Figma CFO: $391M revenue, 43% growth, and 1M weekly MCP users in Q3

Figma posted $391M in revenue for Q3, up 43% year over year, with net dollar retention holding above 130% — at 131% specifically. For a company that came through a failed Adobe acquisition and has been rebuilding its IPO narrative, the numbers are clean.

AI product traction

The more interesting story is the velocity on AI. Figma's agent product went GA on October 6, and credit consumption grew 2.5x through the end of October. The MCP server crossed 1 million weekly active users, with overall MCP usage up 3x and write-back usage to the canvas up 4x. Notably, 5% of all credit consumption in the quarter was orchestrated via MCP — meaning external agents are actively kicking off first-party Figma workflows, not just reading context.

Melwani frames MCP not as a threat but as an acquisition channel. The data backs that up: MCP usage is correlating with higher retention and account expansion, and it's driving new paid seat conversions by giving non-designers — PMs, marketers, researchers — enough reason to hold a full seat rather than a viewer license.

“We put up $391,000,000 in revenue. We grew the business 43% year over year. Our net dollar retention rate's at a 131%. Our MCP server is up to about just over a million weekly active users at this point — we grew it overall usage three x and overall write back usage to the canvas about four x.”

First-party model

Figma is building its own model and doubled the share of actions powered by it over the quarter. The strategic logic is tight: by owning the file system and Figma's design architecture, the company can train a smaller, lower-latency model that beats general-purpose alternatives on speed and cost for in-canvas tasks. The goal is to make first-party AI workflows fast enough and cheap enough that users default to them over routing through an external agent.

Token budgeting and cost efficiency

Figma raised its full-year operating profit guidance by $30M, partly by getting more efficient with AI inference. Some engineers had been spending tens of thousands of dollars on tasks they're now running for a fraction of that cost by switching model types and sharing learnings internally. Figma also hired Casper Sorenson from Coinbase as CIO to systematize that discipline — a hire Melwani flags as directly relevant to AI spend management.

On the demand side, Figma doubled credit allotments on Pro plans and increased allotments on organization plans by 60%, betting that higher limits would drive consumption. Melwani invokes Jevons' paradox by name — and says it played out as expected.

The business is growing faster than it was 12 months ago, its AI products are compounding rather than cannibalizing seat revenue, and the MCP bet looks like a distribution play that's working. The outstanding question is what happens to margin as credit consumption scales — but the $30M guidance raise suggests Figma is managing that curve more carefully than most.