Summary

  • Braze’s fiscal second-quarter revenue rose 26.2% to US$227.230 million, while cost of revenue increased 29.5% to US$75.374 million. Gross profit grew 24.6%, but GAAP gross margin declined 0.9 percentage points to 66.8%.
  • The year-on-year increase in cost of revenue was US$17.153 million. Braze identified US$7.4 million more in third-party premium messaging fees and US$5.2 million more in hosting, infrastructure and other third-party fees. Together, those lines accounted for about 73% of the increase.
  • That calculation locates the principal variable delivery toll; it does not prove that AI generated the cost. Braze said the spending supported overall revenue growth and did not publish AI-only revenue, usage or margin.
  • Management gave a different explanation for the gross-margin decline: personnel associated with an acquired workforce and amortisation of acquired technology, partly offset by technology-stack efficiencies. Absolute cost growth and margin movement therefore need separate bridges.
  • The next useful evidence is a channel-and-workload receipt: premium-message mix, hosted processing volume, unit cost, customer pricing, acquired-cost burden and realised efficiency. Without it, faster engagement can produce both stronger customer value and deeper supplier dependence.

Two invoices inside a growth quarter

The most revealing number in Braze’s fiscal second-quarter results is not the 26.2% revenue increase on its own. It is the movement underneath gross profit. Revenue reached US$227.230 million, up US$47.119 million from a year earlier. Cost of revenue reached US$75.374 million, up US$17.153 million. Gross profit still advanced by US$29.966 million, to US$151.856 million, but the cost base expanded faster than sales.

The Form 10-Q supplies an unusually useful bridge. Third-party premium messaging fees rose by US$7.4 million. Hosting, infrastructure and other third-party fees rose by another US$5.2 million. Their combined US$12.6 million equals about 73.46% of the total increase in cost of revenue.

That ratio is not a gross-margin decomposition. It answers a narrower question: where did most of the additional delivery spending appear? The answer is outside Braze’s own payroll and property, in the channels and computing services through which its product operates.

This is the economic shape of a customer-engagement platform. An email, push notification or in-app interaction can look like software at the user interface, but execution depends on networks, messaging aggregators, cloud capacity, data services and other applications. Premium channels can carry per-message charges. Hosted processing and storage rise with workloads. A successful campaign may therefore enlarge both the value delivered to a customer and the toll paid to an external supplier.

Braze says the two increases supported overall revenue growth. That is important discipline: the costs are not evidence of waste merely because they rose. The investment question is whether revenue, price and technical efficiency can keep outrunning them.

Cost growth is not the same as margin causation

The quarter’s GAAP gross margin was 66.8%, down from 67.7% a year earlier. It would be easy to point from the US$12.6 million third-party increase to the 90-basis-point decline and call the case closed. Braze does not describe the quarter that way.

In its filing, the company attributes the margin decline primarily to personnel costs associated with an acquired workforce and amortisation of acquired technology. Technology-stack efficiencies partly offset those pressures. The third-party lines explain much of the increase in absolute cost of revenue; management’s margin account emphasises acquisition-related costs and amortisation. Both statements can be true because margin depends on the relationship among costs, revenue growth and mix, not simply on which dollar line increased most.

The distinction protects the analysis from two bad shortcuts. First, a variable cost can rise sharply while revenue attached to it rises fast enough to preserve or improve unit economics. Second, a smaller fixed or non-cash line can pressure the margin rate if it is newly added and not yet absorbed by the revenue base. The public numbers do not allocate gross profit among subscription commitments, excess activity, professional services, messaging channels or AI-enabled products.

Professional services and other revenue illustrates the problem. It rose to US$19.6 million from US$8.3 million, an increase of roughly 136%. Subscription revenue reached US$207.7 million, up about 20.9%. Those are very different growth rates, but Braze does not publish a gross margin for each revenue line. It would be speculation to make services mix the cause of the consolidated margin movement.

The acquisition timetable also changes the comparison. Braze completed its acquisition of OfferFit on 2 June 2025 for total consideration recorded at US$303.2 million, comprising US$195.6 million in cash and US$107.6 million in stock. The prior-year quarter included roughly two months after the closing; the current quarter included a full three months. The filing gives acquisition-related explanations, but it does not disclose a standalone OfferFit contribution that would support a precise revenue or margin estimate.

The entitlement model sets the commercial boundary

Braze sells subscription customers committed entitlements. A customer can buy an incremental committed tier, and revenue from that added commitment remains subscription revenue. Activity above the committed amount can also produce excess-usage fees, but Braze says those fees have not been material to date.

