alone sat at roughly $8.47 trillion on Canadian household balance sheets in early 2026. Household credit-market debt reached about $3.25 trillion, close to $1.80 of debt for every dollar of disposable income. Residential mortgage debt passed $2.4 trillion by the end of 2025. When people say real estate is central to the Canadian economy, that is not a vibe. It is a structural fact.

Data centres fit beautifully into that structure.

A single data centre can be pitched as a land development project, a construction project, a commercial real estate asset, a pension cash-flow asset, a sovereign AI project, and a national productivity strategy.

That is politically powerful, because almost every incumbent can find their own story inside it. The bank sees lending. The pension fund sees contracted cash flow. The insurer sees duration. The province sees construction jobs and property-tax base. The federal government sees sovereignty. The utility sees demand growth. The developer sees land value. The tech company sees subsidized capacity.

Nobody on that list has to be wrong, cynical, or corrupt. Each one is doing what its balance sheet already knows how to do. That is exactly why this could become Canada’s default AI strategy even if it is not Canada’s best AI strategy.

The pipeline is already committed

This is not a forecast. It is underway, and this month, for the first time, someone measured it.

A working paper by Alexander Carlo and Lyndsey Rolheiser at York University’s Schulich School of Business maps every Canadian data centre across its full lifecycle, from announcement through construction to activation. Their numbers describe a machine that has already made its decision.

The operational base is modest. 194 active facilities, 1.6 gigawatts of capacity, average size 11.3 megawatts. The announced and under-construction pipeline is 22.2 gigawatts, nearly fourteen times the existing base, at an average facility size of 122 megawatts.

The geography tells you what kind of project this is. Active data centres sit on sites averaging about 13 acres, roughly 19 kilometres from a downtown. Announced facilities average nearly 2,000 acres, 134 kilometres from the nearest major city.

Thirteen acres near a downtown is digital infrastructure. Two thousand acres, two hours out, is land development.

One more number for scale. The researchers calculate that a single average announced facility, at realistic utilization, would draw the electricity of somewhere between 49,000 and 78,000 Canadian households.

Where the machine runs hottest

The federal Sovereign AI Compute Strategy is the visible layer: up to $700 million for domestic compute, up to $1 billion for public supercomputing, up to $300 million for an AI Compute Access Fund. Call it roughly $2 billion.

The real work is provincial and private. Alberta accounts for 92 per cent of Canada’s planned data-centre capacity while hosting about 3 per cent of what is active today. The province is targeting $100 billion in private investment over five years, supported by a concierge program to streamline approvals, municipal property-tax deferrals of up to 15 years, and a deregulated power market that lets developers bring their own natural gas generation instead of waiting in the grid queue.

Ottawa’s program is the press release. Alberta is the machine, already running.

Now hold two numbers next to each other. Canada’s entire sovereign compute strategy is roughly $2 billion. Hyperscaler capital spending is projected at about $725 billion in a single year. Canada’s public play is roughly 0.3 per cent of one year of incumbent spending.

That does not mean Canada should stay out of AI infrastructure. It means Canada cannot afford to confuse participation with strategy. If the scale game is unwinnable, the only serious question left is what strategic capability $2 billion can actually buy. Leverage is in the layers that survive a hardware cycle. Talent that stays. Research and applied capability that compound. Institutions that understand the technology well enough to govern it. More buildings is the most familiar answer available to Canadian finance. It is also the answer that buys the least of any of those things.

The box and the thing inside it

Here is the problem with financing AI like real estate. A data centre is not one asset. It is layers of assets with very different lifespans.

Engineering assessments put the building shell at 50 years or more. The internal installations, power and cooling, last around 20. The servers themselves may last as little as five.

The financial system prices the asset off the 50-year layer. The strategic value lives in the five-year layer.

That mismatch is not a detail. It is the whole problem. Banks, insurers, and pension funds underwrite data centres with the logic of long-duration infrastructure, the logic of a toll road or an office tower. But the layer that makes the asset strategically valuable ages faster than the debt against it. Even the incumbents cannot agree on how fast. Hyperscalers have been extending server depreciation schedules in their own accounting while skeptics argue AI hardware goes obsolete faster than the books admit. The market is openly uncertain about the lifespan of the valuable layer. Canadian institutional capital is underwriting it as though it lasts forever.

Have you noticed how the bet gets sold? Cold climate, cheap power, land, water, political stability. Building data centres sounds like the conservative play, the one that leans on our traditional strengths. But it is only conservative if the current AI paradigm holds. It is exposed to chip architecture changes, cooling changes, model-efficiency gains, inference moving closer to users, export controls, vendor lock-in, and the open question of whether enormous training runs stay the centre of gravity at all.

A bet that is conservative in its inputs and speculative in its assumptions is not a conservative bet.

Sort the risks by which layer they attack and the picture organizes itself. The shell faces grid bottlenecks and siting fights. The installations face expensive retrofits when power density and cooling architecture shift. The servers face obsolescence, export controls, and paradigm risk. Above all of it sits the political risk: a country mistaking hosting compute on Canadian soil for AI sovereignty.

