Canva has cut its annual revenue forecast after revealing the cost of rolling out its highly anticipated AI suite proved far higher than expected, forcing the design giant to slow its launch.
The Australian-founded company, valued at $42 billion (A$59.62 billion), now expects annual revenue growth of around 20 per cent, down from the 30 per cent target it set at the start of the year. The downgrade comes despite quarterly revenue climbing 25 per cent year-on-year to $921.9 million.
In a letter to shareholders, Canva co-founder and CEO Melanie Perkins said the company deliberately delayed the broader rollout of Canva AI after discovering the economics behind serving AI requests at scale simply weren’t sustainable.
“Rather than broadly rolling out a product before the underlying economics were ready, we decided to slow the rollout while we rebuilt the architecture, reduced unit costs and strengthened the business model,” Perkins said.
“This slowed our distribution and impacted our near-term growth, but it also enabled some of the most important technical advances in Canva’s history and put us in a much stronger position to scale AI sustainably.”
The admission offers a rare glimpse into the enormous infrastructure costs facing companies racing to embed generative AI into mainstream products, even as customer demand accelerates.
Canva launched Canva AI 2.0 in April, introducing AI-powered workflows capable of generating fully editable marketing campaigns from a simple prompt while pulling information from connected platforms including Slack and Gmail.
At the time, Perkins labelled it “the most significant moment in Canva’s history”.
However, the CEO said the business quickly realised the average cost of serving each AI request was too high.
“We were relying too heavily on frontier models,” she admitted. “Several of our first-party models were not yet ready for release, and our pricing, consumption model and usage controls had not caught up with the outsized demand we were seeing.”
Rather than continuing to absorb those costs, Canva spent the past three months rebuilding its AI infrastructure, developing more of its own models and reducing its reliance on third-party providers.
According to Perkins, the company has since reduced the cost of processing a single AI task by almost 90 per cent.
Its internally developed style-transfer model is now 23 times cheaper than comparable frontier AI models, while its image generation model is 30 times cheaper and its video model 17 times cheaper.
The savings matter at Canva’s scale.
Users now create more than one billion designs every month, meaning even marginal reductions in the cost of each AI interaction translate into significant savings across the platform.
Canva has also expanded its AI capability internally, with more than 200 researchers, engineers and platform specialists now focused on artificial intelligence, including over 140 working in its dedicated research lab.
The company has also begun prioritising content it already owns, using its library of more than 165 million templates and creative assets wherever possible before generating entirely new AI content.
Canva has begun rolling out Canva AI 2.1, which Perkins described as the first version built on an economic model designed to support the company’s freemium growth strategy.
“We’ll roll out Canva AI 2.1 thoughtfully, expanding access as we continue to monitor quality, engagement and unit economics,” she said.