That architecture matters because an external delivery toll does not automatically pass through to the customer in the same period. A committed price may be set before messaging or cloud unit costs move. A customer may alter its channel mix while staying inside an entitlement. Braze can route or optimise workloads, renegotiate suppliers and redesign the product, but it cannot make every third-party price or customer behaviour disappear.

The company also retains important controls. It chooses how to bundle volume, which premium channels to encourage, how to price incremental tiers and where to deploy optimisation. It can use purchasing scale or technical changes to lower the cost per unit. The stated technology-stack efficiencies are evidence that this control surface exists, though they are not quantified as a standalone saving.

Customers control a different part of the chain. They decide audience size, message frequency, campaign cadence, channel and the extent to which automated decisioning is used. End consumers determine whether engagement converts into attention, a purchase or fatigue. Messaging and cloud suppliers affect the input bill. Economics emerge from the interaction, not from any single party.

AI adoption raises the need for attribution, not the right to assume it

Braze says adoption of BrazeAI Operator, Agent Console and Decisioning Studio is accelerating. That establishes commercial and product momentum. It does not establish an AI-specific revenue line, customer count, workload measure or gross margin. The quarter’s disclosures contain no bridge from AI use to the US$7.4 million premium-messaging increase or the US$5.2 million infrastructure increase.

There are plausible mechanisms in both directions. Better decisioning might reduce irrelevant messages, shift activity toward more effective channels and improve the value earned from each delivery dollar. Agentic tools might also increase the number of tests, audience segments and personalised interactions, producing more processing and channel traffic. The same product can reduce cost per useful outcome while increasing total supplier spend because the useful workload is larger.

That ambiguity is why “AI caused higher cost” is not a finding. Nor is “AI will expand margin”. Both propositions require denominators that Braze has not published: decisions or messages per customer, channel mix, hosted-processing units, cost per unit, incremental revenue or retention, and the support and amortisation attached to the product.

Company-wide customer evidence is encouraging but still too broad for attribution. Braze ended the quarter with 2,789 customers, up from 2,422. The number producing at least US$500,000 of annual recurring revenue rose to 361 from 282. Net dollar retention was 110% for the whole base and 112% for the large-customer cohort. These figures show expansion and scale; they do not say which products, channels or cost inputs produced it.

Growth came from both installed and new customers

The revenue bridge adds another useful boundary. Of the US$47.119 million increase, about 55.3% came from growth among existing customers and 44.7% from new customers. Braze therefore did not rely exclusively on either land or expand.

For delivery economics, those cohorts may behave differently. An established customer can add messages, channels and entitlements while using implementation resources already in place. A new customer may bring onboarding and service needs before reaching a steady workload. A large customer can improve purchasing leverage yet concentrate bargaining power. Aggregate revenue growth hides these paths.

Remaining performance obligations of US$1.0925 billion, including US$691.1 million current, provide contract visibility. They are useful context but not the centre of this analysis. A backlog balance cannot reveal whether a campaign uses a premium messaging channel, how intensively data is processed or whether supplier cost was offset by pricing. The cost bridge asks a different question from the contract balance.

Cash also imposes a useful check. Braze reported US$24.2 million of operating cash flow and US$21.7 million of free cash flow for the quarter. Over six months, operating cash flow was US$52.3 million even after a US$43.8 million net loss, supported by US$109.9 million of non-cash adjustments and partly offset by US$13.8 million of changes in operating assets and liabilities. Stock-based compensation of US$71.5 million and amortisation of deferred contract costs of US$25.1 million were large components of the non-cash bridge. Cash generation is real, but it does not turn non-cash costs into irrelevant ones or isolate delivery efficiency.

The receipt that would resolve the question

A decision-useful delivery receipt can be compact. It would show premium messages by channel, hosted processing or another stable workload unit, and the associated third-party expense. It would separate committed entitlement revenue from incremental tiers and immaterial excess usage. It would place acquired-workforce cost and acquired-technology amortisation beside the variable supplier lines, then show the savings attributed to stack efficiency.

The aim is not disclosure for its own sake. The receipt would distinguish four very different outcomes. Braze might pass higher channel costs through in price. It might steer customers toward lower-cost channels without weakening engagement. It might use better infrastructure to process more activity at a lower unit cost. Or it might deepen a supplier dependency faster than customer economics improve. Consolidated growth and margin cannot identify which mechanism dominates.

Sources