Figma Net Dollar Retention Fell to 136% in Its First Full Quarter of AI Credits

Figma's net dollar retention was 136% at 30 June 2026, down from 139% three months earlier. That was the first full quarter in which Figma billed for AI credits. If usage-based AI add-ons were a new expansion engine, this is the quarter it should have shown up in the retention line. It did not, and the filings explain why seats still carry Figma's expansion story.

The short answer. Figma's net dollar retention is 136% as of 30 June 2026, measured on customers above $10,000 in ARR. AI credit add-ons began contributing in Q2 2026, but the rate fell 3 points that quarter. Figma's own disclosures still point to Full seat expansion as the larger driver, with credits as a supporting line.

What Figma's net dollar retention actually measures

Net dollar retention, often shortened to NDR or NRR, compares what a fixed group of customers pays today with what the same group paid a year ago. A figure above 100% means existing customers grew their spend faster than churn and downgrades removed it.

Figma's version has two definitional choices that matter for any AI credit question. Both are written into the footnotes of its Q2 2026 earnings release, and both are easy to skip.

The $10,000 ARR threshold

Figma calculates the rate only on Paid Customers with more than $10,000 in ARR as of twelve months before the measurement date. At 30 June 2026 that group held 15,964 customers, growing 34% year on year. Figma had roughly 690,000 paid customers in total at the end of March, so the retention metric covers a little over 2% of accounts.

That is a defensible choice. The company says $10,000 is a strong indicator of significant paid usage. It also means the metric is blind to the small Professional teams, which is where the Q1 2026 release placed one of its sharpest AI credit claims.

ARR counts agreements, not consumption

Figma defines ARR as the annualised value of active customer agreements, with an agreement counted as active when seats are provisioned. The definition is written around seats and subscription terms. It does not mention usage.

This is my reading, not Figma's statement: a coterminous credit add-on subscription plausibly lands inside ARR, because it is an agreement with a fixed value. Pay-as-you-go overage plausibly does not. If that reading is right, the purest form of usage-based AI revenue is the part least likely to show up in net dollar retention at all.

Six quarters of Figma net dollar retention

The trend matters more than the single print. Figma's filings give six consecutive readings, and they describe a trough in mid 2025, a recovery into early 2026, and a small step back in the first full quarter of metered credits.

June 2025 customer counts are derived by dividing the June 2026 counts by Figma's stated growth rates of 34% and 46%, so treat them as approximate. All other figures are as reported. The March 2025 reading of 132% comes from the Q1 2026 10-Q comparative and is plotted in the chart only.

One number in that table deserves more attention than the retention rate. Net additions to the $10,000-plus group were about 951 in Q4 2025, 1,357 in Q1 2026 and 746 in Q2 2026. The Q2 figure is roughly 45% below Q1. Figma has not explained the slowdown, and one quarter does not make it a trend. It is still the opposite of what a strong new AI expansion lever would be expected to produce at the threshold.

How Figma meters AI credits by seat

Figma's AI meter is not a separate product. It is an allowance bundled into every paid seat, with overage sold as add-on subscriptions or pay-as-you-go. The CFO told analysts on the Q2 call that "starting in mid-March, we implemented credit limits on all seats." The Q1 release dates that change to 18 March 2026.

Prices and allowances below come from Figma's pricing page and help centre, read on 2 November 2026. The pricing for each credit tier is pulled apart in the comparison of Figma, Adobe and Canva AI credit meters, so this section only covers what the allowance design does to expansion.

Organization and Enterprise prices are billed annually. The Professional page shows a monthly and annual toggle, and the extracted figures did not state which applies. The last column is Zan Digital arithmetic, rounded.

Dev Mode seats get the thinnest allowance

A Dev seat includes Dev Mode, FigJam and Figma Slides, plus view and comment access in Design files. It carries 500 AI credits a month on every plan. A Full seat carries six to eight and a half times that.

I think this is the most deliberate choice in Figma's whole meter. Engineers are the users most likely to run code-heavy AI workflows, and Figma has pointed Make at production codebases since May 2026. Give them the smallest allowance and the first credit wall arrives in the engineering org. That converts AI enthusiasm into either a Full seat upgrade or an add-on purchase, and both show up as expansion.

Beta products burn credits Figma pays for

The largest new AI surfaces are not metered yet. On the Q2 call the CFO listed the Figma agent, Make on local code, Motion, generative plugins and Code Layers as products that "do not currently consume paid credits." He added that Figma bears the cost of inference without offsetting consumption revenue.

Adoption of those surfaces is already wide. Figma said that as of 31 July 2026 more than 50% of customers above $10,000 in ARR used the Figma agent weekly. When those products start drawing on paid credits, the retention line gets a real test. Until then, Q2 tells you how a partial meter behaves, not a full one.

Seats or credits: what moved Figma net dollar retention

Figma does not split retention into a seat component and a credit component. It discloses enough adjacent facts to weigh the two, and the weight falls on seats.

The seat evidence

The CFO said on the Q2 call that approximately two thirds of customers above $10,000 in ARR added Full seats compared with their prior renewal, "consistent with prior quarters." The Q1 release credited its outperformance to "stronger than expected seat expansion across entire organizations."

That is a seat story told in Figma's own words. A consistent two thirds adding Full seats is a stable engine. It does not need AI credits to explain a retention rate in the mid 130s, because a large share of the base is simply buying more seats every year.

The credit evidence

The credit facts are real but softer. Over 80% of customers above $10,000 in ARR were consuming AI credits weekly at the end of Q2. The Q1 release said over 75% of Org and Enterprise users who had exceeded credit limits kept using credits in April 2026, and over 95% of them stayed active.