The landlord and the tenant

Say the buildings get built and filled. Who owns the game inside?

The Schulich paper answers this directly, and the answer sharpens as the facilities get bigger. Among pipeline projects of 100 megawatts and up, US-headquartered firms account for 85 per cent of identified providers, 100 per cent of financial backers, and 100 per cent of end users. The larger the facility, the more foreign the ownership. The authors note that local populations carry the costs of expansion, the electricity demand, the emissions, the water risk, while returns on the underlying assets may accrue outside the province or outside Canada entirely.

That is landlord economics. Canada finances the shell, hosts the power draw, and collects the rent. The tenant keeps the capability, the intellectual property, the pricing power, and the margins, and books the profits wherever suits it. Canada collects property tax on the building and payment for the electricity. The value created by the intelligence inside is taxed somewhere else.

The employment story is just as lopsided, and it is documented. Hyperscale facilities run on skeleton crews: security, facilities staff, technicians. US research cited in the same paper found that tax-incentive-driven data-centre construction changed where facilities got built without producing local tech employment growth. The construction jobs are real, and temporary. That is the employment profile of a real estate project, not an industry.

So the opportunity cost has a shape. Canada would be spending scarce capital, grid capacity, and political attention on the lowest-jobs-per-dollar form of AI participation, while the high-employment layer, the labs and the applied firms and the people, keeps leaking south.

Whose money is on the table

Now the part that involves you whether you follow AI or not.

Canadian institutional capital is already in the trade. CPP Investments committed C$225 million in construction financing for a hyperscale expansion in Cambridge, Ontario. It committed up to roughly C$1 billion alongside data-centre operator CtrlS in India, and joined a multi-billion-dollar data-centre platform in Australia. La Caisse provided $240 million in senior debt for an AI-ready facility in Montréal.

The public-market exposure is bigger, and stranger. Across the observable US equity filings of Canada’s eight largest pension funds, the Maple 8, holdings in AI infrastructure firms grew from about $12.6 billion just before ChatGPT launched to about $64.5 billion by early 2026. A fivefold increase in under four years, concentrated not in data-centre operators but in the cloud platforms and chipmakers sitting above them. For several funds, AI infrastructure now represents fifteen to twenty per cent of their disclosed US equity portfolios.

Put the two facts side by side. Canadian retirement savings are increasingly long the AI boom through shares of the tenants. The physical facilities on Canadian soil are majority foreign-owned, and grow more foreign as they grow larger. Canadians hold slivers of the tenants and almost none of the buildings. Exposed at both ends, in control of neither.

No single one of these positions is reckless. Each is small against the size of the fund holding it. The concern is not any one bet. It is that banks, insurers, pension funds, and governments are converging on the same asset class at the same moment, pricing it with the same long-duration logic, exposed to the same short-lived layer. Diversification across institutions does not help when the institutions are all making the same bet.

Canada has seen this pattern before. No single mortgage was the problem either.

So why does nobody say no?

In a diversified economy, sectors fight. Manufacturing wants cheap land and energy. Tech wants talent and risk capital. Exporters want competitiveness. Banks want credit growth. That friction is messy, but it is also an early-warning system. When a strategy tilts too far toward one interest, some other interest complains loudly and pays lobbyists to keep complaining.

Canada gets described as more socialist than the United States because of universal healthcare. Then I look at housing, and we appear remarkably comfortable treating one of life’s basic necessities as an investment vehicle. That made me wonder whether housing is the exception or the lens.

The more I read of Canada’s AI strategy, the more familiar it sounded. We weren’t really talking about AI. We were talking about land, electricity, construction, financing, leases. Once you notice the pattern you cannot stop seeing it.

The problem is not that Canada’s institutions are identical. It is that too many of the powerful ones have learned to win the same way, through land appreciation, construction, credit creation, and long-duration assets. When the opportunity is phrased as turn AI into land, power, leases, and infrastructure, all of them can read it without translation. The conclusion arrives pre-approved.

The people on the other side of the arrangement exist. Renters. Young workers. Researchers who leave. Manufacturers squeezed on power costs. But they have op-eds, not balance sheets. The friction is real. It just has no institutional voice.

Canada ran this experiment with housing. The warnings were published for twenty years. Look how much they changed.

Policy Horizons already wrote the warning

There is a stranger detail in all this. One arm of the federal government has already described where the road can lead.

Policy Horizons Canada, the federal foresight agency, published a scenario report on the future of social mobility. It describes a plausible 2040 in which most Canadians are locked into the socioeconomic conditions of their birth. Property ownership divides society more than income does. Inheritance becomes the main path to security. The value of human labour shrinks under AI. And people who want to climb emigrate to places where climbing still feels possible.

Reading it, I could not tell whether the report was describing the future or documenting something already underway.

Since 2000, Canadian home prices have risen more than four times faster than household incomes.

