Canva cuts its revenue growth target to about 20% as AI demand drives up compute costs
Canva cut its growth forecast by a third on August 12, 2026, and the reason it gave was not weak demand. It was the opposite: its AI features became more popular than the company could affordably serve.
What Canva said
On August 12 the company lowered its annual revenue growth target from about 30% to about 20%. Since Canva AI 2.0 launched in April, users have been creating three times as many designs with it as with the previous version, and CEO Melanie Perkins said demand “significantly exceeded” expectations.
The catch is that every AI request carries a real compute cost. Serving the surge got expensive enough that Canva slowed the rollout of new AI features. The company says it has since rebuilt its infrastructure and cut the cost of an AI task by almost 90%.
Why the numbers matter beyond Canva
Canva is not a small private experiment. The last employee share sale valued the company at $42 billion, and an IPO is being discussed for 2027. A growth target cut of this size, attributed to the cost of serving AI features, lands in the middle of that conversation.
Figma is working through the same arithmetic. Its free-cash-flow margin dropped from 27% to 14% in a single quarter. Two of the best-known design software companies are now showing the same pattern: AI usage up, margins or growth targets down.
The new SaaS cost problem
Classic software had near-zero marginal cost. Once a feature was built, each additional user cost almost nothing to serve, which is the economic foundation the SaaS model was built on. AI features break that assumption. Every click that triggers a generation bills the vendor for compute, so the most popular AI features are also the most expensive ones to run.
That inverts the usual logic of product success, and Canva’s own framing shows it: demand exceeded expectations, and the forecast went down. The 90% per-task cost reduction is the counter-move, and it points to where attention goes next, making inference cheap enough that usage growth stops being a margin problem. Whether that fix holds as usage keeps climbing, and whether AI features end up priced separately from the base subscription, are the questions the 2027 IPO discussion will have to answer.
Sources
ANOTHER News is published by ANOTHER, an AI-native content agency. Daily coverage also runs on Instagram.
Related
OpenAI launches a separate ChatGPT for teens that asks guiding questions instead of giving answers
On August 18 OpenAI launched a separate ChatGPT experience for 13 to 17 year olds, rolling out globally on Free and paid personal plans. Its Study Mode responds to a problem with guiding questions and step-by-step prompts rather than a finished solution, a Study Hours setting can make that the default during set times, and quizzes and reminders push toward working the problem. Minors are routed to the teen experience automatically, with stronger defaults around self-harm, violence, eating disorders, dangerous activities and explicit content. Parents can link an account and set limits through opt-in parental controls, but cannot read their teen's chats. TechCrunch and Axios reported the launch and noted that teenagers have been using ChatGPT for schoolwork for three years.
A Claude user reports paragraphs from other people's chats appearing in her replies, twice
A Claude user reports that, while asking about a breathing practice, the model's reply contained a new paragraph that began with the word "user," read like a transcribed voice message from a different person, and which she says she never wrote. She says the same thing happened earlier in a chat about a home renovation, when other people's renovation stories appeared in Claude's answers. When she pointed it out, Claude agreed it saw the pattern and also said a leak between accounts is impossible because user data does not cross; that is the model's self-description, not a statement from Anthropic. She has filed a bug report. The screenshots were shared directly; the original chat is in Russian and the quoted text is machine-translated. This remains one user's report, not a confirmed cross-account leak.
Google wins bankruptcy auction for Spirit Airlines' internal data with a $10 million bid
Google won a bankruptcy auction for the internal data of Spirit Airlines with a $10 million bid, beating a $7.5 million offer from AI data startup Mercor. The package includes roughly 100 million employee emails, 500 million Microsoft Teams messages, 17 million OneDrive files, more than 20 million SharePoint items, 516 code repositories holding about 30 million lines of custom software, and billions of flight pricing records. Google says the enterprise dataset should help improve its products and AI models. A third party will strip personally identifiable information before transfer, at Google's expense, and the airline's 97.5 million passenger profiles and 50.2 million loyalty records are excluded. A federal bankruptcy judge still has to approve the sale, reported by Forbes, TechSpot and Tom's Hardware on August 18.