According to Fortune, Brex analyzed credit card and bill pay data from over 35,000 anonymized customers to identify the 50 fastest-growing software vendors of 2025, focusing on real spending to cut through inflated metrics. The top spot went to AI coding tool Cursor, valued over $29 billion, which saw a staggering 1,000% year-over-year spending growth with “spend compounding every single month.” Other coding tools like Windsurf (No. 6), Replit (No. 9), and CodeRabbit (No. 15) also ranked highly, while AI model marketplace OpenRouter took No. 2 with 1,500% growth. The data, weighted to recent months and excluding public companies and those worth over $30 billion, aims to spotlight sustainable growth, with Brex’s head of data, Sumeet Marwaha, predicting visual AI tools like Kling.ai (No. 3) will be key winners in 2026.
Real money talks
Here’s the thing: in the current AI gold rush, a lot of the traditional metrics are basically broken. ARR (Annual Recurring Revenue) can be gamed with pilot projects and one-off deals. Valuation is, well, a story you tell investors. But actual money leaving a company’s bank account? That’s a signal you can trust. Brex’s dataset is fascinating because it’s a ground-level view of what thousands of businesses—from startups to more established companies—are actually buying, not what they’re promising to maybe buy later. It cuts through the hype and shows who’s solving a painful enough problem that managers are approving budgets right now.
The no-friction revolution
The runaway success of the AI coding tools category is the biggest story here. And Marwaha nailed the reason: friction, or the lack of it. Think about it. A developer can download Cursor or a similar tool, start using it locally on their machine, and see an immediate productivity boost. No need to involve IT, security, or procurement. There’s no six-month “evaluation” cycle. It’s a classic bottom-up adoption model, but supercharged by AI’s instant, tangible value. This category “basically didn’t exist two years ago,” as Marwaha said. Now it’s creating billion-dollar companies because it removed every barrier between a user and the value proposition. That’s a powerful playbook.
Infrastructure and the visual future
But it wasn’t just coding. The list is quietly revealing about where the underpinnings of the AI economy are being built. Seeing names like OpenRouter (a model marketplace), Vast.ai (GPU rental), and Groq (inference chips) ranking so high shows that serious money is flowing into the infrastructure layer. These aren’t flashy consumer apps; they’re the picks and shovels. And then there’s Marwaha’s bet for 2026: visual AI. With Kling.ai, Ideogram, and Runway on the list, the logic is that video and image generation could follow the same path. The winners, he argues, won’t be general “make a picture” tools, but ones that nail a specific use case so well that teams can’t imagine going back. Will it be for marketers, product designers, or filmmakers? That’s the billion-dollar question.
What this means for the bubble
So, does this data suggest the AI bubble is fine? Not exactly. It actually highlights the kind of companies that might survive a shakeout. Brex specifically filtered for companies valued between roughly $5 and $25 billion—the “unicorns” that are big enough to matter but absolutely not “too big to fail.” These are the companies most at risk if investor sentiment sours. But if they’re on this list, it means they have a growing base of customers voting with their wallets. That’s real traction. It’s the difference between a company burning venture capital on marketing and one with a product people urgently need. In 2026, that distinction will be everything. The companies that thrive will be the ones that, like the top coding tools, make themselves indispensable and frictionless to use. Everything else is just noise.