If wealth increasingly depends on owning assets rather than earning income, mobility gets harder with each generation. So how speculative is a scenario that says social mobility may decline by 2040?

Now set that scenario beside the strategy. A data-centre-led AI policy produces returns that flow to property and existing capital. It creates few permanent jobs. Its profits are booked mostly elsewhere. Every mechanism runs in the direction the foresight report warns about.

This is not hypocrisy and it is not a conspiracy. It is non-communication. The foresight unit is a smoke detector that was never wired to anything. The warning exists, in the government’s own hand. The machine cannot hear it, because every institution that opens the file arrives at the same familiar answer.

The arena, again

Most criticism of AI data centres focuses on what they consume. Water, electricity, land, quiet. Those concerns are legitimate. But there is another question Canada has barely started asking.

What if the problem is not only what data centres consume, but what they let us avoid building?

If Canada spends its limited sovereign AI capital on the physical layer because that is the layer our financial system understands, we may end up with useful buildings, long leases, and impressive announcements, and none of the capability that makes a country sovereign in anything.

Quebec City’s arena was not useless. It hosts concerts and junior hockey. The construction was real. The civic pride was real. The hope was real.

But the league was somewhere else, and the league decided.

A city can build the perfect container for a future it does not control. So can a country.

• During COVID, governments replaced lost household income because the shutdown was visible and everyone agreed on what had happened. AI removes wages while offices stay open and output keeps rising, so nothing in the surface data triggers a response.

• Corporate models assume productivity gains return to the economy through lower prices, but pass-through only works if consumers still have wages to spend. The same displacement also erodes the payroll base that funds the institutions expected to manage the transition.

• The gap between displaced payroll and preserved purchasing power belongs to no one. Companies own margins, investors own returns, governments own budgets, and workers own the consequences. A gap nobody owns gets discovered late, in defaults and weak consumption rather than in policy.

During COVID we deliberately shut down large parts of the physical economy. Then governments kept demand alive by putting income directly into households and businesses. It was expensive and improvised and it mostly worked, because everyone could see what had happened. The storm was visible. The response was obvious.

AI runs the same problem in reverse. Offices stay open. Production continues. Quarterly numbers look fine. Nothing on the surface says anything is wrong. But underneath, wages can start leaving the system structurally, and no one calls it an emergency because nothing looks like one.

Rather than a sudden collapse, we are witnessing a slow drain on everyday purchasing power. The real danger of job loss isn't just vanished income. It is the systemic disappearance of the consumer base itself.

Where demand comes from

The economy runs on a loop that is almost embarrassingly simple. People earn income. They spend it. That spending becomes revenue for businesses, which pay wages again.

Think of it as a river-powered mill. The river is productive energy moving through the economy. The mill is the labour system that turns that energy into wages and purchasing power. The flour is what the rest of the economy eats.

For most of modern economic history, those flows stayed connected. Businesses became more productive, workers earned more, spent more, and supported further growth. When that balance holds, production and purchasing power reinforce each other. When it weakens, the damage rarely stays contained. It travels through hiring, investment, spending, and confidence.

AI changes the flow

Production can keep rising while less productive energy passes through the labour system. More of the river bypasses the mill entirely.

We have seen a version of this before.

Through the 1920s, American farmers adopted tractors, combines, and large-scale production methods. Agricultural productivity surged and surpluses piled up. But productive capacity grew faster than purchasing power. Crop prices collapsed, farmers could not service their debts, rural banks failed, and the stress moved outward into the financial system.

The problem was never a shortage of production. It was a weakening relationship between production and broad participation.

AI raises the possibility of the same imbalance across the knowledge economy, in a lot of sectors at once.

Most corporate models assume productivity gains eventually return to the economy through lower prices, higher consumption, or new demand. Economists call this pass-through. If AI makes a service dramatically cheaper, consumers should buy more of it or redirect the savings elsewhere.

There is a blind spot in that logic. Price reductions only stimulate demand if consumers still have purchasing power. If the savings on the corporate spreadsheet come from eliminating the customer’s income, the mechanism starts working against itself. Cheaper products matter very little to someone without wages to spend.

Each decision to automate stays locally rational. Collectively, the system starts behaving differently.

The river accelerates. Less of it reaches the mill.

The question markets have to answer

The obvious counterargument is that we have been here before and it turned out fine. Mechanisation, industrialisation, computers. Technology lowers prices, raises productivity, and creates work nobody could have described in advance.

Historically, that has largely been true.

But when agriculture mechanised, the shift from most people farming to almost nobody farming took about a century. Entire generations had time to adjust. Older farmers finished their careers while their children went into factories and offices. The economy changed, and human adaptation moved alongside it.

Software scales globally in months. Accounting, customer support, software development, legal analysis, marketing, and administration may all start changing at the same time.

The risk is not that new work never appears. It is that the retraining cycle may not keep pace with the deployment cycle. A 45-year-old analyst cannot step out of economic participation for a generation while the labour market reorganises. Housing, food, and consumption do not pause.

