Why Shopify Needs to 50x Its Capex

Shopify spent $4 million on capital expenditure last quarter. Four million. On a company generating billions in revenue with a market cap north of $100 billion.

Let that number settle for a second. Then compare it to Amazon at 10-12% capex intensity, or Microsoft and Google at 15-25% as they pour concrete and rack GPUs for the AI era. Shopify's capex intensity ratio sits at roughly 0.2%.

That's not lean. That's structurally exposed.

The Numbers Don't Lie (But They Do Flatter)

Q1 2025: $4M in purchases of property and equipment, down from $6M in Q1 2024. Full-year capital expenditure lands somewhere around $15-25M. Total property and equipment on the balance sheet? $46M net. That's it. The whole physical footprint of the company that processes hundreds of billions in GMV.

R&D was $377M in Q1 alone. The stock buyback programme is measured in billions. Meanwhile, capex — the thing that builds durable infrastructure — is a rounding error.

Free cash flow margins look spectacular precisely because the capex deduction is nearly zero. Wall Street loves it. Every analyst presentation celebrates the "asset-light" model. The market rewards the current multiple.

But asset-light has a second name: asset-dependent on someone else.

The Cloud Rental Trap

Shopify runs on Google Cloud Platform. They don't own data centres. They exited logistics when they sold Deliverr and SFN to Flexport in 2023. Their capex is essentially office fit-outs and minor equipment.

This was a brilliant strategy for the 2015-2023 era. SaaS platforms were supposed to be thin layers of software sitting on rented compute. The economics made perfect sense when your workload was serving web pages and processing payments.

The AI era breaks this model in three specific ways.

First, cloud pricing has been friendly because Shopify is a marquee customer. Google wants Shopify on GCP for the reference case. But as AI workloads scale — inference at the edge, real-time personalisation across millions of storefronts, continuous model training on catalogue data — Google's own demand for that same capacity increases. Shopify's negotiating leverage erodes precisely when their compute needs accelerate. Every hyperscaler is now prioritising their own AI workloads. The queue forms behind the landlord.

Second, inference latency creates compounding advantages for whoever controls the stack. When Shopify runs Sidekick, the Shop app's AI recommendations, and catalogue enrichment features, every millisecond of inference latency matters. Custom silicon, dedicated compute capacity, and co-located data create measurable advantages that renting simply cannot match. Amazon understood this fifteen years ago, which is why AWS exists. They didn't just rent infrastructure — they became the infrastructure.

Third, data gravity. Shopify sits on one of the most valuable commerce datasets on earth. Millions of storefronts, billions of transactions, real-time consumer behaviour signals. The value of that data increases exponentially when you can train models directly on it, in your own environment, without shipping it through someone else's pipes. Running AI workloads on rented cloud means your most strategic asset lives in your landlord's house.

The Stripe Problem

The comparison most people miss isn't Amazon or Google. It's Stripe.

Stripe has been building proprietary infrastructure for financial rails: their own card-present processing stack, their own ML fraud systems, their own compliance engines. They've invested heavily in owning the critical path.

Shopify Payments still runs on top of Stripe in many markets.

Think about what that means. The checkout — Shopify's most strategically important surface, the place where billions in GMV converts — has a dependency on infrastructure Shopify doesn't own. That dependency is hidden inside the "low capex" story. It looks like a partnership on the balance sheet. It's a vulnerability on the strategic one.

Stripe can raise prices. Stripe can prioritise its own features. Stripe can launch competing commerce tools (they already are). Every dollar Shopify didn't spend on payments infrastructure is a dollar of optionality they handed to a potential competitor.

The Buyback Tell

The multi-billion dollar buyback programme while spending $15-25M annually on capex is the financial equivalent of renovating your kitchen while renting the house.

It tells you exactly what management is optimising for: the current stock price, not long-term defensibility. Buybacks are a statement that management believes returning capital to shareholders generates more value than deploying it into infrastructure.

In a mature business, that might be correct. But Shopify is entering what might be the most capital-intensive technology transition since the mobile internet. AI infrastructure isn't a feature you bolt on. It's a foundation you build or you rent. And renting compounds your dependency on every cycle.

The bull case says Shopify's real capital deployment happens through opex — that $377M quarterly R&D number. Software engineers writing code, not pouring concrete. And there's truth to that. Software-defined infrastructure doesn't require physical assets in the same way. This connects to broader trends in infrastructure investment patterns

But that argument had a shelf life, and we're watching it expire. When your AI models need dedicated inference hardware, when your data pipeline requires co-located compute, when your competitive advantage depends on milliseconds of latency — opex alone doesn't cut it.

What 50x Actually Looks Like

If Shopify moved from $20M to $1B in annual capex, it would still be running at roughly 10-12% capex intensity on their current revenue base. That's where Amazon sits. It's actually conservative for a company that wants to be an AI-native platform.

A billion dollars buys:

• Dedicated AI compute clusters — custom inference hardware optimised for commerce workloads, not shared cloud instances

• Edge infrastructure — processing closer to the merchant and buyer, cutting latency for real-time personalisation and fraud detection

• Proprietary payments processing — reducing or eliminating Stripe dependency, capturing the full payments margin

• Data centres with data gravity — training models on commerce data without shipping it through third-party clouds

• Custom silicon R&D — following the Amazon (Graviton, Inferentia) and Google (TPU) playbook for workload-specific chips

None of this is exotic. Every major platform company that's serious about AI is making these investments right now. The question isn't whether Shopify needs to spend more. It's whether they can afford to keep spending this little.

The Counter-Argument (And Why It's Weakening)

Shopify's defenders make three points:

"Asset-light is the whole thesis." It was. When the moat was merchant adoption and app ecosystem lock-in. In the AI era, the moat increasingly depends on who owns the inference stack. The thesis needs updating.

"GCP partnership gives them world-class infrastructure." It gives them access to infrastructure that Google can reprice, reprioritise, or compete with at any time. Access is not ownership. Ask any company that built on Twitter's API how that worked out.

"R&D spending is the real investment." R&D writes software. But software needs hardware to run. And when your hardware is rented from a company that's simultaneously building competing AI commerce features (Google Shopping, anyone?), your R&D output is subject to someone else's strategic decisions.

What This Means For Merchants

If you're running a store on Shopify, this isn't abstract corporate finance. It's your platform risk.

When Shopify's AI features are slower than Amazon's because they're running on shared cloud instead of dedicated silicon, your conversion rate pays the price. When Shopify Payments is constrained by Stripe's roadmap, your checkout experience is capped by someone else's priorities. When the next wave of AI-native commerce tools requires infrastructure Shopify doesn't own, your platform can't compete.

None of this happens tomorrow. But the capex decisions Shopify makes in 2025-2026 determine the platform's competitive position in 2028-2030. Infrastructure investment compounds. Infrastructure neglect compounds too, just in the wrong direction.

The Bottom Line

Shopify has the revenue, the free cash flow, and the strategic imperative to invest dramatically more in infrastructure. They're choosing not to. Instead, they're buying back stock and celebrating free cash flow margins that look great precisely because they're not investing in the foundation.

At 0.2% capex intensity, Shopify isn't an asset-light platform. It's a platform that's made a bet: that renting infrastructure will remain cheaper and more flexible than owning it, indefinitely, even as AI transforms every layer of the stack.

History suggests that bet has a time limit. Ask any company that outsourced a core competency and then tried to bring it back in-house. The longer you rent, the harder it gets to own.

50x isn't aggressive. It's the minimum to stay in the game.

Related: Agentic Commerce Is Here. Your Shopify Store Isn't Ready.