Canva lowers growth forecast due to high AI implementation costs

The high cost of artificial intelligence implementation is prompting enterprises to seek greater financial sovereignty over their AI stacks. Canva Inc. serves as a primary example, having reportedly lowered its 2026 revenue-growth forecast from 30% to 20% due to unexpectedly high input costs associated with AI features.

To mitigate these expenses, Canva transitioned from a heavy reliance on expensive third-party frontier models to a custom stack utilizing in-house models, Leonardo.AI, and task-level routing. This shift reportedly reduced the cost of AI tasks by approximately 90%, with video and image models becoming 17 to 30 times cheaper than frontier alternatives.

Industry experts and leaders, including Palantir Technologies Inc. CEO Alex Karp, argue against ‘tokenmaxxing’—the practice of optimizing vendor bills rather than retaining enterprise value. Instead, the focus is shifting toward ‘sovereign alpha,’ where companies maintain control over their data, evaluations, and cost telemetry to ensure they capture the economic benefits of their AI workflows.