So pass-through depends on something deeper than lower prices. It depends on consumers still participating in the loop.

The issue is not whether AI expands productive capacity. It almost certainly will. The issue is whether purchasing power keeps circulating widely enough to absorb what gets produced.

The missing economic infrastructure

Modern economies were built on the assumption that productive activity and human payrolls stay connected. Governments fund social infrastructure through systems tied to wages: income taxes, payroll taxes, pension contributions, employment insurance, and the consumer spending of employed workers.

That works as long as the river keeps passing through the mill.

AI introduces a payroll logic failure. If productive capacity keeps growing while payroll participation weakens, the system starves the institutions expected to stabilise the transition. Displacement increases the demand for support and reduces the revenue that funds it at the same time.

This is where a different kind of mechanism becomes necessary. I call it the Agentic Economic Contribution, or AEC.

The AEC is not a robot tax and it is not a penalty for using AI. It is a transition mechanism tied to the economic effects of payroll displacement. When AI-driven productivity rises while human payroll falls, part of the displaced flow gets redirected back into the infrastructure that keeps demand alive.

To make this concrete, I built a simple demand-impact model. It does not try to predict the future of work. It isolates one mechanism: what happens to demand when payroll is displaced faster than purchasing power is replaced.

The model compares two scenarios. In the first, displaced payroll simply leaves the demand cycle. In the second, a modest AEC recirculates part of it.

The AEC does not solve the problem. Even in the model, most of the demand loss remains. What it does is slow the bleed while leaving most of the company’s incentive to adopt AI intact. The spreadsheet is available here for anyone who wants to change the assumptions, including displacement rate, spending rate, pass-through, contribution tiers, and multipliers.

The trigger would be the Labour Displacement Ratio, or LDR: the share of a company’s baseline payroll displaced by AI-attributed systems.

There is an obvious objection. How would anyone know whether payroll was displaced by AI rather than by a downturn, a restructuring, or a change in strategy?

That is what disclosure is for. In an earlier piece, Let AI Shovel the Snow, I argued that AI-driven displacement should begin with a standardised corporate disclosure form covering the function affected, the roles reduced, the system involved, and the estimated payroll value displaced. The point is not to track every software licence. It is to create an auditable record when AI materially reduces human roles, which becomes the empirical basis for the LDR.

A business saving a few hours a week stays below the threshold. Large institutional deployments become visible.

A company that displaces 3 percent of its payroll is not doing the same thing as one that displaces 40 percent. The first is having a productive year. The second is restructuring its relationship with the broader economy.

Rather than a flat penalty, the AEC would work like a set of sluice gates. As displacement rises, contribution rates open gradually. Small displacement triggers little or nothing. Larger displacement triggers more, but only on the portion above each threshold.

The logic is the same as marginal income tax. Crossing into a higher bracket does not tax everything at the top rate. Only the next layer is treated differently.

So a company displacing 12 percent of payroll faces something modest. At 40 percent it becomes meaningful, and it still keeps most of the savings. The framework has no cliff.

There is a jurisdictional problem too. If the mechanism applied only where AI servers physically sit, companies would move infrastructure into low-tax jurisdictions and keep selling into large consumer markets elsewhere.

That is why the AEC needs a dual nexus. The first is the activity nexus: where the automated productive activity occurs or is operationally controlled. The second is the consumer nexus: where the customers and economic beneficiaries are located. A company using AI infrastructure in one country to replace payroll in another while selling into a third would have obligations that follow the economic flow rather than the server rack.

The purpose is not perfect precision. No tax system has that. The purpose is to close the obvious loophole: automating globally while routing the economic activity through whichever jurisdiction asks for the least.

None of this requires treating AI as a person or as a corporation. It uses tools governments already run: marginal rates, payroll calculations, nexus rules. It just recognises that agentic systems now perform economically meaningful work, and that when that work displaces payroll at scale, the missing flow has to be accounted for somewhere.

Why companies would still adopt AI

The standard objection is that any contribution framework discourages innovation. That concern is fair. Designed badly, this becomes a drag on productivity instead of a stabiliser.

But the AEC is built around marginal displacement. Even at significant displacement levels, companies retain most of the financial benefit of adopting AI. The model shows a company displacing 40 percent of payroll still keeping the majority of its savings. The framework does not erase the incentive to automate. It prices part of the transition cost back into the decision.

It might also change how automation gets evaluated internally. Without a contribution mechanism, the spreadsheet is simple. If AI can do the job cheaper, replacement wins. With a rising marginal contribution attached to large-scale displacement, full replacement starts competing against hybrid models where workers use AI to become far more productive.

In roles built on judgment, trust, accountability, or client relationships, that combination may already be the better option. The best use of AI is not always replacing people. Sometimes it is expanding what they can do.

A contribution framework would not force that choice. It would make it more honest. Replace at scale and you contribute to the transition costs you created. Use AI to keep people economically participating and you contribute less.

