Shopify reported second-quarter revenue of $3.58 billion on Wednesday, up 34% from a year earlier, and the stock ran as much as 26% higher. In a week when Wall Street punished almost every company that mentioned artificial intelligence, Shopify got rerated for the same theme, and the difference comes down to who is paying for the infrastructure.
The Numbers, Briefly
Revenue of $3.58 billion beat the $3.45 billion analysts had modeled. Gross merchandise volume rose 32% to roughly $116 billion. Free cash flow margin landed at 18%, and cumulative GMV since inception crossed $1.8 trillion.
The guidance did the heavy lifting. Shopify told investors to expect third-quarter revenue growth in the low thirties, against a consensus that had penciled in 26.3%, which as The Globe and Mail noted in its results coverage implies something close to $3.75 billion versus the $3.59 billion the street was carrying. Raising a growth rate that was already accelerating is a rare move, and it is the single line that explains the size of Wednesday’s reaction.
Everyone Else Is Buying the Picks
Sit this quarter next to the rest of the week and the asymmetry gets loud.
AMD beat on revenue, earnings, and guidance, then fell 8% because capital expenditures came in at nearly triple what analysts had modeled. SpaceX grew revenue 92% and lost roughly a tenth of its market value across two sessions after posting $18.4 billion of quarterly capex, $15.83 billion of it on AI infrastructure. Sandisk and Western Digital took comparable treatment. The market’s message has been consistent: we will not keep funding the buildout on faith.
Shopify is on the other side of that ledger entirely. It does not train frontier models. It does not lease data centers or negotiate for power. The AI systems generating its incremental demand belong to OpenAI, Google, Perplexity, and Anthropic, and those companies are absorbing the capital cost of building the discovery layer while Shopify collects a take rate on whatever converts through it. The buildout is somebody else’s balance sheet problem.
That is the cleanest position in the AI trade right now, and almost nobody is framing Shopify that way. It is not an AI company. It is a toll booth on AI commerce.
The Statistic That Actually Matters
Buried in the disclosures is a number with real structural weight: AI-driven traffic and AI-driven orders each tripled year over year, and 75% of AI-attributed purchases came from outside Shopify’s top 100 product categories. PYMNTS pulled that detail out of the earnings call, and it deserves more attention than the headline growth rate.
Conventional search concentrates demand. A shopper types a category term, sees the same handful of brands that won the SEO and ad auction, and buys from one of them. Scale wins, and it compounds. Every incumbent advantage in e-commerce over the past fifteen years was built on that dynamic.
A shopper describing what they want in a sentence to a language model gets a different answer, because the model is matching a description rather than ranking bidders. Long-tail products that never had a chance in a keyword auction surface on relevance instead. If three quarters of AI-driven purchases are landing outside the hundred biggest categories, AI search is functioning as a redistribution mechanism, pushing demand down toward smaller merchants rather than up toward the largest.
Shopify hosts the long tail. That is the whole company. Which means the shift is not merely a tailwind for Shopify, it is disproportionately a tailwind for Shopify specifically, in a way it is not for a marketplace whose economics depend on volume concentration.
New buyer orders arriving through AI channels are also converting nearly twice as fast as those from other channels, which suggests the intent quality is genuinely higher and not just novelty traffic.
The Google Question, Answered Carefully
Shopify’s other notable claim is that AI search is adding traffic rather than replacing Google, which TechCrunch reported straight from management. Treat that with the skepticism any self-interested framing deserves, and then notice that Shopify is one of very few companies with the transaction data to actually know.
We have tracked the earlier signs of this reordering, from ChatGPT putting a buy button directly inside the chat window to the measurable migration away from conventional search. The additive read is plausible for now because AI shopping is early and incremental. It gets harder to sustain the moment AI referrals start growing while search referrals shrink in absolute terms, and Shopify will see that crossover in its own dashboards before the rest of the market does.
On the merchant side, the AI assistant Sidekick handled close to 34 million merchant conversations in the quarter with daily active usage up 3.6 times year over year. That is a genuine retention mechanism. A merchant who runs their storefront through your assistant is considerably harder to migrate to a competitor.
What Would Break the Thesis
Two things, and neither shows up yet.
The first is disintermediation. Shopify’s position is comfortable precisely because the AI platforms send traffic and let merchants close the sale. If OpenAI or Google decides that owning checkout is worth more than owning referral, the toll booth becomes a competitor rather than a partner. Instant checkout inside a chat interface is the early shape of exactly that, and Shopify’s current advantage is that nobody else has the merchant catalog. Catalogs are not permanently defensible.
The second is a take-rate squeeze. More than 14% of US e-commerce now runs through Shopify, and market share at that level tends to attract both regulatory interest and pricing pressure from the platforms feeding it demand.
For now, though, Shopify has produced the thing this market has been asking for since AI capital budgets crossed into the hundreds of billions: a company demonstrating AI revenue, at scale, on somebody else’s capex. Investors paid 25% for it in a single session. The interesting question is how many other businesses sit in that same position and have not worked out how to say so.