Figma (FIG) Q2 2026 Update: The GAAP Loss Spooked the Market — But It's Not the Number That Matters

Revenue up 48% year-over-year, accelerating for a third straight quarter — yet the stock fell 15-16%. Over a third of the $117 million GAAP operating loss is stock-based compensation, and there's a more important date waiting at the end of August.

Chapter 1 | Earnings Checkup: Demand Beat Across the Board, the Problem Lies Elsewhere

There's no crack anywhere on the demand side: revenue accelerated for a third straight quarter, large-account growth kept accelerating, NDR held at 136%, and full-year guidance wasn't cut — it was raised. The problem is entirely in the lower half of the income statement — a GAAP operating-loss headline that overshadowed what is, above the line, a clean report card.

Chapter 2 | Why the Stock Fell 15-16%: The GAAP Headline vs. the Real Numbers

The market's gut reaction was simple: GAAP operating expenses exploded from $219.7 million a year ago to $426.9 million (+94%), flipping the company from a small profit to a $117.3 million loss — a contrast stark enough to overshadow 48% revenue acceleration. But "GAAP opex up 94%" and "the company is burning cash out of control" are not the same statement, and there's a critical breakdown in between.

CFO Praveer Melwani, verbatim, on the earnings call: "Even including the cost of Config, our Q2 operating expense growth was still below our revenue growth." That statement is about non-GAAP operating expenses, not GAAP. The gap between the two is mostly that $147.6 million of stock-based compensation — an accounting reality every recently-IPO'd company runs into: RSUs with their own vesting schedule stack on top of a fresh round of post-IPO equity awards, creating a huge but non-cash-consuming expense line. Non-GAAP operating income was still positive at $36 million; free cash flow was still positive at $53 million. Those two numbers are the ones that actually answer "is the company burning cash" — and the answer is no.

Chapter 3 | Gross Margin Under the Hood: AI Inference Costs Are a Choice, Not a Runaway

Beyond operating expenses, cost of revenue (COGS) also doubled to $60.5 million, pulling GAAP gross margin down from 89% a year ago to 84%. This line maps more directly to AI inference cost, and it's worth unpacking because the CFO gave a very specific explanation.

Figma Agent, the local-code version of Figma Make, Motion, generative plugins, Code Layers — all the new products unveiled at Config 2026 are still in beta. CFO, verbatim: "We're not currently charging customers for usage of these beta products, but we are bearing the inference cost without offsetting consumption revenue." In other words, this isn't cost running out of control — it's a deliberate invest-first, monetize-later playbook. The company is choosing to let customers try compute-heavy AI features for free first, build up usage, then phase in monetization. In the same quarter, non-GAAP gross margin actually rose 2.5 percentage points (250 basis points) sequentially to 85%, and gross profit dollar growth also accelerated 9 points sequentially — this is exactly the re-rating trigger the July article flagged as "AI credit monetizing across a full quarter," and this quarter delivered the closest thing yet to direct evidence: Q2 is the officially confirmed first full quarter in which "customers began paying for AI credit embedded in their plans," and as of end-June, over 80% of paying customers with $10K+ ARR were consuming AI credit weekly. To be candid about the limits here: the company hasn't disclosed a standalone revenue figure for AI credit itself, nor broken out what share it represents of the $370.1 million in total Q2 revenue — what's visible right now is the indirect evidence of "structurally improving gross margin plus high penetration," not a revenue number you can plug straight into a model. That's exactly why it stays on the Chapter 8 watch list rather than getting checked off as resolved.

Whether this strategy pays off ultimately comes down to conversion: can the beta products be successfully converted into paid features over the coming quarters. But the evidence visible today is that this is a choice with a timeline and a logic that management explained clearly — not an accidental cost blowup.

Two details from the Q&A portion of the earnings call are worth adding — neither appeared in the company's own press release or prepared remarks. First, the CFO explicitly stated that this quarter's $40 million full-year revenue guidance raise does not include any contribution from beta/early-access products (Figma Agent, Make on local code, etc.) — those products currently represent upside, not revenue already baked into guidance. In other words, if the conversion to paid goes well, there's room for guidance to move up again, and that upside isn't priced in yet. Second, when CEO Dylan Field was asked about crowded competition and the rise of vibe-coding platforms, he didn't name any specific competitor or push back directly — he simply reiterated the existing pitch of "performance, professional-grade canvas, and full creative control." That suggests the company currently prefers to win the argument through product differentiation rather than attacking rivals head-on — worth continuing to watch. Separately, when a Piper Sandler analyst asked about headcount, management acknowledged they've hired fewer people than originally planned, having used their own AI tools to boost team efficiency instead — the same discipline as the "invest first, monetize later" logic above: spending is targeted, not indiscriminate expansion.

