Vibe coding is one of those tech trends that sounds almost magical. You type in plain English (or any language, really), and the AI turns it into working code. There are quite a few players in this game right now. Lovable promises to help anyone—not just developers—create full-stack apps with just a few prompts. Replit makes quick web prototypes a breeze, especially for beginners or hobbyists. Similar to these are: Bolt, v0, and more. Then there are the heavy-duty developer tools like Cursor and Windsurf, aimed at speeding up professional coding work.
Different flavors, same goal: make software creation faster, easier, and a little more fun.
But Here’s the Catch…
All these tools—whether they’re targeting non-tech creators like Lovable, casual coders on Replit, or pro developers using Cursor and Windsurf—face a surprisingly similar challenge: making money without going broke in the process.
The problem is the cost of running the AI models behind the magic. The best ones right now—like Claude or GPT-5—are expensive to use. And users expect access to the very best models without paying eye-watering subscription fees. That leaves companies with a tricky choice: charge more and risk losing customers, or keep prices low and burn through cash.
Windsurf’s Wild Ride
If you’ve been following Windsurf’s story, you know it’s been a bit of a rollercoaster. Once valued at nearly $3 billion, it was almost sold to OpenAI before the deal fell apart one month ago (11 July). Then Google swooped in to hire some of the top talent for $2.4 billion (!!!), while the rest of the company was bought by Cognition for $250 million – a fraction of that earlier valuation.
Windsurf is still alive, still adding features (including GPT-5 support), but it’s gone through layoffs and a big leadership change. And all this makes you wonder: if vibe coding is such a healthy business, why would Google want only the main staff and show no interest in the business?
Vibe Coding: Margins That Make Investors Nervous
Cursor reportedly brings in about $100 million a year, Windsurf about $50 million. Sounds great… until you see the margins. Running these tools costs so much that profits stay negative. Even Lovable, which is growing fast, faces the same issue: high AI costs eat into revenue before it ever reaches the bottom line.
Some companies are trying to solve this by building their own AI models to cut costs. That’s a huge bet—training competitive AI isn’t just expensive, it’s a constant race to keep up with the newest tech.
Can Vibe Coding Survive?
The answer is absolutely yes—one way or another. However, the present players in this area must find a way to cut costs for expensive AI calls if they ever are going to reach breakeven. The way to do that is probably by strategic alliances with the owners of the large models.
The magic of vibe coding is real. But maybe it will go the same way as many other AI applications lately – in the end just be a part of the web browser that you use for free. Time will tell.