Canva’s valuation reset leaves AI economics, not one price, as the IPO test

The spread is a reason to resist two symmetrical errors. The reductions do not prove a collapse, and the previous US$42 billion mark does not establish a future IPO price. They show that the company and its investors are working with less generous assumptions while the product transition is unfinished.

The AI cost problem is an operating claim, not a solved one

Perkins described the pause as a deliberate attempt to get the economics right: average AI-task serving costs were too high, she wrote, and the delay slowed distribution and near-term growth. A separate report likewise linked the reassessment to the difficulty of rolling out AI-powered tools while paying for frontier models.

That is more specific than a generic AI investment story. Canva is not merely adding a feature; it is trying to alter the cost of delivering a service it expects users to employ at scale. Lower per-task cost could permit broader distribution. But until the revised system is used widely enough to test that claim, the projected benefit remains a company and investor assertion rather than independently audited operating evidence.

Blackbird partner Rick Baker said in the report that the company had reduced AI serving costs by 90% and that early user signals for Canva AI were strong. The figures and the assessment are the investor’s description of the company’s progress. They are encouraging counterevidence to a simple decline narrative, but the reporting does not disclose the starting cost, the usage volume or the resulting margin.

Growth and rivals complicate the bearish reading

Canva still had substantial operating momentum in the reporting period. In the reporting, its June-quarter revenue grew 25.2% year on year to US$921.9 million; the company had been profitable for nine years and ended the quarter with US$1.47 billion in cash. Those reported results explain why a reduced internal share valuation should not be confused with evidence of a near-term liquidity problem.

The competitive context also argues against treating a single private mark as a clean company-specific judgment. The reporting describes Figma’s shares as more than 70% below their post-IPO highs and says Adobe had also been heavily marked down. Those companies are not like-for-like valuation comparables, but their performance is relevant to the investor debate because it changes the public-market backdrop for design-software businesses pursuing AI strategies.

The reported US$12 billion Canva valuation suggested in an investor podcast discussion is best understood as a hypothetical, not a transaction price or consensus estimate. It rested on a revenue-multiple comparison and an argument that AI could weaken Canva’s position; another participant said the value could be higher if the company overcame that risk. The disagreement is evidence of uncertainty, not a defensible replacement for the current marks.

What would resolve the IPO question

The commonly discussed 2027 US listing remains tentative. The next useful evidence is not another isolated private valuation but whether Canva can demonstrate both parts of Perkins’s proposition after the rollout resumes:

• revenue growth that holds or improves after the product delay;
• AI serving costs that remain lower as usage rises, with enough disclosure to judge the unit economics; and
• a price from a financing, liquid secondary sale or eventual listing that puts willing buyers and sellers on a comparable basis.

Until then, US$31 billion, US$34.9 billion and roughly US$30 billion are signals produced by different mechanisms. Canva’s more consequential test is whether its architectural rebuild turns the claimed cost reduction into durable growth rather than a temporary pause in a tougher market.