AI Application Companies Face First "Check-up" on Gross Profitability: Canva Voluntarily Slows Down, Figma Absorbs Inference Costs

The two leading companies in the design software sector are exposing the most hidden costs of AI transformation through real financial data.

Canva issued a warning to investors this week that annual revenue growth will slow to 20% because the company proactively halted a planned rollout of AI features intended to drive paid subscription growth—demand far exceeded expectations, and so did service costs. Figma released its earnings report on Thursday, revealing that it expects revenue growth in the third quarter to slow from 48% in the June quarter to 36%, and admitted that many of its AI tools are still in the beta phase and have not yet established an effective commercialization path. After the announcement, Figma's stock price plummeted about 15% in a single day.

According to technology media The Information, both companies are facing the same structural problem: The popularity of AI products does not automatically equate to a healthy unit economics model. Under the narrative framework where investors previously placed high expectations on AI application companies, this serves as a real-world stress test from the frontlines.

Canva: Overwhelming Demand Turns into a Burden, Proactively Hits the Brakes

Canva's dilemma is quite dramatic—not because there aren't enough users, but precisely because there are too many.

According to reports, Canva COO Cliff Obrecht told colleagues that before AI was introduced, the cost of serving a large number of free users was "very low." But after adding AI features, "those costs soared, fundamentally changing the unit economics, making reducing AI costs far more important than before."

Based on this judgment, Canva chose to proactively slow down the planned rollout of AI features, rather than continue pursuing user growth at a high cost. This decision directly lowered the company's full-year revenue growth expectations, from previously higher market expectations down to 20%.

To fundamentally address the cost issue, Canva is pushing hard to develop its own AI models. The company claims its in-house models generate images and videos faster and cheaper than those of top AI labs. However, these models were not ready in time for the current AI feature rollout window, forcing the company to still rely on higher-cost external models at key moments.

Figma: Beta Products Offered Free, Company Bears All Inference Costs

Unlike Canva's proactive slowdown, Figma is facing a different form of cost pressure—paying the price for products that have not yet been commercialized.

Figma CFO Praveer Melwani explained this logic directly during an investor call:

"We currently do not charge customers for using products in the beta phase; we bear the inference costs ourselves, with no revenue to offset those costs."

This means Figma is subsidizing users' utilization of its AI features out of its own pocket, while waiting for these tools to move from beta to full commercialization. The company expects this cost structure will put pressure on gross margins.

Figma is also ramping up its investment in proprietary AI models and has started combining internal models with cutting-edge ones to support its newly launched Figma AI agent. But training proprietary models itself requires time and capital, and the capital market's patience is obviously limited—after the performance guidance was released, Figma's stock price fell by about 15% in one day.

Shared Dilemma: The Cost Structure of AI Transformation Remains Unresolved

The report notes that Canva and Figma's cases reveal a common and current contradiction facing AI application companies: product attractiveness has been validated, but the path to sustainable profitability remains unclear.

Both companies see proprietary models as the key to breaking through—reducing reliance on costly third-party models and fundamentally improving unit economics. However, there is an unavoidable time gap between building these models and the market's immediate demand for performance growth.

Analysts believe these two earnings reports offer an important reference: The cost of AI transformation is reflected not only in R&D investment, but is more deeply embedded in the inference costs behind every user interaction. Before the commercialization model matures, these costs will continue to erode gross margins.

The AI narrative of software companies is now undergoing its first real financial check-up.