Figma bets on AI credits and in-house design model to lift margin

What Figma announced

In an interview with Reuters this Wednesday (November 4, 2026), Figma's Chief Financial Officer, Praveer Melwani, detailed how the company is monetizing its artificial intelligence features: through credits paid by customers and through an AI model developed in-house to run tasks within the design software, according to ET Tech.

The company raised its annual revenue forecast for 2026, from a range between $1.463 billion and $1.467 billion to a range between $1.497 billion and $1.499 billion. The expected adjusted operating margin rose from 9% to 11%, with adjusted operating income projected between $158 million and $163 million (up from $125 million to $135 million previously).

In the third quarter, revenue grew 43% year-over-year to $391.4 million, beating market expectations. Despite this, quarterly gross margin came in at 80%, slightly below the 81% forecast, and the company's shares fell 8% in after-hours trading.

How AI credit billing works

The revenue engine Figma highlights is AI credit consumption: as of October, more than 80% of the company's large paying customers were using credits weekly, according to Melwani. This recurring use is the indicator the company uses to justify monetizing features like Figma Agent, its AI-based design assistant.

The company also reworked package pricing, delivering more AI credits than customers originally paid for. The effect on consumption was direct:

We've seen utilization rate of those add-ons increase by 40% since we re-priced it.

Praveer Melwani, Figma's Chief Financial Officer

In other words, giving more credits for the same price did not reduce expected revenue: it increased engagement with AI tools, which is what the company monetizes at contract renewal and through paid add-ons.

The in-house model behind Figma Agent

The second front of the strategy is technical: Figma more than doubled the share of work on the platform processed by its proprietary model, post-trained specifically for design tasks, instead of relying solely on third-party models. According to Melwani:

We're increasingly serving a lot of our product with our first-party model over the course of the third quarter, and we're able to better serve customers with lower latency, higher quality and a more efficient price point.

Praveer Melwani, Figma's Chief Financial Officer

In practice, this describes a common move among SaaS companies that have embedded generative AI into their products over the past two years: relying on third-party models (such as those from major LLM providers) is expensive and variable in latency; training an in-house model, fine-tuned for the product's specific domain (here, generating and manipulating design components), makes it possible to cut cost per call and keep control over the experience.

What changes for people designing products in Brazil

For designers, PMs, and product teams using Figma daily, this news has three practical implications:

• AI budget becomes its own line item. If credit consumption keeps growing and the company is actively encouraging this use (with repricing that increased utilization by 40%), teams that already use Figma Agent to generate variants, prototypes, or specs should expect AI consumption to show up as a significant item in the tools budget, not as a marginal extra.
• Purchase decisions move up a level. When 80% of large paying customers already consume credits weekly, the choice of which AI add-ons to contract stops being an isolated decision for the designer and starts involving whoever approves the corporate contract, especially for teams negotiating their plan with Figma in Brazil.
• Quality and latency as an internal sales argument. Figma's claim that its in-house model delivers lower latency and higher quality is, for those deciding on product tools, a concrete criterion for comparing Figma Agent with alternatives that rely on generic third-party models plugged in via integration.

It's worth noting that the source does not detail how Figma Agent differs in the designer's workflow compared to competitors that also embed generative AI into interface tools; what the company disclosed was business metrics (revenue, margin, utilization), not a functional comparison.

What remains open

The 8% drop in shares in after-hours trading, despite revenue coming in above expectations, signals that the financial market is paying more attention to gross margin (which came in below estimate) than to revenue growth alone. This is an indirect signal for those following the design tools sector: the cost of running generative AI models, even proprietary ones, still pressures these companies' gross margin, despite the latency and pricing efficiency gains mentioned by Melwani.

ET Tech, citing Reuters, did not detail specific per-credit pricing figures or the exact size of the feature catalog covered by Figma Agent. The source also does not provide a direct cost comparison between the in-house model and the third-party models Figma previously used. These are points that should surface in upcoming quarterly results and any future product announcements from the company.