Chapter 4 | Did July's Watch List Actually Deliver This Quarter?

The July update's watch items centered on whether the Config 2026 product lineup (Code Layers, Motion, Shaders, generative plugins, Figma Agent) could genuinely extend Figma's position as the "product-creation control plane," standing up to the stress test posed by AI prototyping tools like Claude Design. This quarter's official earnings call transcript delivers concrete, verifiable evidence of follow-through — not just product launches.

All five watch items now have concrete, verifiable customer cases or usage numbers behind them — not just management saying so. This means the cost increases visible in the Q2 report and the strategic direction set out in the July article are two sides of the same coin: the company is genuinely following through on what it said, with visible adoption evidence to show for it. That's also why the "invest first, monetize later" logic in Chapter 3 holds up: without real adoption numbers behind it, that investment logic would just be talk. Now there are three lines of evidence converging at once — MCP usage, Agent penetration, and customer case studies — confirming the direction is right.

Chapter 5 | Stress-Testing the Bear Case: Model Dependency, Claude Design, and "Overvalued" — Do Any of the Three Hold Up?

Post-earnings, the market split into two camps: one leaning bullish (BofA reinstated coverage at Buy, RBC raised its price target, JPMorgan and Piper Sandler started reframing AI from "threat" to "potential tailwind"); the other still insisting the structural risks haven't gone away. Three specific arguments from that camp deserve individual scrutiny rather than a blanket dismissal.

Model Dependency Risk: If LLMs Become Electricity, Who Gets to Charge for It?

One of the most common bearish arguments is that "Figma's AI features run on external foundation models from Anthropic, OpenAI, and the like — Figma doesn't own the underlying intelligence layer" — implying Figma's moat is built on someone else's land. The inverse of that argument is actually worth taking more seriously: if LLMs ultimately commoditize into general-purpose infrastructure, like electricity, who actually gets to charge for value? Utility companies don't capture Apple's margins, because electricity is an undifferentiated input — Apple owns product design, brand, and the customer relationship. Apply the same logic to the AI era: if the model layer gets competed down to near cost, value naturally migrates upstream to whoever owns the customer relationship, the proprietary data, and the workflow embedding — and that's exactly Figma's position, not the model vendors'.

More important, Figma laid out its own response strategy directly on the earnings call. CFO, verbatim: "We route across multiple models based on task complexity, optimize across vendors through a model-agnostic architecture, and have already started bringing a first-party model trained on Figma's own design data into specific design tasks inside Figma Agent." This kind of "don't lock into a single model, dynamically route across vendors" playbook already has dedicated infrastructure providers supporting it in the industry (Together AI, for instance, offers a unified multi-model API that routes traffic to different underlying models by task) — in other words, what Figma is describing isn't empty rhetoric, it's an architectural choice that already has mature infrastructure to back it up. This directly answers the "dependent on external models" concern: Figma has deliberately avoided locking into a single model vendor, and is already training a proprietary model on design data no general-purpose foundation model can replicate — a data moat. The observation that "the AI runs on someone else's models" isn't wrong, but the implication that "so Figma can't capture pricing power" doesn't square with the company's actual architectural choices.

Two Quarters of Claude Design: The Numbers Haven't Moved

Another frequently cited threat is Anthropic's Claude Design — capable of generating an interactive interface prototype from a single prompt in minutes, directly challenging the premise that "product design has to start in Figma." But this threat has now been in the market for more than two quarters, and it can be tested directly against Figma's own numbers: Q2 marks a third consecutive quarter of revenue acceleration, not deceleration. If Claude Design were genuinely pulling meaningful budget or use cases away from Figma, this stretch should have shown slowing growth, softening NDR, or a pullback in large-account expansion — none of the three has shown up. NDR sits at 136%, and $100K+ ARR customers grew 46% year-over-year — both accelerating, if anything. A threat existing isn't the same as a threat materializing; two quarters in, the evidence of it materializing still hasn't shown up in the numbers. As for Google, there's no comparable dedicated design-AI product visibly making a quantifiable dent in the market yet, so that line doesn't currently rise to a variable worth taking seriously.

"Overvalued": Does That Argument Still Hold After the Stock Was Cut in Half, Twice?