There is a reputational dimension as well. Consumers are already starting to distinguish between companies that use AI to extend human work and companies that use it to remove human contribution. The backlash against AI-generated art in games and media is an early version of this. People are not only asking whether AI was used. They are asking whether anyone was left in the process.

Environmental policy followed a similar arc. It started as compliance cost and became brand value, investor confidence, and consumer trust. AI may follow the same path, where firms get judged not on whether they use it but on how much participation they preserved.

The missing half of AI infrastructure

The imbalance is already visible in how governments are preparing.

Canada is funding expanded access to AI compute. The United Kingdom has invested heavily in national compute capacity through Isambard-AI and the AI Research Resource. France is positioning itself as a European infrastructure hub, with money flowing into sovereign compute, data centres, and firms like Mistral. The United States piloted the National AI Research Resource to widen researcher access to compute and data.

That investment may well be necessary. If AI becomes a major productive layer, it will need enormous physical infrastructure.

But almost all of the institutional energy is pointed at productive capacity. Very little is pointed at how societies maintain purchasing power, tax capacity, and participation if displacement accelerates.

We are building the rivers before we build the sluice gates.

This is not because companies are villains or governments are asleep. It is because every incentive currently points the same way. Companies are rewarded for efficiency. Investors reward margin expansion. Consultants are paid to help firms capture productivity gains. Governments want investment, data centres, energy projects, and technological standing.

Each of those makes sense on its own. Together they produce a one-sided tug of war. Nearly everyone is pulling toward more capacity and more speed. Far fewer institutions are pulling with equal force toward the systems needed if that deployment weakens the wage base demand depends on.

That is the missing half of AI infrastructure.

The visible half is easy to count. Servers, chips, power lines, announcements. The invisible half is harder: income continuity, retraining capacity, payroll replacement, demand stabilisation, and jurisdictional rules that stop displaced economic flow from vanishing into the least accountable channel.

Counting is where most of the conversation currently stops. A recent Atlantic cover story asked how soon AI would take American jobs, and the pattern in it was familiar. Economists, policymakers, labour leaders, and executives all agreed the risk was serious enough to discuss. The responses stayed mostly in the realm of measurement, retraining, wage insurance, shorter workweeks, or broad political aspiration.

Those conversations matter. But counting is not a response. It is the beginning of one.

No one owns the demand gap.

Companies own their margins. Investors own their returns. Governments own their budgets. Workers own the consequences. The space between displaced payroll and preserved purchasing power does not clearly belong to anyone, which is why it stays easy to ignore until it shows up in unemployment data, weaker consumption, defaults, or political anger.

The purpose of the AEC is to make that gap visible while it is still cheap to look at.

The real choice

This framework does not define the end state. It does not prescribe how governments should redistribute what gets captured, and it does not assume one model works across every jurisdiction and sector.

What it does is create the mechanism that makes informed choices possible.

Some governments may direct AEC flows toward retraining and workforce transition. Others may build public AI infrastructure that serves citizens directly. Others may choose income support or community stabilisation, or some combination that changes as the data improves.

The AEC does not dictate the answer. It creates the captured flow that lets answers emerge.

That is the minimum viable policy. Not a finished system. A starting point that generates the data, the revenue, and the institutional capacity to adapt as the transition unfolds.

Some economists argue it is too early to build systems around AI-driven displacement. They may turn out to be right.

But societies almost never build stabilising infrastructure after certainty arrives.

The question is not whether AI transforms the economy. That is already underway. The question is whether we build the gates before the pressure does the deciding for us.

The Instinct to Adapt

Early in my career, a VP described how our actuaries approached problems.

They were brilliant at what they did. But they had spent their entire careers refining a very specific way of solving problems.

When conditions changed, their instinct wasn’t to redesign the solution. It was to adapt what they knew.

He put it simply: if you handed them a pair of jeans, they wouldn’t rethink the product. They would cut it into smaller pieces—because they knew how to make socks.

I’ve been thinking about that story since OpenAI published its economic policy paper.

Their proposal is ambitious. A public wealth fund. Shorter workweeks. Adaptive safety nets. Taxes on automated labor. The instinct behind it is right : AI is going to reshape who creates value and who benefits from it.

But the solution they’re reaching for is a familiar one.

More cash. Better distributed.

Before we ask whether that’s the right answer, it’s worth asking what cash was designed to solve in the first place.

A Note on Who This Is Written For

A lot of these ideas, including OpenAI’s, are explained upward. To governments, investors, and executives. The language is fluent in capital markets, policy frameworks, and shareholder value.

But the people most affected by these changes aren’t in those rooms.

If a proposal can’t be explained in a way that makes sense to them, it’s probably missing something.

This one tries to do that. It starts where money started.

Cash Solved a Simple Problem: Misaligned Needs

Barter failed because it required something economists call a double coincidence of wants.

You have fish. I have wheat. But I don’t need fish today. The exchange doesn’t happen — not because value doesn’t exist, but because coordination fails.

Coins solved that. A shared medium everyone agreed had value meant exchange could happen across time, across distance, across mismatched needs. The fisherman sells today, holds the coin, buys wheat next month.

