Canva has slowed its plans to expand AI features across its platform. The company cited rising inference costs as a key factor behind the decision. Inference costs refer to the computing expense of running AI models each time a user requests a feature, such as image generation or text editing.
These costs differ from the one-time expense of training a model. Inference happens continuously, scaling with every user interaction. For a platform with millions of active users, that expense can grow quickly and unpredictably.
Figma, a competitor in the design software space, has taken a different approach. The company is reportedly absorbing the cost of running its newer AI tools rather than immediately passing those expenses to customers through higher fees. This suggests Figma views near-term AI adoption as a priority, even if it pressures margins in the short term.
The divergence between the two companies illustrates a broader tension facing software firms that have layered generative AI into their products. Many companies rushed to add AI features over the past two years to stay competitive. Fewer have fully solved how to pay for the computing power those features require at scale.
Inference costs depend heavily on the pricing of cloud computing and specialized chips, particularly graphics processing units used to run AI models. Demand for that hardware has remained strong, keeping prices elevated for many providers. Companies without their own infrastructure often rely on major cloud vendors, which can limit their ability to control costs directly.
Canva's decision to slow its AI rollout suggests the company is reassessing how quickly it can expand these features without hurting profitability. A slower pace could mean fewer new AI tools reaching users in the near term, or existing tools being offered more selectively. Figma's choice to absorb costs, by contrast, signals a bet that maintaining momentum with AI adoption is worth the near-term financial trade-off.
Both strategies reflect the uncertainty many software companies face as they try to balance user demand for AI capabilities against the real costs of delivering them. Neither approach guarantees long-term success. Cost structures for AI computing remain in flux, shaped by chip supply, cloud pricing, and the pace of efficiency improvements in AI models themselves.
The situation also underscores how AI features, once seen mainly as a competitive differentiator, are now forcing companies to make harder decisions about margins, pricing, and product strategy. How Canva and Figma navigate this in the coming months could offer a signal for how the broader software industry handles similar pressures.
Market Impact
For software and technology investors, this divergence offers a case study in how AI monetization is playing out unevenly across companies. Firms that slow AI rollouts to protect margins may face criticism for falling behind on features, while those absorbing costs risk margin compression if usage scales faster than efficiency gains in AI computing.
The broader implication touches sectors tied to AI infrastructure, including cloud computing providers and chipmakers supplying the hardware behind inference workloads. Continued high demand for that computing capacity, paired with software companies searching for ways to manage costs, could keep pressure on pricing dynamics across the AI supply chain for the foreseeable future.
As AI features become standard in software products, companies are being forced to choose between slower rollouts and absorbed costs. Canva and Figma's differing responses may foreshadow decisions many other firms will soon face.
Frequently Asked Questions
What are inference costs in AI?
Inference costs are the computing expenses incurred each time an AI model processes a user request, such as generating an image or editing text, as opposed to the one-time cost of training the model.
Why did Canva slow its AI feature rollout?
Canva cited rising inference costs as a reason for slowing the expansion of AI-powered features across its platform, according to reporting on the matter.
How is Figma handling AI-related costs differently?
Figma is reportedly absorbing the expense of running its newer AI tools internally, rather than immediately passing those costs on to customers.
Does this affect all AI-powered software companies?
The facts reported concern Canva and Figma specifically, but the underlying challenge of managing inference costs applies broadly across companies building AI features into their products.