Figma and Canva Hit AI Cost Walls as Design Giants Rethink Their Generative Bets

Design software leaders Figma and Canva poured resources into artificial intelligence features over the past year. They promised faster creation, smarter suggestions and broader accessibility. Yet fresh data shows the pivot carries real trade-offs. Both companies now face higher expenses than anticipated. Growth projections have come down. And users sometimes bypass the new tools for simpler alternatives like ChatGPT.

The signals emerged clearly in recent weeks. On Thursday, Figma told investors it expects revenue growth for the September quarter to ease to 36 percent from 48 percent in the prior period. The Information reported the figure alongside similar pressures at Canva. The Information highlighted how popular AI capabilities can bring painful economics. Canva slowed its rollout of certain generative functions to manage expenses. Figma, meanwhile, continues heavy investment in agentic tools that let AI models edit directly on its canvas.

But the numbers tell a complicated story. Canva reached $4 billion in annual recurring revenue by the end of 2025 with 265 million monthly active users. The company added more than 31 million paid subscribers that year. TechCrunch noted the surge came partly from AI adoption. Co-founder and COO Cliff Obrecht described the platform as evolving toward a “design agency in your pocket.” Yet by mid-2026, Canva trimmed its full-year growth target to about 20 percent from an earlier 30 percent goal. The Australian reported the revision after the firm conceded it could not afford to run its AI suite at scale for a user base approaching one billion monthly visitors. Average cost per AI task ran higher than planned. Some products announced in April now sit up to six months behind schedule.

Figma posted stronger topline figures. The company generated $1.056 billion in revenue for fiscal 2025, up 41 percent year over year. First-quarter 2026 sales hit $333.4 million, a 46 percent increase. Bloomberg noted Figma topped analyst estimates in May and touted early traction from paid AI add-ons. Still, the September slowdown suggests usage patterns and infrastructure costs create friction even for the more enterprise-focused player. Net dollar retention stood at 136 percent in one recent period, but that metric may face pressure if AI features drive disproportionate compute demands without matching price increases.

Both organizations built much of their initial AI on third-party frontier models. That decision delivered quick capabilities. It also exposed them to volatile pricing and unpredictable demand. X posts reacting to The Information’s reporting captured the dynamic. One user observed that “a lot of companies are about to go through a similar arch: new ai features are too costly to run on frontier models.” The poster predicted a shift toward training custom models and smarter routing to cheaper alternatives. Another noted Canva users opting out of its tools in favor of GPT-generated images. The pattern repeats across the industry. Heavy usage spikes when features launch. Bills arrive later.

Figma responded with deeper platform integration. It acquired Weavy to bring generative editing into the open canvas. The firm launched Figma Make, an AI coding companion powered by Anthropic’s Claude. Users write natural language prompts to generate prototypes, apps or iterations. TechCrunch reported the tool now works across more subscription tiers, though full publishing rights remain limited to higher plans. In June the company added code layers, motion support, shader effects and custom AI skills. Its 2026 AI Report, drawn from over 8,400 survey responses and 639 interviews, identified four adoption patterns: grassroots, directive, nascent and unified. Researchers Shane Johnston and Christa Simon presented findings showing 41 percent of respondents believe AI now meaningfully changes team collaboration. Yet friction persists. A principal product designer quoted in Figma’s materials said, “Ultimately it’s more of an organizational gap, exposing some of our communication and knowledge gaps across squads.”

Canva took a different route. It developed its own design model and released AI 2.0 with persistent memory that learns user styles over time. The Verge described an orchestration layer allowing access to the entire tool set through one conversational interface. Prompt-based editing lets creators adjust work by describing changes. The company also introduced mini-app creation and 3D object generation. These moves helped drive user growth. But the expense side proved stubborn. AI Weekly summarized that the AI suite “proved far more expensive to run than management planned.” Some features now roll out more gradually. Others receive cost optimizations before wider distribution.

The competitive field grew crowded. New specialized tools target pain points where Figma and Canva still struggle. A Medium analysis listed seven AI design entrants gaining traction in 2026. The global AI-powered design tools market reached $8.22 billion this year and heads toward $18.16 billion by 2030. Much of that expansion comes from point solutions for brand asset consistency, rapid wireframing or non-designer workflows. Figma countered by opening its canvas to external AI agents. Models can now read design systems, apply variables and edit components without producing generic output. Muz.li quoted the shift as addressing the core issue: “AI-generated UI still feels generic, detached from the conventions teams spend months building.”

Adobe provides a contrast. The established player raised its own growth outlook on stronger Firefly adoption. It avoids subsidizing AI features to the same degree. X commentary pointed out the difference. One account noted Adobe’s approach while questioning whether Canva can truly disrupt the incumbent. Figma, for its part, positions itself as the hub for product teams rather than a general creative suite. One-third of its users are developers. Features like Dev Mode and MCP server improvements let AI models access underlying code directly. Kris Rasmussen, Figma’s technology chief, explained in a blog that the system “indexes the code in your Make file so you and your favorite platforms can request exactly what’s needed.”

Executives at both firms acknowledge the tension. Melanie Perkins, Canva’s co-founder, cited the high cost of serving AI tasks as the primary driver for the forecast change. Products lag because engineering teams focus on efficiency first. Figma’s leadership highlights organizational challenges revealed by AI. The 2026 report shows adoption varies widely. Grassroots efforts bubble from individual designers. Directive companies push from the top. In both cases teams lack shared playbooks. The result appears in uneven productivity gains and occasional resistance.

So what comes next. Both companies race to build cheaper, more targeted models. Canva works on first-party systems to reduce reliance on expensive APIs. Figma expands agent capabilities and connector skills for tools like GitHub, Notion and Excel. The goal remains tighter integration between design, code and data. Yet the economics lesson lands hard. Generative features that feel magical to users can erode margins quickly. Rollouts must balance excitement with sustainability. Early subsidies to drive adoption now require careful calibration.

Industry observers watch closely. The Australian valued Canva at $42 billion despite the reset. Figma maintains momentum in enterprise accounts. Neither appears at risk of displacement. But the episode reveals limits to bolting advanced AI onto existing platforms without rethinking cost structures and pricing. Users flock to features that save time. They abandon them when alternatives prove cheaper or faster. ChatGPT’s image tools siphon usage. Smaller AI design apps nibble at niches.

The coming quarters will test execution. Canva must deliver on delayed products while containing expenses. Figma needs to convert its AI investments into sustained acceleration rather than the projected September deceleration. Both will likely introduce usage-based pricing tiers or model routing optimizations. The winners may be those who master not just the technology but the unit economics behind it. For an industry that once viewed AI as an unqualified accelerator, the message is more measured. Progress arrives. Costs accumulate. Strategy adjusts. And the design tools that survive will combine creativity with careful calculation.