Canva's hardest year shows what AI really costs the creator toolkit
Australia's design giant is restructuring 1,271 roles after AI computing costs forced it to cut its growth forecast by a third. Its answer, cheaper in-house models trained partly on platform data, raises new questions for the creators and brands who use it every day.

For a decade, Canva was the Australian start-up that seemed to get everything right. It was profitable, growing fast and used by more than 260 million people a month. Many of them were creators, small brands and social media managers making thumbnails, carousels, media kits and pitch decks without a designer.
On November 4, Sydney time, co-founder and chief executive Melanie Perkins told staff that the run had ended. "This is without doubt the hardest moment in our 13-year history," she wrote in a company-wide note. She said growth had slowed for the first time in a decade, that "small changes won't be enough", and that the company needed to "refound Canva, getting back to our roots of small, empowered teams".
The restructure affects 1,271 roles, about 22% of a workforce that AAP put at roughly 5,700 people in 37 countries. Canva has not said how many of those people will lose their jobs and how many will change roles or be asked to move. According to The Information, as reported by The Next Web, these are the first job cuts since Canva was founded in 2013.
Perkins's note never mentioned AI, but the company's own explanations of its slowdown keep coming back to it. For the creators and marketers who use Canva every day, the real question is what that cost problem means for their tools.
The bill behind the slowdown
The trouble first became public in August. In a shareholder update reported by Startup Daily, Canva cut its 2026 revenue growth forecast from 30% to 20%, a reduction of a third. TechCrunch has reported that Canva ended 2025 with about US$4 billion in annualised revenue.
The business was not shrinking. Revenue for the June quarter rose 25.2% from a year earlier to US$921.9 million, and Canva said it had been profitable for nine years and held US$1.47 billion in cash. The problem was margin. Perkins told shareholders that the "average cost of serving an AI task was too high" and that Canva had been "relying too heavily on frontier models", the large general-purpose models it pays outside companies to run.
She told Fortune that demand for the new AI features launched in April as Canva AI 2.0 had "significantly exceeded" expectations. "This validated the demand, but also showed us we needed to reduce the cost of completing an AI task," she wrote. Instead of "broadly rolling out a product before the underlying economics were ready," Canva slowed the rollout.
Derek Hernandez, a senior analyst at PitchBook, gave Fortune the industry view: "AI is making SaaS no longer a zero marginal cost solution." Traditional software costs almost nothing to serve to one more user. An AI feature, by contrast, runs up a computing bill every time someone clicks "generate."
Investors have repriced the company accordingly. Startup Daily reported that the Australian venture firms Blackbird and Airtree marked their Canva holdings down 17%, from a peak of US$42 billion to US$34.9 billion. Canva has been widely expected to list in the US, but Hernandez told Fortune it might now aim for next year.
Canva's answer: build its own models
Canva's plan is to depend less on other companies' models. In the August update, it said its own video model was 17 times cheaper to run than a frontier model and its image model 30 times cheaper, and that it had cut the average cost of serving an AI task by about 90%.
In the fortnight before the restructure it released two in-house models. Craft, launched on October 22, is the first image-generation model Canva has built itself. Brush, announced a week later, is a design-editing model. It can read a whole layout and change it, for example restyling a design, updating headers across a multi-page deck or turning a presentation slide into an Instagram graphic. Canva says Brush does general editing tasks up to 69% faster than its previous design assistant at 46% lower cost per task. It is in beta for Canva Pro subscribers.
According to The Letter Two's account of the launch, Brush was built by Canva Research, a 160-person AI lab, starting from an existing open-weights model. Its training data included in-house design expertise and anonymised insights from users who had opted in to help improve Canva AI.
Where creators fit in
Creators are both Canva's customers and part of what it sells. Through its Canva Creators program, the company pays designers royalties for templates and graphic elements used on the platform.
In October 2023, when it launched its first big AI suite, Canva said it would pay US$200 million in content and AI royalties to that community over three years. The money would go to creators who agreed to let their work train its models, and those who did not agree would be left out of training by default. That three-year window has now ended. We could not find any public statement from Canva on whether the fund has been renewed or how much of it has been paid.
That matters more now that Canva's strategy depends on its own models. Pricier outside models give way to in-house ones trained partly on design data from its own platform and users. So creators who supply templates and elements have a fair question to ask: how will that training be paid for after the original fund?
The competition is also changing. TechCrunch noted that Canva now competes with general-purpose chatbots such as ChatGPT and Gemini, whose image tools let people put together posters and presentations easily. That is exactly the kind of quick, free graphic that first drew many creators to Canva. Adam Barty, managing director of the Australian digital agency Revium, told AAP: "It's pretty hard not to view AI as an existential threat to Canva's core business model."

"Not using AI to do the work of people who are leaving"
Canva says the cuts are not a straight swap of staff for software. A spokesperson told AAP that "AI is changing what's possible for our community and the kinds of products we're now building". The spokesperson also said the company was "not using AI to do the work of people who are leaving".
The union Professionals Australia is not convinced. "It's really important that large tech companies like Canva are honest about the impact that AI is having on jobs," Paul Inglis, its tech industry director, told AAP. "They owe it to all of us." The union welcomed the severance terms.
Departing staff get at least six months' salary and 12 months of healthcare where it was already provided. A new Canva Collective Fund offers US$10,000 grants, with no equity taken, to former staff who start businesses or community projects. Sydney becomes the main base for more product teams.
What it means for creators
Expect the AI features in the tools you use every day to keep changing in price, limits and quality as companies work out what they can afford to give away. If you depend on any single design tool's AI features for client work, keep a back-up.
If you sell templates or elements on Canva, read your contributor terms and AI-training settings closely. The company's new models are trained partly on its platform's design data. Ask the program directly what happens to AI royalties now that the 2023 fund has run its course.
What it means for brands and agencies
Many in-house social teams produce most of their creator-facing assets, briefs and brand kits in Canva. The company is not going anywhere, but a leaner Canva will put AI features where they earn the most. So far, Brush is reserved for paying Pro users. Budget for AI costs in creative tools as a recurring cost, not a free extra, and check how any tool trains on the brand assets you upload.
Why it matters: Canva is the clearest sign yet that the AI features creators now expect are expensive to run. The company that made design cheap and easy for millions of creators is cutting a fifth of its roles and building its own models to make AI affordable at scale. How it shares that cost, and the value it gets from creators' work, with the people who use and supply its platform will set an example for every creator tool.