The coordination problem is solved.

Paper money and digital cash extended the same logic. More portable. More divisible. More abstract. But still solving the same fundamental problem: how do you coordinate exchange between people with different skills, different needs, and different timing.

Every version of this system carries one assumption:

That value is created by human participation.

Someone caught the fish. Someone grew the wheat. The exchange medium exists to coordinate between human producers and human consumers.

AI Begins to Alter That Assumption

Not everywhere. Not immediately. But directionally and at scale.

When AI systems draft the documents, process claims, staff the support lines, and generate the code — the coordination problem between human producers starts to look different. You are no longer primarily exchanging value between people with mismatched skills. You are distributing output from a system that has no needs of its own.

Cash was designed to solve a human coordination problem.

It may be only a partial answer to a post-human-production problem.

OpenAI’s public wealth fund takes the existing fabric and cuts it into smaller pieces. It redistributes the output of a changing system without asking whether the system of exchange itself needs to change.

That’s not a criticism of the people proposing it. It’s a description of how smart people respond when the conditions change faster than the mental models do.

They make socks.

The Part Nobody Plans For

This doesn’t mean cash disappears. It won’t — not for decades, and perhaps never entirely.

What’s changing is its domain.

We didn’t replace barter with coins overnight. There was a long period where both existed simultaneously. Trust in the new medium had to be built. Old coordination mechanisms still worked for some exchanges while new ones emerged for others.

The transition was the hard part. It always is. And it’s also the part nobody plans for.

Most economic proposals describe the destination — what the system should look like once the shift is complete. What they skip is the journey. Which is exactly where the policy failures happen.

The challenge isn’t choosing one system over the other. It’s managing the transition between them.

That transition is hard to design in advance. Most economic proposals try to define the end state. But transitions don’t work that way.

Startups don’t ship the final product. They ship a minimum viable product and iterate based on real-world feedback.

Economic transitions require the same approach.

Not a fully designed end state — but a minimum viable policy that can evolve as the system changes.

The disclosure framework proposed from my first article makes that possible. It turns economic change into feedback, allowing policy to adapt as new patterns emerge.

But the disclosure form does something more than generate fiscal data. It tells governments not just how much to redistribute — but what form that redistribution should take.

The Honest Middle Road

Replacing cash entirely is likely one of the last steps in this transition — if it happens at all. What’s more realistic, and more useful to plan for, is a gradual boundary shift.

Cash UBI covers what remains genuinely market-mediated. The things where price signals still serve a useful coordination function, where individual choice still matters, where human exchange still drives value.

Public AI infrastructure covers something different. The sectors where automation has driven marginal cost low enough that direct provision beats cash transfer. Where the market mechanism adds friction without adding value.

We already have precedents for this boundary. Public libraries. Free public education. Municipal water. Interstate highways. These are goods societies decided shouldn’t be fully mediated through cash exchange — not because cash disappeared, but because universal access to certain things was deemed more important than market efficiency in distributing them.

AI doesn’t create that idea. It massively expands the category of things that could work that way.

The disclosure form is the instrument that identifies which future categories meet that threshold. Not ideologically. Empirically. When farming, food transportation, and distribution jobs are displaced at scale, the data seeds the case for a public AI food access program — direct provision rather than cash transfer.

This Isn’t a Thought Experiment

It’s already happening.

New York City just announced its first municipally owned grocery store. The goal is one in each of the five boroughs. The argument is simple: food access is too important to leave entirely to market pricing.

This is the most expensive version of the model.

Human staff. Physical infrastructure. Tens of millions per location. No AI-driven supply chain. No automated logistics. No marginal cost approaching zero.

Critics point out the obvious challenges: cost, efficiency, execution. A publicly run grocery system built on today’s economics risks being expensive, difficult to scale, and vulnerable to the same inefficiencies markets are designed to avoid.

All of that is true.

And it still makes the case.

Because the constraint here isn’t the idea. It’s the cost structure.

Now imagine the same model in a system where AI has displaced a significant share of farming, transportation, and distribution work.

The disclosure data shows governments when that threshold is crossed—sector by sector, region by region.

At that point, the economics change.

The overhead that makes this model difficult today begins to fall. Labor, logistics, coordination—areas where inefficiencies compound—become increasingly automatable.

What looks expensive and impractical now becomes something else: viable infrastructure.

Mamdani’s store is a proof of direction.

AI changes the cost of following it.

The most expensive version of this model is already being built. The question is what remains when the cost is non longer the constraint.

What We’re Actually Building

The jeans-into-socks instinct is understandable. These are smart people working with the best tools they know. But the underlying conditions are changing. The coordination problem that cash was invented to solve is shifting shape.

The honest answer isn’t to throw away the fabric. It’s to ask, clearly and without ideological commitment, which parts of the economy still need cash to function — and which parts might be better served by something else.

That question can’t be answered from theory alone. It requires data. Real displacement numbers, by sector, by jurisdiction, over time. A minimum viable policy that generates feedback rather than prescribing outcomes.

