Paper announced a $34 million Series A on Wednesday, led by Accel and ICONIQ, with Designer Fund, Michael Grinich of WorkOS, Anton Osika of Lovable, and individual engineers and designers from Anthropic and OpenAI participating. Thirty-four million dollars is a rounding error in the current funding environment. Figma booked $333.4 million of revenue in the first quarter of 2026 alone. And yet Figma’s equity has lost roughly half its value over six months, trading below $22 against a post-IPO peak above $115, on precisely the thesis this round exists to fund.
That asymmetry is the story. The market has already repriced Figma for AI-native disruption. What it has not yet done is watch the disruption arrive with a balance sheet, a customer list, and named tier-one sponsors. This is the first time the abstraction has a company attached to it.
The technical claim is the part worth reading closely, because it is not a feature difference. Paper renders with HTML and CSS rather than a proprietary canvas. Founder and CEO Stephen Haney frames the goal as letting designers work seamlessly with coding agents like Claude Code and Codex, and Accel’s Dan Levine describes it as a new category built for a world where humans and AI create together rather than one where humans do all the work. ICONIQ’s Mariano Payano called it the first canvas built for where software is going. Strip the venture language and the architectural bet is simple: if agents are increasingly the thing that writes production code, then the design artifact should be something an agent can read natively, which means the web platform, not a proprietary file format.
That bet aims directly at the least-discussed part of Figma’s moat. The conventional description of Figma’s advantage is multiplayer collaboration and the browser-native canvas, and both are real. But the durable asset is the file. Figma files become the design system of record for an organization — the components, the tokens, the version history, the review threads, the institutional memory of every decision a product team made. Procurement inertia does the rest. A tool that renders to HTML and CSS does not try to build a better proprietary file; it argues the proprietary file should not exist. If design output is just code, there is nothing to be locked into.
Which is also, and this is the part the funding announcement will not tell you, the structural weakness of Paper’s own position. A moat built on the absence of lock-in is not a moat. If the entire proposition is that designs are HTML and CSS that agents can consume, then Vercel can ship it, Adobe can ship it, Anthropic’s Claude Design already operates in adjacent territory, and Figma itself can ship it — it has Dev Mode, a Model Context Protocol server, and Figma Make already pointed at the design-to-code seam. Paper is competing on taste, speed, and the quality of the agent workflow, not on switching costs. Those are real competitive assets and they are the ones that historically decay fastest under well-funded imitation.
The traction figures deserve the same skepticism. Paper reports ARR up 25x since the early-2026 launch of Paper Desktop and four consecutive months on Ramp’s fastest-growing-companies list. Twenty-five times from an undisclosed base is the standard construction for a number that is small in absolute terms; a company with meaningful revenue states the revenue. The customer list — Ramp, Lovable, Vercel, PostHog, Quartr, Y Combinator — is genuinely impressive and genuinely unrepresentative. These are San Francisco, engineering-led, agent-forward companies whose entire cultural disposition is to adopt the newest tool. They are the leading indicator for developer software and they are approximately zero percent of the enterprise seat base that generates Figma’s revenue. Figma’s customers producing more than $100,000 of annual recurring revenue grew 48% year over year. Those buyers move on procurement cycles, not on Guillermo Rauch endorsements.
So the fair reading of this round is that it is a strong signal about direction and a weak signal about timing, and Figma’s stock has been trading as though both signals were strong.
The numbers underneath the FIG selloff are the argument for that. First quarter 2026 revenue grew 46% year over year to $333.4 million. Non-GAAP EPS of $0.10 beat the $0.06 consensus. Net dollar retention reached 139%, the strongest reading in more than two years, and management raised full-year revenue guidance by $55 million to roughly $1.425 billion, implying about 35% growth against $1.056 billion in fiscal 2025. Seventy-five percent of enterprise customers who exceeded their AI credit limits bought more credits, and over 95% stayed active. None of that describes a business being disintermediated. It describes a business whose customers are paying more for the very capability that is supposed to kill it.
The stock fell 44% in the first half anyway, with more than 20% of that coming in June alone on AI-competition fear rather than on any operating datapoint, and Findell Capital has been publicly agitating. The shares rebounded from a 52-week low of $16.60 to roughly $25 earlier this month before sliding back below $22 as the same fear resurfaced.
Two things are legitimately deteriorating and they are worth separating from the narrative. Non-GAAP gross margin fell to 82% in the first quarter from the high-80s in late 2025, because AI inference costs money and Figma is serving more of it. Consensus 2026 EPS sits at $0.23, a 23% year-over-year decline. Figma is buying its AI defense with margin, which is the correct decision and an expensive one. That is a real, quantified cost of competition — and notably, it is showing up in the P&L a full year before any AI-native competitor shows up in the revenue line.
On valuation, the relevant anchor for a GAAP-unprofitable compounder is forward enterprise value to sales rather than earnings. Bank of America values Figma at roughly 8 times estimated 2027 EV/sales against a peer average near 5.9 times, justifying the premium on growth of about 35.6% in 2026 and 23% in 2027 versus peers around 19%. Street targets have compressed but not collapsed: Citi at $36, Bank of America initiating at $30, Piper Sandler cut to $30 from $35 at Overweight, Stifel to $25 from $30 at Hold. Against a low-$20s print, that is a market that thinks the company is fine and the multiple is not.
Base case for FIG, $20 to $28. Growth stays in the mid-30s, AI credit monetization continues converting, and the stock chops inside the de-rated range while the market waits for evidence either way. The catalyst is the August 14 earnings report. Bull case, $35 to $40, requires net dollar retention holding near 139% alongside re-accelerating $100,000-plus customer counts, which would demonstrate that AI expands the seat base rather than replacing it — the Bank of America thesis, that more people creating digital products means more demand for the place they collaborate. Bear case, $14 to $17, is a retest of the low, and it is cohort derating rather than a company-specific break: unprofitable software has been the worst-positioned asset class of this tape, and a stock that fell 20% in a month on a competitor’s press cycle will fall further on the next one. The company-specific accelerant would be gross margin dropping below 80% while growth decelerates, which is the combination that turns a premium multiple into a discount very quickly.
For Paper itself, the honest assessment is that $34 million from Accel and ICONIQ buys roughly two years of runway to prove that agent-native design is a category rather than a feature. The distinction matters enormously and will be settled by whether anyone other than the San Francisco early-adopter cohort buys it.
Watch one number on August 14. Figma’s net dollar retention was 139%. Existing customers spending 39% more year over year is definitionally incompatible with being displaced. If that figure holds above 130%, the market has repriced the equity for a disruption that has not started. If it breaks below, Paper’s $34 million was early money on the right thesis.
