Canva has spent a decade convincing the world that anyone can be a designer. Now the Sydney company faces a harder question: whether it can convince its own users to pay for the artificial intelligence that is rapidly rewriting what a design tool even is. The stakes are not modest. A wave of AI image and document generators has arrived at exactly the moment Canva reached global scale, and the technology that makes the platform smarter also threatens to make parts of it redundant.
According to the Australian Financial Review, the company is running a quiet experiment to work out how to embed generative AI across its products without blowing up the economics that made it one of the country’s most valuable private businesses. The tension is simple to describe and fiendishly hard to solve. Canva built its empire on a generous free tier, luring hundreds of millions of casual users who cost very little to serve. AI features do not follow that logic. Every image conjured, every document drafted and every background removed by a large model carries a real compute bill, and that bill lands whether the user is a paying customer or not.
Why AI is both the opportunity and the threat
For most of its life, Canva enjoyed the delightful economics of software. Once the product was built, serving one more user cost almost nothing. Generative AI upends that arrangement. The models that power text-to-image tools and automated layouts run on expensive graphics chips, and inference costs scale with usage rather than sitting as a fixed overhead. A free user who generates fifty images in an afternoon is now a genuine cost centre, not a marketing expense that might one day convert.
That is the existential piece. If a new generation of AI-native tools can produce a polished flyer or a social post from a single sentence, the very act of dragging elements around a canvas starts to look old-fashioned. Canva cannot afford to let a nimbler rival define the AI design category, yet it also cannot afford to hand out unlimited AI generation to a user base measured in the hundreds of millions. Get the balance wrong in one direction and it bleeds cash. Get it wrong in the other and it cedes the future to competitors.
The company’s answer, as reported, has been to test carefully rather than to bet the house in public. Canva has folded AI into features such as Magic Studio and its design assistants, while watching closely how usage translates into cost and, crucially, into upgrades. The experiment is really about pricing and packaging as much as technology: how much AI to give away to keep the funnel full, and how much to reserve for the paid tiers that actually cover the compute.
Two ways to read the gamble
Optimists argue Canva is better placed than almost anyone to make this transition. It has the distribution, the brand and a habit of shipping AI features that ordinary people actually use rather than tech demos that impress only engineers. Co-founder and chief executive Melanie Perkins has long framed the company as a platform that democratises design, and on that reading AI is simply the next tool in the kit, one that deepens the moat by making the product indispensable to schools, small businesses and marketing teams.
Sceptics see a classic innovator’s dilemma. The free tier that built Canva’s growth is precisely the thing that makes an aggressive AI pivot financially painful, and companies rarely cannibalise the model that made them successful without a fight. There is also the uncomfortable possibility that AI commoditises design itself. If generation becomes cheap and universal, the premium once attached to a slick template evaporates, and Canva ends up competing on the cost of compute against far larger players who own their own chips and data centres. The company’s strength, its enormous base of casual users, could become the anchor that slows it down.
What it means for Australia
Canva is not just another software company to Australia. It is the local tech industry’s proof point, the business every founder cites when arguing that a globally significant technology firm can be built out of Sydney rather than San Francisco. Its private valuation has run into the tens of billions of US dollars, and its success has helped seed a generation of local startups through both talent and the wealth of its early backers. How it navigates the AI shift therefore carries a weight beyond its own balance sheet.
The compute question also lands squarely on a national nerve. Australia is in the middle of a scramble to build AI infrastructure, with billions of dollars flowing into data centres and so-called AI factories, and the cost of inference is the reason those investments matter. A homegrown champion that suddenly needs vast amounts of processing power is a live example of why sovereign capacity and access to affordable compute have become policy priorities rather than abstractions. If even Canva feels the squeeze, smaller Australian firms trying to build AI products will feel it far harder.
There is a workforce angle too. Canva’s tools sit in the hands of countless Australian small businesses, marketers, teachers and community groups who use them precisely because professional design help is out of reach. If AI makes those tools more capable, it lowers the barrier further. If it makes them more expensive, it risks pricing out the very users who made the platform a household name. The company’s pricing decisions will quietly shape how a large slice of the Australian economy produces its everyday communication.
What happens next
The near-term signal to watch is packaging. Expect Canva to keep drawing sharper lines between what free users get and what sits behind a subscription, using AI as the lever that nudges people to upgrade. The company will also be under pressure to show that its AI features drive genuine conversion rather than simply inflating costs, a metric its investors will scrutinise closely ahead of any eventual public listing.
Longer term, the real test is whether Canva can turn AI from a cost problem into a durable advantage. If it manages to make generation feel native, useful and worth paying for, the free base becomes an asset again, a vast audience ready to be upsold. If it cannot, the same technology that promised to supercharge the platform could hollow it out. Either way, the experiment underway in Sydney is one of the more consequential bets in Australian technology, and the outcome will say a great deal about whether the country’s biggest software success can reinvent itself for a second act.
Update, 19 August 2026: Canva says AI costs have fallen 90 per cent
That balance between compute bills and free-tier generosity has now been given a hard number. In comments reported on 19 August, a Canva founder said the company had driven the cost of running its generative AI features down by about 90 per cent, after those bills climbed so steeply that they forced the business to delay putting some AI products in front of users at all.
The admission is striking because it confirms, from inside the company, exactly the squeeze this article described. Canva did not simply fold AI into Magic Studio and wait for upgrades to follow: it hit a wall where the inference costs of serving hundreds of millions of users made a wider rollout uneconomic until the engineering caught up. The reported 90 per cent reduction is less a victory lap than an account of how much work it took to make the numbers survivable in the first place, according to Forbes Australia.
For the wider Australian sector, the detail matters more than the headline figure. If the country’s best-resourced tech champion had to defer shipping AI while it wrestled its compute bill down by an order of magnitude, the barrier facing smaller local firms is considerably steeper, and it sharpens the case that affordable inference and sovereign capacity are commercial necessities rather than policy talking points.
Sources: Australian Financial Review (Technology). · GNews · Canva-Atlassian AI
