Figma has made an unusually explicit trade: it is willing to accept lower gross margins while it expands AI adoption. Chief Financial Officer Praveer Melwani told Reuters on October 15 that a return to the company’s historical low-90% gross margin is uncertain—and not the metric he would prioritize if lower margins help Figma reach a much larger market.
The disclosure turns a familiar AI strategy claim into a finance problem that can be measured. Figma’s gross margin was 84% on a GAAP basis and 85% on a non-GAAP basis in the second quarter of 2026, while revenue rose 48% year over year. The company now sells extra AI credits through subscriptions or pay-as-you-go billing. That combination creates the beginnings of a unit-economics test: can paid usage and customer expansion eventually outgrow the inference and infrastructure costs that are compressing margin?
The judgment
- Implication: Lower gross margin can be a rational investment when it buys durable adoption and pricing power, but only if finance can separate strategic customer acquisition from an open-ended subsidy.
- What would change the conclusion: Evidence that paid AI-credit usage, retention and gross profit per active AI user are rising faster than cost to serve would strengthen the case. Weak conversion to paid credits or persistent margin erosion without customer expansion would weaken it.
- Management action: Report AI unit economics as a cohort: included credits, paid-credit conversion, gross profit per active user, retention and payback. Do not evaluate the strategy through consolidated gross margin alone.
The margin trade is now explicit
Reuters reported that Figma’s quarterly gross margins have moved into the mid-80% range from above 90% historically as the company spends on AI assistants and prompt-driven prototyping and development tools. Melwani’s stated objective is not to preserve the old margin at any cost; it is to use AI to attract more users, widen the addressable market and create room for future pricing.
That framing matters because gross margin is not simply a score. It is a summary of product architecture, pricing, customer behavior and cost allocation. A software company can protect the percentage temporarily by limiting an expensive feature. It can also sacrifice margin to create a larger, more valuable product surface. The latter can be sensible, but the word “investment” does not make the cost self-justifying.
Figma’s official second-quarter results show both sides of the argument. Revenue reached $370.1 million, up 48% from a year earlier. GAAP gross profit was $309.6 million and non-GAAP gross profit was $314.0 million, producing 84% and 85% gross margins, respectively. Growth remained strong even as the mix shifted toward more compute-intensive products.
Usage pricing makes the subsidy measurable
Figma’s credit model gives management a way to connect AI consumption to revenue. Each seat includes a monthly credit allotment; customers that exceed it can buy a recurring credit subscription or use pay-as-you-go billing subject to a spending limit. Enterprise customers can allocate credits across teams or cost centers and obtain usage data through an API.
For finance, that design is more important than the label “usage-based pricing.” It allows the company to observe where consumption occurs, which customers pay beyond the included allowance and whether heavier usage improves retention or expansion. It also lets customers treat AI consumption as a controlled operating expense rather than an invisible feature embedded in a seat price.
The embedded allowance still acts like a subsidy. That may be appropriate during adoption, especially when users need time to change workflows. The discipline is to define what the subsidy is expected to buy and when management will reassess it. If free or included credits create durable habits that convert to paid usage, the margin compression may have a measurable return. If they merely raise cost without changing retention, expansion or willingness to pay, the program is not an investment; it is a discount with an expensive delivery mechanism.
A CFO scorecard for AI unit economics
Figma has not published the detailed unit economics needed to resolve that question. The following is Numbers & Judgment analysis—not company guidance—but it describes the evidence a finance team should ask for:
- Cost per unit of useful consumption. Track model, infrastructure and support cost per credit or per completed workflow, not just total AI spend.
- Conversion from included to paid usage. Measure the share of active users and accounts that buy incremental credits after consuming the included allowance.
- Gross profit by cohort. Compare customers by adoption date, product mix and usage intensity. A consolidated margin can hide improving economics in newer cohorts or deteriorating economics in the heaviest users.
- Retention and expansion. Test whether AI-active customers renew, add seats or buy more products at higher rates than comparable non-users.
- Payback on the margin concession. Estimate how long incremental gross profit takes to recover the cost of included credits, onboarding and product support. The threshold should reflect the company’s capital priorities and uncertainty, not an invented industry benchmark.
This scorecard also prevents a common analytical error: treating every percentage point of gross margin decline as equally bad. A temporary decline caused by rapidly growing, high-retention usage is different from a structural decline caused by weak pricing power. Both appear in the same headline percentage.
Pricing power must be earned
Management’s expectation that the company can raise prices later is plausible, but it is not yet proof. Pricing power usually follows demonstrated workflow value, switching costs or scarce capability. It is harder to claim when customers can compare outputs across competing models and tools or when AI features remain easy to substitute.
The strongest version of Figma’s strategy is therefore not “accept lower margins now and raise prices later.” It is “use a controlled allowance to increase adoption, learn the cost curve, convert valuable usage to paid credits and retain enough of the productivity gain to rebuild gross profit.” Each step should leave evidence.
For finance leaders evaluating their own AI portfolios, the lesson is broader. The right question is not whether an AI initiative lowers a margin this quarter. It is whether the company can trace that lower margin to a customer behavior that has a credible path to gross profit. The same discipline applies to internal automation: our free AI Automation ROI Calculator helps separate labor savings, implementation cost and recurring model expense.
Related analysis: AI Scaling Falls Short of the Business Case and AI ROI Is a Finance Problem, Not a Technology Problem.
Leave a Reply