Box, Inc.

Q2 revenue was $321 million, up 9% year over year and 11% in constant currency. Q2 gross margin was 81.2%; operating income was $95 million and operating margin was 29.4%, up 90 basis points from the prior year period and above guidance; EPS was $0.40, above guidance of $0.39. Billings were $310 million, up 17% year over year (16% constant currency), RPO was $1.7 billion, up 15% year over year (17% constant currency), and net retention was 106% versus 103% a year ago. For Q3, Box guided to revenue of about $329 million, growth of about 9% year over year (11% constant currency), gross margin of about 80.5%, operating margin of about 28%, and EPS of about $0.39. For FY27, Box raised revenue guidance by $10 million to about $1.29 billion, representing 10% year-over-year growth (11% constant currency), and reiterated gross margin of about 80.5%, operating margin of about 28%, and EPS of about $1.54.

Aaron Levie framed Box as being at the center of enterprise AI transformation, emphasizing secure, governed access to unstructured data for AI agents and model-neutral workflows. He said the company is seeing stronger demand for Enterprise Advanced, AI units, and migrations off legacy on-prem systems as customers want to bring more content and workflows into a single secure platform. His tone was confident and expansive, with repeated references to “massive opportunity,” customer momentum, and a heavy upcoming product pipeline, including BoxWorks announcements.

Dylan Smith focused on the clean execution in Q2 and the underlying demand indicators: $321 million of revenue, $310 million of billings, $1.7 billion of RPO, 81.2% gross margin, 29.4% operating margin, and $0.40 EPS. He noted free cash flow of $60 million and cash from operations of $71 million, with $446 million in cash and investments at quarter-end, and said Box repurchased 2.6 million shares for about $66 million. On guidance, he raised FY27 revenue by $10 million to about $1.29 billion, kept full-year operating margin at about 28%, and said the share count outlook falls to about 141 million, which he described as a significant reduction from 149 million in the prior year.

Analysts focused on token costs, open-source model adoption, legacy migrations, security/agent governance, wall-to-wall deployments, zero data retention, FDE expansion, consumption trends, and billings phasing. Management said longer agentic workflows are token-intensive, which makes Box’s model-neutral layer valuable because it can route workloads to the cheapest model that still meets accuracy needs; they also said rising AI competition can lower costs or improve consumption. On security, Aaron said agent governance requires new guardrails, alerting, anomaly detection, and identity controls beyond traditional human-user controls. On billings, Dylan said Q3’s apparent step-down is mostly quarter-to-quarter variability, FX, and comparables, not a change in underlying demand.