The strongest monetisation claim is from Q1. Pro teams that bought AI credit add-ons had average ARR more than three times that of teams that had not. I would not read that as credits causing expansion. Bigger teams use more AI and run out first. The add-on is a marker of size as much as a driver of it.

The CFO's Q2 summary sits between the two. Retention "remained strong at 136% as customers expanded both seats and AI credit add-ons." That sentence names both levers and quantifies neither. The broader pattern of credit add-ons sitting on top of seats, rather than replacing them, is the subject of the analysis of hybrid pricing and the retention gap.

What AI credits cost Figma's gross margin

If AI credits are not yet visible in retention, they are very visible in cost. The Q2 2026 10-Q reports cost of revenue of $60.5 million, against $27.9 million a year earlier. That is a 117% increase while revenue grew 48%.

Gross profit grew 40%, so the AI build is already compressing the ratio. Non-GAAP gross margin was 90% in Q2 2025, 86% in both Q3 and Q4 2025, 82% in Q1 2026 and 85% in Q2 2026.

The Q2 recovery is the interesting part. The CFO attributed it to "our first full quarter of AI credit monetization." That is the clearest evidence in the filings that credits are doing a job. Their first measurable effect is to recover some of the inference cost Figma was absorbing, not to lift what customers pay in aggregate.

That framing matters for anyone modelling a credit meter of their own. A credit add-on that pays for inference is a margin defence. Calling it an expansion engine before the retention line moves is a category error. The wider pattern of AI features eating software gross margin is laid out in the breakdown of AI cost of goods and real gross margin.

The Q3 guide implies almost no sequential growth

Figma reports Q3 2026 results on 4 November 2026. Its guidance for the quarter is revenue of $373 million to $375 million. At the midpoint that is about $3.9 million above the $370.1 million it reported for Q2.

Sequential growth over the previous three quarters was $29.6 million, $29.6 million and $36.7 million. A guide that implies roughly a tenth of that pace is either very conservative or signals something real.

The conservative reading has history behind it. Figma guided Q1 2026 to $315 million to $317 million and reported $333.4 million, about $17.4 million above the midpoint. It guided Q2 to $348 million to $350 million and reported $370.1 million, about $21.1 million above. A beat of the same size in Q3 would put revenue near $395 million. That is arithmetic on past behaviour, not a forecast.

The cautious reading is also available. Credit consumption is less predictable than seat billing, and Q2 was the first full quarter of credit revenue. A finance team that cannot yet forecast usage will guide to the seat base and let credits surprise. I lean towards this reading, and it implies credits are now material enough to make guidance harder. The same pattern across other software filers is tracked in the SaaS earnings guidance scorecard.

Where this reading of Figma's retention is weakest

The argument above says credits have not yet lifted Figma net dollar retention. There are two strong reasons it may be wrong, and a reader should weigh both before repeating it.

The pricing anniversary pulls the rate down mechanically

The CFO said the 136% held "even as we begin to anniversary our pricing and packaging changes from March 2025." Retention compares ARR with a year earlier. Once a price change is a year old, it sits on both sides of the ratio and stops inflating the result. A 3-point dip in the quarter that laps a repricing is close to what you would expect with no credit effect at all.

Read that way, credits may have offset a larger fall. I could not open Figma's help article describing the March 2025 changes, so I cannot size the anniversary effect. That gap is the single biggest hole in this post, and it cuts against my conclusion.

One quarter and an unclear definition

Q2 2026 is one data point. The meter was live for one full quarter, the beta products were free, and user-level limits only began rolling out in August. Any conclusion from a single quarter of a partial meter is provisional.

My point about pay-as-you-go sitting outside ARR is an inference from the definition, not a Figma statement. If Figma includes usage in ARR in a way the footnote does not describe, the metric sees more of the credit story than I have assumed. Watch the Q3 call for a direct answer.

What a SaaS founder should copy from Figma's credit design

You are probably not pricing a design tool. You may well be deciding whether to bundle AI usage into seats or sell it on a separate meter. Figma's filings offer three lessons that transfer.

First, bundle an allowance into the seat and make it unequal by role. Figma's 500-credit Dev seat is a quiet upsell path. It costs the light user nothing and turns the heavy user into an add-on buyer. That beats a flat allowance, which either gives margin away to light users or frustrates heavy ones.

Second, do not expect the meter to show up in net dollar retention quickly. If your ARR definition counts agreements, usage overage may never appear in it. Decide now whether you want credit add-ons to be coterminous subscriptions, which land in ARR, or pure pay-as-you-go, which may not. Figma's CFO described most current structures as add-ons rather than pure consumption.

Third, report the cost side honestly. Figma's margin moved from 90% to 82% and back to 85% within a year. If your board sees a credit meter only as upside, the first quarter of unbilled beta inference will be an unpleasant surprise. The wider shift from seats to consumption across software vendors is mapped in the analysis of seat compression and SaaS pricing.

Frequently asked questions

Where to start before 4 November

If you track Figma, read three lines in the Q3 release and ignore the rest. The retention rate, against 136%. Net additions to the $10,000-plus group, against 746. And any sentence that separates credit add-on ARR from seat ARR, which would settle the question this post cannot.

If you run a SaaS business with an AI meter, pull your own retention cohort and split the expansion into seat changes, price changes and credit purchases. If you cannot produce that split, Figma's investors are asking the same question of Figma, and yours will ask it of you.

Related reading