Figma beat estimates, then fell as AI credits became the story

What Figma actually reported

Figma reported adjusted earnings per share of $0.10 for the third quarter of fiscal 2026, clearing the consensus estimate of $0.05 by 83%. Revenue of $391.4 million grew 42.8% year over year and came in $16.4 million above the high end of prior guidance.

Those are strong numbers for a company that listed recently and still trades far below its early highs. On the headline, the design software maker beat on both lines, and it also lifted its full-year revenue outlook by $33 million to roughly $1.5 billion, which is why the negative share reaction needs explaining.

Why the stock fell anyway

Figma shares fell about 7% in extended trading after the report. The company sees fourth-quarter revenue growth slowing to about 33%. Analysts polled by LSEG had predicted 28% growth, so the guidance was not a miss, but it described a decelerating business rather than an accelerating one.

The other reason is cost. Figma is spending more on AI to keep its product competitive, and it told investors that the near-term margin trade is deliberate. In a market that has spent a year rewarding AI revenue and punishing AI spending, a beat with softer growth is an awkward combination. The wider AI earnings picture has looked similar all year: revenue growth is real, but the cost of serving it now shows up on every income statement.

The pattern is familiar from the rest of enterprise software this year: AI features lift engagement and revenue, and then the infrastructure bill arrives a quarter later, forcing companies to choose between margin and momentum.

AI credits as the engine

The clearest driver behind the quarter was AI credit consumption. By the end of October, more than 80% of customers spending over $10,000 annually were actively consuming AI credits, and weekly active users in that cohort rose nearly 50% between July and October.

Net dollar retention of 131% shows the same pattern from another angle: existing customers spent more, not just new ones. Figma ended the quarter with 16,566 paid customers above the $10,000 annual recurring revenue threshold, up more than 3,650 year over year, while customers above $100,000 in ARR reached 1,717. Investors have been pricing that arithmetic for AI vendors all year, most visibly in fal's inference valuation.

The trade Figma chose

During the quarter, the company started giving more AI credits to clients while maintaining prices. That is a deliberate trade: Figma gives away some near-term revenue to keep customers using the AI features that make the product stickier, then charges for the usage once it is part of the workflow.

Figma now anticipates an 11% operating margin, up from 9% as of August, and $1.497 billion to $1.499 billion in revenue, compared with $1.463 billion to $1.467 billion before. Free cash flow was $41.5 million, but the margin slipped to 11% from 18% a year earlier.

That trade is visible in the free cash flow line. Figma still generates cash, but the margin compression tells you the AI bill is real, and the company is choosing to absorb it now rather than pass it to customers who might churn if the price moved.

The in-house model bet

Figma is also building its own AI model rather than licensing all of it. Management framed that as the core bet of the next phase, pushing the company into territory that Adobe and a crowd of AI-native design tools also want, and it is the clearest signal that Figma sees model access as a cost centre it wants to own. The same question about how much of the stack a company should own has followed Anthropic's enterprise revenue through its own reporting.

The number of weekly users at large customers who used AI credits rose almost 50% during the quarter, Figma said, and the share of work its own model performs doubled. That is the metric to watch: if the in-house model carries more of the workload, the cost per AI action falls and the margin story improves.

Security joins the pitch

Figma is increasing its focus on cybersecurity, co-founder and chief executive Dylan Field said. In recent months, AI agents have broken into systems at Hugging Face and at other companies, and buyers now ask vendors how those agents are contained before they sign.

Clients are not asking Figma to become a security vendor, but the company must scrutinise the security of its own vendors, Field said. For a design platform that increasingly runs AI on customer files, that is a procurement question as much as an engineering one.

The security question also explains why buyers care about which model runs where. If a design platform sends customer files to a third-party model, the vendor list becomes part of the customer's own risk register.

What is established and what is not

Established: Figma beat on revenue and earnings, raised full-year guidance, guided fourth-quarter growth to about 33%, and reported that AI credit usage is spreading through its largest customers. Established too: free cash flow margin fell to 11% from 18% as AI spending rose.

Not established: whether AI credits become a durable margin engine or a discount that never reprices, how much of the workload the in-house model can take, and whether slower growth is a one-quarter effect. The stock reaction says investors want those answers before they pay for the AI story again.

The next two quarters will settle the argument. If AI credit consumption keeps rising while margins recover, the bulls have their case; if the credits keep growing and margins do not, the market will treat that as a discount that never repays.