AI is beating down many public software companies. But not many of the leaders servicing the physical world.
Some is that they don’t tie pricing models primarily to seats. But it isn’t only the pricing model. Their customers’ end products aren’t the ones AI is eating.
Shopify reported strong and accelerated earnings this past quarter. Revenue $3.58B, up 34%, 33% in constant currency. GMV $115.6B, up 32%, the fifth consecutive quarter above 30%. Operating income $488M against $291M a year ago. Free cash flow $654M at an 18% margin, up from 16%. The stock went up almost 20% on the print.
Toast reported no deceleration and a record number of net new restaurant additions. Revenue $1.91B, up 23%. ARR $2.4B, up 25%. A record 9,500 net new locations in a single quarter. Recurring gross profit streams up 28%, faster than revenue. They raised full-year recurring gross profit growth guidance to 23-25% from 21-23%.
Samsara selling to the connected transportation world and related segments is not slowing down. ARR $1.99B, up 30%, with revenue up 31%. Net new ARR of $101M, itself up 30%. ARR from $1M+ customers up 62% and accelerating for the fourth straight quarter. Third consecutive quarter of GAAP profitability.
For comparison, the median public B2B company is growing about 13%. Salesforce’s most recent quarter was up 13%, and roughly 4 points of that came from the Informatica acquisition. Its full-year guide is 11%, including about 3 points of Informatica. Restaurants, shipped goods, and truck fleets are outgrowing CRM by two to three times. Two things are driving it, and the second one matters more.
1. None Of Them Are Primarily Priced Per Seat #
Shopify takes a percentage of GMV, processing $78B through Shopify Payments in the quarter at 68% penetration, with cumulative payment volume now past $1 trillion. Toast charges per location plus payments, and added 9,500 locations. Samsara charges per connected asset, so a truck, a trailer, a piece of equipment.
Seats are not entirely absent from any of them. Toast sells payroll priced per employee. Shopify’s plans meter staff accounts. But headcount is not the engine, and none of the three lead with it. What moves the number is throughput. They grow when the customer sells more, opens more, or moves more, not when the customer adds people.
That difference is showing up across the whole market. Total software spend is forecast up 15.5% this year to $1.468 trillion, while CRM, sales, marketing, CX, collaboration, and productivity are all in single digits. The market didn’t stop growing. The unit did.
When your pricing unit is a person, and your customer is under pressure to hold headcount flat while output rises, your growth is capped by their org chart. When your pricing unit is a transaction, a location, or a vehicle, you capture the productivity gain instead of getting squeezed by it.
It also leaves somewhere to go.
Shopify’s $3.58B of revenue is about 3.1% of the $115.6B of GMV crossing its platform. Toast’s $1.91B is about 3.1% of $60.7B of gross payment volume. ServiceTitan earns roughly 1.2% of the $21.7B of transaction volume its contractors invoice, against a stated ceiling closer to 2% at full platform adoption. Growth in this model comes from two places at once, the customer’s volume rising and the slice of it getting bigger, and neither one is near done.
2. Their Customers’ End Product Isn’t What AI Is Eating #
This is the bigger one.
AI is compressing the categories where the deliverable is text, code, tickets, and documents. Support software, content tools, dev seats, and anything sold on “we help your team produce more of X” where X is knowledge work. The buyer is now asking whether they need fewer seats, or whether an agent replaces the workflow the software was wrapping.
None of that applies to a restaurant. AI does not reduce the number of meals served. It does not reduce the number of packages that have to be picked, boxed, and shipped. It does not reduce the number of trucks that have to physically move freight.
ServiceTitan is a useful test of where the line actually sits. Their pricing scales with technician count, so in a narrow sense it looks like a seat model, and they grew 25% last quarter on revenue of $268.8M with gross transaction volume up 23% to $21.7B. But a technician is a unit of physical capacity, not a knowledge worker. No agent is going to crawl under a house. That seat count goes up when demand for the trade goes up, and demand for the trades is not being deflated by a model release.
Shopify’s quarter makes the point from the other direction. Their merchants are seeing a reported 3x surge in traffic from AI shopping surfaces. Agentic commerce is currently a demand tailwind, not a threat, because at the end of an agentic transaction somebody still has to ship an actual object. The agent can find the product. It cannot manufacture it, warehouse it, or put it on a truck.
Compare that to a software company whose customer’s entire output is something an agent can now produce. The end market is being repriced underneath them, and execution doesn’t fix that.
The Objection: “These Are Payments Companies With Software Multiples” #
Worth addressing directly, because it is the strongest pushback.
A large share of Shopify’s and Toast’s revenue is payments, and payments carries lower gross margin than subscription software. Shopify’s merchant solutions revenue was $2.78B against $802M of subscription solutions.
The answer is in the profit lines, not the revenue line. Toast’s recurring gross profit streams grew 28%, ahead of its 23% revenue growth, which is why they raised the recurring gross profit guide rather than just the revenue guide. Operating income was $152M against $80M a year ago.
Shopify’s gross margins did compress, from 48.6% to 47.8%, on a heavier payments mix. Gross profit grew 31% against 34% revenue growth. But operating income grew 68% and free cash flow margin went from 16% to 18%. The payments mix costs a little at the gross line and pays it back several times over below it.
The better framing: these companies made the transaction their pricing unit. Payments isn’t a separate business bolted onto a software business. It is how they charge for the software.
The Control Case: Procore #
Procore reported on July 29. Revenue $375.2M, up 16%, with its first quarter of GAAP operating profitability, a 21% non-GAAP operating margin, and an $845M deal to acquire DroneDeploy.
Construction is about as un-disruptable by AI as any market gets, and Procore is the category leader. But it prices on users and modules, so it captures the industry’s activity only indirectly. Growth is less than half of Shopify’s.
One company does not settle the question. But it is the cleanest available test of which half of the thesis does the work. Being attached to a physical end market gets you a durable market. Pricing on that market’s throughput is what gets you 30% growth at $2B of scale.
It is also worth noting that Shopify and Toast are SMB-heavy while Samsara sells to large fleets, and all three land in the same place. The segment isn’t the variable. The pricing unit is.
What Could Break This #
Consumer spending. Toast’s revenue is levered to restaurant traffic and Shopify’s to consumer purchasing. A real downturn hits both faster than it hits an enterprise contract, and most of Toast’s customers are small businesses that are harder and costlier to retain when times get tight.
Agentic disintermediation. Today agentic commerce sends traffic to Shopify merchants. If the agent layer consolidates demand and starts dictating terms, the merchant becomes a supplier and the margin moves. Shopify is investing hard to be the infrastructure rather than the disintermediated party, but it is a live risk.
Deceleration math. Toast’s revenue growth has come down slightly from roughly 25% in mid-2025. Growth at this scale is hard to hold, and all three are valued on the assumption that it does. Samsara and Toast have both derated sharply over the past year even as the businesses compounded.
The Two Questions Worth Asking About Your Own Company #
Look at your pricing unit and ask what it is a proxy for. If the answer is “how many people my customer employs,” you are pricing against the strongest cost-cutting pressure in a decade. If it is “how much my customer sold, shipped, served, or moved,” you are pricing against the thing they are trying to grow.
Then ask whether an agent can produce your customer’s output. If it can, your TAM is being repriced whether or not your product is good.
Neither answer is easy to change after the fact. That is why the gap between 34% and 13% is likely to widen before it closes.