Canva cuts revenue growth forecast on AI cost blowout

Canva just did something unusual for the AI-first design era: it cut its own growth guidance. The Information first reported that the $42 billion Australian company 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, after the AI suite it rolled out proved far more expensive to run than management planned.

The specifics, as summarised by B&T: second-quarter revenue landed at $921.9 million, up 25.2 per cent year-on-year, which is a strong number in isolation but a miss against Canva's own guidance. Co-founder and CEO Melanie Perkins told investors the company had been 'relying too heavily on frontier models,' its first-party models weren't ready, and pricing and usage controls hadn't caught up with demand. Some AI features are reportedly running as much as six months behind schedule.

The interesting part is the fix. Canva says it spent the past three months rebuilding its AI infrastructure and moving more inference onto its own models, and the company's claim is a 90 per cent reduction in the cost of serving an AI task. If that holds up as usage scales, it is a meaningful change in unit economics rather than a one-quarter trick. The power-user tier, AI Pass, is priced at USD $100 per person per month on top of Pro or Business.

ChatGPT is the other half of the story, and the picture there is more nuanced than 'users are leaving.' Canva reported more than 26 million conversations with its ChatGPT app by October 2025 and claims to be a top-10 referred domain from ChatGPT, which is real distribution, while also being the same channel that lets people generate a design without ever opening Canva.

The honest caveat is that this is a private company briefing its own investors, so we are seeing the numbers Canva chose to share and the framing Canva chose to tell them in. What the reporting doesn't give you is AI Pass adoption, gross margin on AI usage, or how much of the miss traces to the $100 tier landing softly versus features simply arriving late. Still, if a category leader is telling its backers that AI unit costs forced it to slow down and rebuild, that is the datapoint every other SaaS company staring at the same bill should sit with.