Canva expands OpenAI partnership to lower AI costs

A strategic shift in AI spending

Canva is navigating a difficult balancing act between providing high-end AI features and managing the massive costs associated with running them. The design platform recently expanded its partnership with OpenAI to gain access to newer models like GPT-6 Astra and GPT-5.6. This move comes as the company tries to find a more economical way to power its suite of tools.

The rising price of running these advanced models forced Canva to rethink its business strategy. In August, CEO Melanie Perkins admitted that the company was relying too heavily on frontier models, which are the most powerful but also most expensive AI systems available. This realization led to a downward revision of revenue growth forecasts from 30% to 20%.

To address these financial pressures, Canva is slowing down some new AI rollouts. The goal is to ensure that the company's own technology and pricing can keep up with the high demand for generative features. By diversifying its model usage, Canva hopes to maintain its competitive edge without sacrificing its profit margins.

The rise of in-house models

While Canva continues to work with OpenAI, it is also doubling down on its own proprietary technology. Just days before the new partnership news, the company launched Brush, its second home-grown AI model. This follows the previous release of Craft, which was designed specifically to generate images for users.

Unlike general models that create content from scratch, Brush is designed for design-editing tasks. It allows users to make specific changes to existing projects, such as updating headlines in a presentation without the AI having to recreate the entire slide. This targeted approach makes the tool more practical for professional workflows where precision is more important than total generation.

Canva's researchers developed Brush by taking an existing open-weights model—a type of AI where the underlying parameters are available for others to see and use—and training it further on specific design tasks. This method allows the company to create a specialized tool that is significantly more efficient than a general-purpose model.

Efficiency and performance gains

The move toward specialized models like Brush is already showing measurable results in terms of speed and cost. Canva reports that Brush is up to 69% faster than the design agent it replaces. Furthermore, the new model costs 46% less per edit, which is a critical metric for a company scaling AI to millions of users.

These improvements are particularly important for Canva Pro subscribers who rely on these tools for high-volume work. By reducing the cost per edit, Canva can offer more robust features without passing the full brunt of AI infrastructure costs onto the customer. It represents a shift toward a more sustainable way to deploy AI in a software-as-a-service business.

Danny Wu, the head of AI product at Canva, noted that the company's long-standing relationship with OpenAI began with GPT-3. He stated that OpenAI's models continue to power a growing number of AI features at Canva, reaching more people. The company is now also expanding the use of ChatGPT and Codex for its internal staff to improve team productivity.

Internal testing and scale

The partnership with OpenAI isn't just about external features; it is also being used to overhaul how Canva operates internally. In one internal test, the GPT-6 Astra model was used to navigate more than 80 million lines of Canva's code. It was also used to write and analyze more than 1,000 queries against the company's data warehouse.

This level of integration shows how Canva plans to use frontier models for complex engineering tasks that require a deep understanding of large datasets. By using these models to help their own engineers, they can potentially build better products faster. It highlights a two-pronged strategy: using high-end models for internal development and custom models for high-volume user tasks.

Canva is currently one of the five most-installed apps in ChatGPT, which has brought millions of new users to its products. This high level of visibility makes it essential for Canva to find the right mix of technology. They must balance the power of OpenAI's ecosystem with the cost-effectiveness of their own specialized models.

Restructuring and the road ahead

Despite the technological expansion, Canva is also undergoing significant internal changes. Reports from Capital Brief revealed that the company made 32 roles redundant in its finance division in January. There is also ongoing speculation regarding a possible hiring freeze or further restructuring as the company adjusts to the new economic reality of AI.

These moves suggest that while Canva is investing heavily in AI technology, it is also being very disciplined about its headcount and spending. The company is trying to ensure that its growth remains sustainable as it navigates the high costs of the generative AI era. This focus on efficiency is likely to dictate how they roll out new features in the coming months.

Moving forward, the success of Canva's strategy will depend on how well it can balance these two worlds. They need to keep the