The Seeking Alpha piece titled "From High-Growth Darling to Overvaluation Risk" sounds like it's describing the present, but arguments like it were mostly written while the stock still traded near its IPO highs (closing as high as $122, up more than 270% from the $33 IPO price) — back when revenue was running at just over a billion dollars annualized, at a far higher valuation multiple than today. After the stock fell roughly 81% from its closing high, EV/Revenue now sits at 9.9x — while over the same stretch, the revenue growth rate hasn't slowed, it's accelerated from around the IPO period to today's 48%, a third consecutive quarterly high. Using a rough rule-of-thumb combined metric of "growth rate + free cash flow margin": 48% growth plus 14% FCF margin sums to 62 — a combination that, in typical high-growth software valuation frameworks, usually corresponds to a multiple well above 10x revenue, not where the stock sits now.

Chapter 6 | The Next Variable: VC Lockup Expires August 31

Figma went public on July 31, 2025, with lockups releasing in stages: an early release of 25% tied to performance in early September, another batch tied to Q3 earnings in November, and the standard 180-day lockup expiring January 27, 2026. VC shareholders holding roughly 54.1% of shares separately signed an extended lockup agreement, releasing shares in quarterly batches through June 2026. The final tranche lands on August 31, 2026 — at which point the venture shareholders holding a majority of shares will all be eligible to sell.

This is a variable entirely unrelated to this quarter's earnings quality, but its timing happens to land less than a month after the report. With the stock already down more than 80% from its post-IPO closing high, if VC shareholders choose to unload positions at this point, it would create direct supply-side pressure on the stock — unrelated to fundamentals, purely a share-structure issue.

Chapter 7 | Valuation: Down ~81% From the IPO High, Trading at 9.9x EV/Revenue

With the stock down more than 80% from its post-IPO high, the market's current pricing of Figma to some extent already reflects the impression of "a company still losing money" — but this report shows the company is profitable on a non-GAAP basis, free cash flow is positive, and full-year guidance is still being raised. If non-GAAP operating income and the beta-to-paid conversion continue validating over the next one to two quarters, there's room for the current valuation to recover; if the August 31 lockup expiration does trigger heavy selling pressure, near-term price action could decouple from fundamentals for a stretch.

Chapter 8 | Key Risks and What Would Break This Thesis

1. Whether beta products convert to paid features: Figma Agent, Make on local code, Motion, generative plugins, and Code Layers are all free today. If a sustainable path to monetization fails to materialize, the "invest first, monetize later" logic in Chapter 3 falls apart.
2. Actual selling pressure at the 8/31 lockup expiration: Whether VC shareholders holding 54.1% of shares choose to sell in concentrated fashion will directly affect near-term share-supply structure.
3. Whether non-GAAP operating income stays positive: Full-year guidance of $125-135 million hasn't been cut, but if future quarters fail to deliver on it, the "disciplined investment" call needs to be revisited.
4. Whether the GAAP-to-non-GAAP gap keeps narrowing: Full-year 2025 SBC once spiked to 129% of revenue, driven mainly by pre-IPO RSUs carrying "double-trigger" vesting conditions (requiring both a service period and a liquidity event before vesting), which meant almost none of it had to be recognized under GAAP before the IPO. Once the company actually went public in July 2025, years of accumulated expense hit the books all at once in Q3 2025 — $975.7 million recognized in a single quarter. That's a one-time accounting event tied to the exit, not the ongoing cash-burn rate. Strip out that base-period distortion and Q2 2026 SBC has already fallen to 39.9% of revenue — moving down, not continuing to deteriorate. What actually needs watching isn't "whether it spirals out of control" but whether it's converging fast enough — mature SaaS peers like Adobe run SBC in the single digits as a percentage of revenue, and how many quarters it takes Figma to approach that level is the thing to keep validating. Even as a non-cash item, it still dilutes shareholders over the long run and can't be ignored indefinitely.

Chapter 9 | Watch List and Conclusion

• The actual reaction in share price and volume after the 8/31 VC lockup expires.
• Whether Q3 earnings disclose beta products (Agent, Make on local code, Motion, Code Layers) starting to charge, or provide conversion-rate figures.
• Whether non-GAAP operating income and free cash flow stay positive, and whether full-year guidance holds.
• Whether MCP usage and Figma Agent penetration sustain this quarter's growth rate.
• Whether stock-based compensation as a percentage of revenue continues Q2's downward trend (129% → 39.9%), keeps converging toward the single-digit levels of mature peers like Adobe.

Sources