For most of the past decade, Melanie Perkins has been the face of Australian tech success. Canva, the online design platform she co-founded, became one of the world’s most valuable private software companies. In 2026, though, she has had to make the kind of decision that tests a founder: slowing growth to fix the economics of an expensive bet on artificial intelligence.
From yearbooks to a global platform
The Canva story is well known. Perkins, who grew up in Perth, was frustrated by how hard design software was to use. With Cliff Obrecht, now Canva’s chief operating officer, she first tested the idea through Fusion Books, an online tool that let schools design their own yearbooks. Canva was founded in 2012 with third co-founder Cameron Adams and launched in 2013.
Its rise was steep. In September 2021 it raised US$200 million at a US$40 billion valuation, and Perkins and Obrecht pledged about 30% of the company’s equity to the Canva Foundation. In February 2026, Obrecht disclosed that Canva had finished 2025 with US$4 billion in annual recurring revenue, more than 265 million monthly active users and more than 31 million paying users.
The AI cost shock
The trouble came with Canva’s push to put AI at the centre of its product. Demand for its new AI features was strong, but so were the bills for using frontier AI models supplied by other companies.
In her update to shareholders for the June quarter, Perkins was unusually frank. The “average cost of serving an AI task was too high”, she wrote, and Canva had been “relying too heavily on frontier models.” She described cutting the 2026 revenue growth forecast from 30% to 20% as a “deliberate decision to get the economics right.”
“We decided to slow the rollout while we rebuilt the architecture, reduced unit costs and strengthened the business model,” she told investors. Revenue for the June quarter still rose 25.2% to US$921.9 million, but that was below Canva’s own target.
Canva has since invested heavily in its own models. It has publicised an in-house AI lab, Canva Research, with more than 160 scientists and engineers, and its first image model, Canva Craft. According to The Sydney Morning Herald, the company has cut the cost of running its AI by about 90%.
The valuation markdown
Markets noticed. In August, two of Canva’s longest-standing backers, Blackbird and AirTree, cut the value at which they hold the company to US$34.9 billion, from US$42 billion, according to reporting by the Australian Financial Review. An independent valuation used to price staff shares fell to US$31 billion, from US$38.9 billion a year earlier. None of these were transactions; Canva’s last priced deal was an employee share sale in 2025 at US$42 billion.
Perkins has defended the trade-off as consistent with how Canva has always operated. “We knew this transition would mean balancing our near-term growth with investing in the enormous opportunity ahead, and we’ve always been willing to make those long-term decisions,” she told the Herald.
The IPO question
The slowdown matters most for Canva’s long-awaited share market listing. In April, Obrecht told Capital Brief the company was “fully IPO ready” and that he assumed it would list in 2027, once its move to an AI-credit business model had bedded down. In September, Capital Brief reported that Canva had been hiring for roles associated with running a US-listed company, while a spokesperson said there was nothing new to share about an IPO.
For Australian investors, the stakes are significant. A Canva float would be among the largest exits ever for an Australian venture-backed company, and would test whether public markets reward a founder who chose margins over speed.
The bigger lesson
Perkins’ year is a case study for every software company racing to add AI. Usage-based AI costs can grow faster than revenue, and pricing and controls need to keep up. Canva’s willingness to say so publicly, and to absorb the hit to its valuation, may prove to be one of her more important decisions as chief executive.