Canadian SMEs Didn't Invest in US AI Infrastructure, But They're Still Funding Its Development

March 2026 | midagent | John Pope

US Hyperscalers are Funding the Biggest Infrastructure Buildout in Human History

Google, Meta, and Amazon are building the most expensive infrastructure in corporate history. Your business is funding it. You were never asked.

Somewhere in the budget documents of a small Canadian business — a physiotherapy clinic in Guelph, a furniture maker in Granby, a boutique travel agency in Halifax — there is a line item that has been quietly growing for three years. It doesn't say "AI infrastructure levy." It doesn't say "Silicon Valley capital subsidy." It says "digital advertising" or "platform fees," and it looks, on the surface, like a normal cost of doing business in the modern economy.

It is not a normal cost of doing business. It is something structurally different, and understanding what it actually is may be the single most important economic literacy exercise that Canadian policymakers undertake in 2026.

The Numbers That Should Stop a Room

Let's start with scale, because without scale the argument remains abstract.

Bain & Company calculates that justifying the current AI infrastructure buildout requires $2 trillion in new annual revenue by 2030 — against a current AI revenue baseline of roughly $20 billion. Bain & Company That is not a rounding error. That is a 100-fold gap between what the hyperscalers are spending and what the market for AI services is currently generating.

The five largest hyperscalers — Amazon, Microsoft, Google, Meta, and Oracle — are projected to spend $602 billion on capital expenditures in 2026, with Amazon, Microsoft, Google, and Meta each individually exceeding $100 billion. Introl To put the financing dimension in context: hyperscalers raised $108 billion in debt in 2025 alone — more than three times the historical annual average. Bank of America estimates that combined capex will consume 94% of these companies' operating cash flow. Comsoc

These are not normal investment cycles. The combined 2026 capital expenditure projections for Amazon, Google, Meta, and Microsoft exceed the GDP of all but the top 20 national economies. Tech Insider

The question this post is asking is deceptively simple: where does the rest of the money come from?

The answer is already in your accounts payable.

The Pre-Loaded Bill

Canada's digital advertising market reached approximately $18.2 billion in 2024, with Google commanding roughly 50 percent of the total and Meta capturing another substantial share Made in CA — meaning two American companies control the majority of the channel through which 1.19 million Canadian small businesses must reach their customers.

Cost per click increased for 87 percent of industries on Google Ads between 2024 and 2025, with some sectors like Education and Beauty seeing increases exceeding 40 percent. WordStream Meta's average cost per ad increased 14 percent in the fourth quarter of 2024 alone. Overdrive Interactive These are not isolated fluctuations. They are the continuation of a multi-year trend: the average cost per click across Google's ad network more than doubled between 2015 and 2024, representing a compound annual growth rate of approximately 8.2 percent. Rankfuse

The platforms' official explanation is that AI-powered targeting improvements make their ads more effective, justifying higher prices. This is partially true and largely irrelevant, for reasons we will come to. The deeper story is in the timing and the direction of the cash.

Google's parent Alphabet is spending $175 to $185 billion on capex in 2026. Meta's operating margin on advertising revenue expanded from 38 percent in 2024 to approximately 42 percent in 2025, even as it poured tens of billions into infrastructure — and 97 percent of its revenue is advertising-dependent. MEXC Mark Zuckerberg himself told investors that advertising remains "the most important driver of growth" as the company spends hundreds of billions on AI infrastructure. Marketing Brew

The accounting is not subtle. Advertising revenue goes up. Infrastructure spending goes up. The platforms are not funding their AI buildout from air. They are funding it from the businesses that have no alternative but to advertise on their platforms.

The bill for the AI arms race has been pre-loaded onto the cost structure of the global small business economy — and it is being collected now, years before the AI services it is funding will be delivered.

The Forced IPO Nobody Signed

Here is the framing that makes this argument most legible to a business owner rather than an economist.

When a technology company wants to raise capital for a major infrastructure investment, it has options. It can issue equity — an IPO or secondary offering — in which investors choose to participate, understand the risk, and receive an ownership stake in exchange for their capital. It can issue debt, in which lenders choose to participate and receive a contractual return. Both mechanisms involve consent, disclosure, and some form of proportional benefit to the capital providers.

What Google, Meta, and Amazon have done instead is neither of these things. They have raised capital for their AI infrastructure buildout by raising the prices charged to businesses that cannot say no, cannot negotiate terms, cannot receive equity, and receive no proportional benefit from the assets being built.

The physiotherapy clinic in Guelph cannot decline to advertise on Google. Her patients search for physiotherapists on Google. If she stops bidding on those keywords, she becomes invisible. She has no alternative search engine with comparable reach. She has no alternative social platform with comparable targeting. She does not choose to participate in this arrangement. She participates because the alternative — invisibility — is commercially untenable.

She has involuntarily funded a fraction of a data centre in Virginia. She received no shares. She will receive no dividends. The data centre will be used, in part, to improve the AI systems that will make Google's advertising more precise — which will make her advertising more expensive next year.

This is the Forced IPO. Every small business that advertises on these platforms has been inducted into it, without a prospectus, without consent, and without any mechanism to exit.

The Circular Economy of Extraction

The Forced IPO is damaging on its own terms. But it becomes genuinely dangerous when you understand the flywheel it sets in motion — particularly for what comes after the buildout is complete.

The cycle works like this.

AI improves ad targeting. Platforms use this improvement to justify higher prices per impression. Small businesses pay higher prices because they have no alternative. The revenue funds more AI infrastructure. More AI infrastructure produces better targeting. Better targeting justifies higher prices still.

At every rotation of this cycle, more value is extracted from outside the platforms and concentrated inside them. The platform's moat — its ability to target, to reach, to optimize — deepens with every dollar of infrastructure investment. The small business's competitive position relative to the platform weakens correspondingly.