The disclosure framework is that instrument. It doesn’t tell us where we’re going. It tells us where we are — and how fast we’re moving.

That’s enough to start.

The questions this raises — how to measure, how to coordinate globally, how to govern the boundary between cash and direct provision — each deserve their own treatment. This is the second of several.

The Question Nobody Is Asking

For years, the technology industry has been consumed by a single question: how do we put guardrails on AI?

A reasonable question.

Meanwhile, no one has asked whether the economy those systems operate in has any guardrails at all.

Companies are racing toward full automation with the logic of an arms race, not because it’s collectively wise, but because no competitor can afford to stop first. The first movers capture extraordinary returns. Stock prices soar. Everyone follows.

And somewhere in that sequence, quietly, the customers begin to disappear.

An AI system that replaces 1,000 workers doesn’t just reduce costs. It removes 1,000 incomes from the economy. Income that would have been spent, circulated, and taxed.

This isn’t a new problem.

A century ago, Henry Ford understood that mass production only works if workers earn enough to buy what they produce. The system worked because labor and consumption were tightly linked.

Automation breaks that link.

You’re building the machine that eliminates your own demand. And you can’t stop, because your competitor won’t.

My First Five Bucks

When I was a kid, my dad used to tell me it would help if I shoveled the snow once in a while.

I was always too busy. Homework. Going out with friends. Something else always came up.

Then one day, he offered me five bucks.

I remember it clearly. Suddenly, the work hadn’t changed but the equation had.

I did the job.
I got paid.

And that money didn’t just sit there. I spent it. Probably on something useless. But it went back into the world.

That’s how the system works:

Work creates income.
Income creates spending.
Spending sustains everything else.

Now imagine the same driveway. The snow still needs to be cleared. The work still gets done.

But no one gets paid:

No five dollars.
No spending.
No participation.

The job exists. The output exists.
The snow still gets cleared.
But the economic loop is broken.

The Worker Who Arrived Without a Paycheck

So who’s new in this equation?

Someone is doing the work. Drafting the emails, processing the claims, writing the code, staffing the support lines. The output is real. The productivity is real. The profits are real.

But there’s no paycheck. No rent paid. No groceries bought. No tax withheld.

The worker arrived. The worker’s economic participation didn’t.

This Doesn’t Slow the Race

Some policymakers have started to respond to the speed of AI by proposing to pause the construction of new data centers.

Slow it down.
Build guardrails first.
Give society time to catch up.

It’s an understandable reaction.

When a system moves faster than our ability to govern it, the instinct is to stop the system.

But the race isn’t something any one country or company can pause.

The incentives are global. The competition is structural.
If one actor slows down, another accelerates.

The question is not whether the race continues. It will.
The question is whether it can sustain itself.

This framework does not ask companies to slow down. It accepts the reality of the race and builds the infrastructure to support it.

An economy that cannot maintain demand will not sustain innovation for long.

This is not a constraint on progress.
It is a condition for its continuity.

We know how to do this. We have done it before.

Social Security wasn’t built after the crisis.
Seatbelts weren’t mandated after the fatalities.

We build infrastructure before the system breaks.

The same logic applies here. We figured out how to measure emissions across the global economy. It wasn’t perfect. It didn’t start perfect. But we built it anyway. We can do the same for AI.

Why We Created Corporations

Economic systems evolve when the units they’re built on no longer fit reality.

At one point, individuals were enough. But as commerce scaled, something broke.

Projects grew larger. Capital had to be pooled. Risk had to be contained. Coordination extended beyond any single person.

So we created a new unit of economic activity:

The corporation.

It extended economic participation beyond what individuals alone could support. Corporations can generate income, own assets, enter contracts, and crucially, be taxed. That structure didn’t slow growth. It made it possible.

Corporation vs. Contractor

A corporation is a full legal entity. It owns assets, takes on liabilities, can sue and be sued.

That’s more infrastructure than we need — and it opens doors we don’t want to open. Questions of AI rights. AI ownership. Legal personhood.

A contractor is simpler.

A contractor performs work, generates income from that work, and that income carries fiscal obligations in the jurisdiction where the work is performed.

No personhood required.
No rights implied.
Just economic activity with a ledger attached.

That is much closer to what’s needed here.

The Balance Sheet Implication

If AI systems are treated like contractors for tax purposes, their activity becomes measurable as a notional revenue stream.

Not because the AI owns anything. But because the work it performs has value. That value is measurable. The wage equivalent of the human labor it replaces. Grounded in existing wage data. Auditable using systems that already exist.

The Income Statement Writes Itself

Revenue: the imputed labor value of tasks performed, by jurisdiction.
Remittance: a portion of that value, returned to the jurisdiction’s public infrastructure.

No new accounting paradigm. Just a new layer of attribution.

The Line That Connects Everything

We already know how to handle this:

• When a contractor works in your jurisdiction, they remit.

• When a corporation operates in your market, it remits.

