Canva AI Costs Cut a Third From Its 2026 Growth Forecast

Canva has cut its 2026 revenue growth forecast from roughly 30% to about 20%, a rare guidance reduction for a company built on profitable expansion. The trigger is Canva AI costs: the agentic upgrade released in April, Canva AI 2.0, pushed users to generate three times as many designs, and every one of those designs consumes inference compute. CEO Melanie Perkins indicated this week that a public listing is now more likely next year than in 2026, a timeline that puts the company's unit economics directly under investor scrutiny.

Canva revised its full-year outlook in early August after second-quarter revenue of $921.9 million, up 25.2% year over year, came in short of internal expectations. The miss is measured against a base of roughly $4 billion in annualized revenue and more than 200 million users, the scale that made Canva the rare venture-backed software company to pair fast growth with steady cash generation.

What the Guidance Cut Shows

The downgrade is not a demand problem. Usage of the AI suite exceeded every internal projection after Canva AI 2.0 shipped in April, and the resulting compute bill forced Canva to reset its near-term trajectory. The company chose to absorb the hit in its growth number rather than let costs run ahead of revenue.

The cut lands at a delicate moment in the freemium model that built Canva. AI features are effectively free at the point of use, so engagement cannot be rationed by price, and a hit tool can move a quarterly revenue line already approaching a billion dollars. Teams that ship popular AI features must be ready for the usage those features create.

Inference also changes the shape of software margins. The classic SaaS model carried near-zero marginal cost per additional user, which is what made software margins structurally high. Generative features reintroduce a real cost per action, and under a freemium model that cost lands before the user pays anything. That reversal is why AI spend now has to be modeled as cost of goods sold rather than a hosting line.

Why Canva AI Costs Kept Climbing

Two disclosed numbers frame the bind. Canva says cost per task has fallen by nearly 90% since April, the payoff from re-architected model routing that sends each job to the cheapest model that can handle it. Yet AI users now produce roughly three times as many designs, so the volume gain swamps the per-task savings and total inference spend stays high. Canva AI costs remained elevated precisely because the efficiency win and the usage surge compounded in opposite directions.

The agentic design of Canva AI 2.0 makes the math worse. An agentic feature can fire multiple model calls for a single user request, so the cost of each completed task is a chain of inferences rather than a single one. Product teams should therefore model spend as a function of usage per user, not cost per task alone.

Three Levers: Rollout Pace, Pricing, Model Choice

Canva has responded on all three fronts. New AI features are being rolled out more slowly, in part because the company is reworking the cost structure of existing ones before adding more surface area. Heavier workloads are shifting toward usage-based pricing. And Canva is testing smaller and in-house models as substitutes for frontier models wherever quality holds up.

Perkins has framed the slowdown as a deliberate trade. The pause cost near-term distribution and growth, but it produced what she describes as the most important technical work in Canva's history and left the company better positioned to scale AI on sustainable economics. The framing is a candid acknowledgment that product ambition and inference budgets are now negotiated against each other at the roadmap level.

The consequences for users are concrete. Slower rollouts mean the most expensive AI capabilities arrive late or in stages, and usage-based pricing means heavy workloads may carry per-use charges instead of being bundled into a flat plan. For a platform of more than 200 million users, each delay or new charge is also a retention decision, because a heavy user can move the same work to ChatGPT.

Figma and ChatGPT Frame a Sector Problem

Canva is not alone in absorbing this shock. Figma, the design rival that is now Canva's public-market benchmark, chose the opposite route and swallowed the full inference cost of its new AI tools instead of passing it to users. The result shows in its margins: free cash flow margin fell to 14% in the second quarter from 27% in the first, and its shares dropped 15% after the disclosure. Figma's third-quarter income growth guidance of 36% is a clear step down from the pace it had been posting.

The two strategies carry different price tags at different scales. Canva's AI rollout serves a base of more than 200 million users, while Figma's subsidy applies to a professional design tool with a much smaller audience. Absorbing inference costs is workable when the user base is narrow, and close to impossible when every point of engagement is multiplied across a mass market.

Competition adds a second pressure point. Some Canva users moved design work to ChatGPT after OpenAI improved its image generation, which means Canva faces rising costs on its own platform and price pressure from a rival that bundles design capability into a general-purpose subscription. Canva has chosen to make its AI cheaper to run before it tries to reaccelerate growth, a sequencing that puts unit economics ahead of near-term share gains.

The competition angle raises the cost of slowing down. Every AI feature Canva delays is a window for ChatGPT to absorb more design workflows, so the cost problem and the competitive problem pull in opposite directions. That tension is why the next two quarters matter: Canva has to prove it can cut inference cost without losing perceived product momentum.

IPO Timing Hangs on Unit Economics

The stakes are set by Canva's valuation trajectory. A 2025 employee share sale valued the company at $42 billion, and Perkins now expects a listing next year rather than in 2026. Public investors will judge Canva AI costs against that mark, asking whether features that triple engagement can still clear a gross-margin hurdle once inference is counted as cost of goods sold.

Figma's 15% share drop after its AI cost disclosure is a preview of that scrutiny: public markets now punish software companies that let inference spending surprise them. The ten-point haircut, from 30% to 20%, is a concrete signal that AI cost of goods sold has become an earnings risk for software at scale. Any consumer-scale AI-first company with a freemium base faces the same structural bind, and Canva's response is becoming the reference case for resolving it.

At roughly $4 billion in annualized revenue, a 20% growth year still adds on the order of $800 million in new revenue. The argument is no longer whether Canva grows, but what margin that growth carries once inference is priced in.

The practical takeaway for product and finance leaders is that inference spend now needs the same forecasting discipline as headcount. Track cost per task and cost per active AI user as first-class metrics. Gate feature rollouts on the unit economics of each feature. Decide early whether heavy workloads will be priced per use or subsidized to chase engagement. Canva is running that playbook now, and its next two quarters will show whether the 20% guidance holds.

Why this matters

Canva's cut turns AI from a feature story into a cost line, and the company's response is becoming the template for every AI-first software business facing the same math. If Canva rebuilds its unit economics and still takes the platform public, it validates a model where AI growth is paced by inference budgets. If it cannot, the $42 billion valuation and the IPO timeline will both come under renewed pressure.

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.