And then — this is the part that matters most for Canadian policymakers — the cycle enters its second phase.

Once the infrastructure is built, the platforms will begin selling AI services directly to businesses. The same small business that funded the infrastructure through advertising premiums will now be asked to pay subscription fees for AI tools — inventory optimization, customer service automation, marketing generation — built on that infrastructure. The capital has already been extracted. The services will be priced, once again, at monopoly rates. The small business will pay twice: once to build the infrastructure, and again to use it.

AWS, Microsoft Azure, and Google Cloud together held a backlog of nearly $700 billion in committed enterprise AI contracts by late 2025 Tech Insider — representing the forward revenue that justifies the capex. That backlog will be filled, in large part, by businesses that have spent the past three years involuntarily funding the infrastructure those services will run on.

This is not speculation about a distant future. It is the current trajectory, playing out in real time, in the accounts payable of every Canadian SME.

What This Means for Your Constituents: Three Time Horizons

For policymakers, the relevance of this argument is not abstract. It maps directly onto the lived experience of the 1.19 million small and medium businesses that employ the majority of working Canadians. Let's be precise about the timeline.

Right now, during the buildout: Canadian SMEs are experiencing real margin compression from advertising inflation that directly funds foreign AI infrastructure. Small business failure rates in Canada sit at 21.5 percent in the first year, with cash flow cited as the primary challenge for a third of business owners. MARKAGE The advertising cost escalation documented in this series is not a peripheral concern for these businesses — it is a front-line survival issue. Every dollar extracted through monopoly ad pricing is a dollar not available for payroll, inventory, or the kind of capital investment that drives productivity growth. Canada's productivity gap with the United States, documented extensively in previous posts, is partly a story of capital that never got invested because it was extracted instead.

In the medium term, as AI services go to market: The businesses that have been funding AI infrastructure through advertising premiums will face a second extraction event as that infrastructure is monetized through AI product subscriptions. A Canadian retailer currently paying 30 to 45 percent of gross revenue in platform fees and advertising costs will face an additional layer of AI tooling costs — priced, once again, by companies with monopoly market power. AI-related services currently generate only about $25 billion in revenue against $400 billion in annual infrastructure spend — meaning the monetisation wave has barely started. TradingView The businesses paying to build the infrastructure today will be the primary revenue targets when that wave arrives.

In the long term, if Canada does not act: The competitive moats built with this infrastructure will be used to entrench exactly the dependency patterns this series has spent months documenting. Canadian businesses will become increasingly reliant on AI services they do not own, built on infrastructure they involuntarily funded, governed by foreign law, priced at rates set by companies with no competitive constraint. The data generated by Canadian commercial activity will train the next generation of foreign AI models. The intelligence produced by Canadian government and institutional activity will flow through foreign-governed systems. The compounding logic that has made Google and Meta and Amazon into the most profitable companies in corporate history will continue to compound — at Canadian expense.

Bain estimates the industry still faces an $800 billion annual revenue shortfall against what is needed to fund the full buildout. Bain & Company That gap will be closed. The question is whether it will be closed by charging Canadian businesses more — or whether Canada will have built the sovereign commercial and intelligence infrastructure to route that spending differently.

The Clarity That Policy Demands

There is a word for a levy that is imposed without consent, collected from parties that have no power to refuse it, and used to build assets that will then be deployed against the interests of those who paid for them.

We are too polite to use it. So let's say this instead.

Canadian competition law, consumer protection law, and digital markets regulation were all built for a world in which the mechanisms of economic harm are visible. A price-fixing cartel is illegal because it extracts value from buyers without competitive justification. A monopolist is constrained because market dominance, left unchecked, produces prices that reflect power rather than value. A fraudulent financial product is prohibited because it collects capital from participants without disclosing the terms of the arrangement.

Every one of those legal frameworks applies, in principle, to what Google, Meta, and Amazon have done with advertising pricing during the AI infrastructure buildout cycle of 2023 to 2026. The harm is real, documented, and quantifiable. The mechanism is clear. The causal chain from monopoly pricing power to AI capex funding to future monopoly deepening is not hypothetical — it is the stated business logic of these companies, communicated to their shareholders in every earnings call.

What is missing is not evidence. What is missing is the political will to look at the evidence, name what it describes, and act.

A Call to Action

To every Member of Parliament reading this, and to every business leader who has watched their advertising budget grow and their margins shrink and told themselves it was just the cost of doing business in the digital economy:

It is not the cost of doing business. It is a transfer — from your business, from your employees' wages, from your customers' purchasing power — to five American corporations building infrastructure that will be used to charge you more. The transfer is structural, ongoing, and accelerating.

Canada has, right now, a rare convergence of political will and institutional capability to respond. We have a competition bureau that has begun to find its voice. We have a sovereign AI strategy with real funding. We have the Maple Eight pension funds with the capital to build alternatives. We have Cohere, ThinkOn, HyperTec, and Bell — a sovereign technology stack that is functionally complete. We have a Prime Minister who stood at Davos and told the world that Canada is done pretending the old order is sustainable.

What we need is the decision to connect those assets into a functioning alternative — one that gives Canadian businesses somewhere to route their advertising spend, their commerce transactions, and eventually their AI service needs, without every dollar flowing to infrastructure owned by someone else, governed by foreign law, and designed to extract more next year than it extracted this year.

The AI arms race has already drafted your constituents as its financiers. They didn't volunteer. They weren't asked. They are paying now for services that haven't arrived, and they will pay again when those services do.

The question for every politician in this country is not whether that is acceptable.

Everyone already knows it is not.

The question is what you are going to do about it — and whether you are going to do it before the flywheel makes the answer irrelevant.