AI systems are closer to contractors than corporations. And we already have the tools to handle both.

Before We Can Fix It, We Need to See It

Before we can redesign the system, we need to see it clearly.

Today, governments have detailed visibility into employment but almost none into displacement. When a role disappears because of automation, it’s recorded as a layoff, a restructuring, or simply absorbed into productivity gains.

The cause is lost. The snow gets shoveled, but no one gets paid.

A simple first step would change that.

Require companies to file a standardized disclosure when AI systems replace or materially reduce human roles. Not as a penalty. Not as a restriction. As infrastructure.

A form.

Access the full form here.

• What function was replaced.

• How many roles were affected.

• What system now performs the work.

• What level of income was displaced.

• Where the economic activity continues.

We already require disclosure for far less transformative events.

This form is not just transparency. It is the empirical foundation for everything that follows.

Without it, debates about remittance rates, attribution methodology, and jurisdictional thresholds are speculation. With it, they become engineering problems : solvable, adjustable, and grounded in observed economic reality.

The IRS, the CRA, and equivalent agencies in every jurisdiction already audit payroll remittances. This framework doesn’t require a new enforcement body. It requires a new line item for existing ones.

If AI is becoming a core driver of economic output, then tracking its impact on labor is foundational.

This can be implemented within the next year.

A prototype of such a disclosure form is included here.

Where the Work Happens

Take a simple example.

An AI customer service agent answers calls and chats for Amazon customers.

Where is that work happening?
In Canada? In the United States? In Brazil?

The answer isn’t just one thing. It’s two.

Where the work is performed: the jurisdiction where the AI system is actively processing, deciding, generating output.

And where the customer is located: the jurisdiction where the economic benefit is received.

Both matter. Both generate obligations.

Where both are present, both jurisdictions have a claim, apportioned accordingly. Where only one is present, that jurisdiction holds the obligation alone.

This dual nexus closes the obvious avoidance gap. A company cannot declare its AI infrastructure resident in a low-tax jurisdiction and route all obligations there. The Canadian customer interaction generates a Canadian obligation, regardless of where the server is.

Not where the server is. Not where the company is incorporated. Not where the system was built or trained.

Where the work is delivered. Where the value is created. Where a human worker would have been.

That is where economic activity occurs. That is where it should be recognized.

In 2017, Canadian tax lawyer H. Michael Dolson proposed exactly this mechanism in response to Bill Gates’ proposal to tax the robots. Dolson concluded it would be extraordinarily difficult to build a mechanism that would allow a robot to subject to personal income tax on a notional wage equivalent to a similarly skilled human worker.

The framing has since been overtaken by reality.

A robot replaces one worker. A single AI system can replace hundreds or thousands. The unit of measurement isn’t the individual worker displaced. It’s the total wage bill eliminated.

An AI platform handling 50,000 interactions a day across three countries doesn’t have a human equivalent. It has a payroll equivalent — the combined wages of every human worker who would otherwise have been hired to do that work, in the jurisdictions where that work occurs.

That is the notional income. That is the base.

How It’s Collected

The mechanics don’t need to be invented. They already exist.

Every company already runs payroll.
Every company already remits taxes on behalf of workers.

The same logic can apply here. When an AI system performs economically valuable work — work that would otherwise require human labor — that activity is attributable economic output. It can be treated for tax purposes as income.

The company doesn’t lose that revenue. But it remits a portion of it, on behalf of the AI system, just as it would for an employee.

The exact attribution method will vary — displaced labor cost, revenue contribution, or activity metrics are all viable starting points. The rate itself will need to be set jurisdiction by jurisdiction, and it will change over time as the disclosure data matures. That’s not a weakness. Tax structures adapt almost every year. This one will too.

One important threshold: this framework is not designed to burden small companies using AI tools to augment a handful of roles. It is designed for deployments that materially displace human labor at scale. A disclosure threshold tied to revenue or the number of roles affected would protect smaller operators while capturing the economic activity that actually moves markets.

The corporate tax objection is worth addressing directly: companies already pay tax on AI-driven profits. That is true. But corporate tax captures profit, not displacement. A company can offshore its profits. It cannot offshore the customer service call that happened in Vancouver. These are different obligations addressing different economic realities.

No new infrastructure.
No speculative enforcement model.
Just an extension of systems that already process billions of transactions every year.

The Snow Still Gets Cleared

The worker who gets paid participates in the economy. The worker who doesn’t, doesn’t.

That principle held when it was a kid and a driveway. It holds when it’s an AI system and a global supply chain.

If the worker participates in production, it must participate in the economy.

The framework proposed here isn’t radical. It’s conservative in the most literal sense — it conserves the economic logic that made growth possible in the first place.

We created corporations because economic reality outgrew the individual. We are creating AI systems that are outgrowing the corporation as the unit of economic accountability.

The answer isn’t to slow the snow from falling.

It’s to make sure someone still gets paid to shovel it

The questions this raises — how to measure, how to distribute, how to coordinate globally — each deserve their own treatment. This is the first of several.