Summary

  • Box's remaining performance obligations reached US$1.7 billion at 31 July 2026, up 15% year on year. Short-term RPO was US$904.7 million and long-term RPO was US$787.0 million.
  • RPO measures contracted revenue not yet recognised. It does not measure how many files customers store, how intensively they use API or AI entitlements, what cloud capacity they consume, or whether the resulting work is accepted and retained.
  • Box kept GAAP gross margin at 79.1%, but disclosed a US$3.2 million rise in quarterly public-cloud infrastructure cost and warned that customers using Box for a larger share of their content storage could lower gross and operating margins.

Box's contract book is advancing faster than its income statement. Fiscal-second-quarter revenue rose 9% to US$321.1 million, while remaining performance obligations rose 15% to US$1.7 billion. At constant currency the respective growth rates were 11% and 17%. That is commercially encouraging. It is not, by itself, a measurement of the work Box will have to perform to earn the contracted revenue.

The distinction is embedded in Box's latest Form 10-Q. RPO is contracted revenue not yet recognised. It combines deferred revenue with backlog that Box deems certain to be invoiced and recognised under non-cancellable arrangements. Short-term RPO — the portion expected to turn into revenue within twelve months — was US$904.7 million, up 11%. Long-term RPO was US$787.0 million, up 18%.

These figures answer a valuable question: how much revenue is attached to qualifying contracts and remains to be recognised? They do not answer a different one: how much content, retrieval, API activity and AI work will pass through the platform while Box satisfies those contracts?

One contract figure contains several mechanisms

Box says the rise in RPO came from expansion inside existing customers as they broadened deployment, conversion to multi-product Suites, the timing of customer-driven renewals, longer average contract terms and new customers. Currency reduced the reported growth rate by about 290 basis points.

Those mechanisms do not carry the same workload implication. A longer contract can increase long-term RPO without changing this month's file volume. A Suite conversion can add functionality and price while actual use develops gradually. A larger user entitlement can remain partly unused. Conversely, one customer can place substantially more content into Box or run more AI work before a renewal changes the contract value.

Box's own pricing description makes the boundary clearer. Subscription fees depend on customer requirements including user numbers, API and AI-unit entitlements, and deployed functionality. The operative word is entitlement. It describes a contractual right or allowance, not a public receipt of actual consumption.

An investor can therefore observe the contract perimeter but not the consumption perimeter. The filing does not provide active users by paid cohort, content stored or transferred, API calls, AI units consumed, model workload, cloud cost by product, or gross margin by account. It also does not connect those inputs to a completed customer task. RPO cannot fill those cells simply because it is large and growing.

The margin did not fall; the exposure is still real

The current financial counterevidence is strong. GAAP gross profit was US$254.0 million and gross margin was 79.1%, unchanged from the comparable quarter. GAAP operating margin improved to 10.2% from 7.0%. The earnings release also reported billings of US$309.5 million, up 17%, operating cash flow of US$70.8 million, up 54%, and non-GAAP free cash flow of US$59.7 million, up 67%.

This is not evidence of present margin deterioration. It is the right baseline against which to read an explicit future risk.

Quarterly cost of revenue rose US$5.7 million, or 9%. Box attributed the increase primarily to US$3.7 million more amortisation of capitalised software and US$3.2 million more public-cloud infrastructure cost, partly offset by declines elsewhere. Cost of revenue remained flat as a share of revenue. Box says strong and growing use of its platform and Box AI, combined with public-cloud capacity dynamics, should lift cost of revenue in absolute dollars over time, although the percentage may fluctuate.

The filing then supplies the decisive sentence. Only a small percentage of customers currently use Box to organise all their internal files. If more organisations, particularly large enterprises, use the service as a larger component of content storage, Box says gross and operating margins could be lower.

That warning describes an economic success that can carry its own cost. Winning a larger content footprint can deepen retention, improve the context available to AI products and make Box harder to replace. It can also increase storage, indexing, transfer, retrieval, resilience and compute requirements. Contract expansion and margin expansion are related outcomes, not synonyms.

Hosting concentration adds a capacity clock

Box outsources a substantial majority of its cloud hosting to Google Cloud Platform. Its filing discusses shortages and lead times for AI processors, memory and optical-networking equipment, along with data-centre power constraints. It must secure enough capacity to serve customers without paying unnecessarily for excess.

At quarter-end, Box had US$517.444 million of non-cancellable purchase obligations, primarily for public-cloud hosting and IT software and support, over terms of two to five years. The fiscal-2026 Form 10-K provides the prior contractual context.

Those supplier obligations are not a hedge against RPO and cannot be netted with it. They involve different counterparties, recognition rules and time horizons. Their relevance is narrower: Box has a supplier-side capacity clock as well as a customer-side revenue clock, and the public filing does not join them at the customer or workload level.

The missing receipt begins after entitlement

A decision-grade bridge would retain contract identity, product and Suite, term and price. It would then add paid and active users, content footprint, storage and transfer, API and AI entitlements, actual consumption, hosting and software cost, completed customer task, exception rate, renewal or expansion, and account or cohort gross margin.

No single row needs to disclose a customer's identity publicly. Aggregated cohorts could show whether broad content adoption produces operating leverage, whether AI use arrives inside prepaid allowances, and whether higher consumption predicts retention strongly enough to pay for the capacity it uses.

Until that bridge exists, the disciplined conclusion has two halves. Box's US$1.7 billion RPO is credible evidence of contracted revenue and the current accounts show resilient profitability. But wider content and AI adoption sits outside RPO's perimeter. The same adoption can strengthen the franchise while increasing the work required to earn each contracted dollar.

Sources