This piece of writing aims to predict who will win the ai coding space and why.
Github copilot started it off, then Cursor hit $500m arr, Lovable hit $100m arr, your mama hitting $100t arr, everybody is talking about claude code, then there are the smaller players like windsurf, cognition and open source ones like codex, gemini cli, aider, cline, roo, kilo, openhands and many others.
None of them are making profits and you might as well replace the “ARR” with the word “subsidy given out to users”. Some even have negative gross margins! We’re in the $5 ride Uber phase all over again. For Uber, it worked out ok for them, last year they recorded $12B in net income and it looks like it has hit a scale and pricing point where it can become a real business!
Will the same thing happen in the AI coding space? And who will be crown king?
There are a few differences, for Uber it was all about network effects. More drivers, more riders, more money for drivers, less wait times for riders. A win win. It was also in a commodity market where the only thing you can compete on is costs. Yes there are slight differences in the UI of the apps, etc, but the most important thing is riders just want the cheapest ride and drivers want to earn as much as possible. These two reasons justify Uber to sell their dollar bills for 90 cents. (probably even less than 90 cents) If you can get the biggest network first, it’s very hard for anybody else to compete and then you can have pricing power and earn some profits.
So will the same thing happen in the ai coding space? Nope! The reason is there is differentiation in this space. It’s not a commodity. Some want a browser solution, some want a UI, some want something that runs in the terminal. Some need open source for security. Some are for webapps, some are built for mobile apps. Some work better on big codebases, some with MVPs. There are so many characteristics to differentiate that many players can carve out a sizeable chunk of the market and build a brand around them. Lovable is doing this with non coders to a certain extent. Cursor, Claude code, Github copilot are all fighting it out to win the professional developers market, but again the high level of differences in preferability with each developer means that this will not be a winner take all market. Perhaps it will end up like the cloud with an oligopoly market structure.
The only path in which there will be one winner is if a model developer also created a winner application and uses that data to train a model that is in a class of its own relative to others and cut off this model to all competitors. So they own the best coding model and the best application, by a long shot. If this happens then yes, we could be looking at a winner take all, but most likely this won’t happen.
The most likely outcome is there will be a few big winners that will serve the majority of professionals, probably the same names you see now. Then there will also be sizable applications serving niches. So web, mobile, internal tools, design and more. I predict a few players will make out nicely after the AI bubble bursts. The subsidies will end at some point and I expect profits to come out of it.
A comparison I like to use is cereals vs airlines. AI coding is more like cereals. Cereals are not a commodity, people buy each one for many reasons, tastes, packaging and brand. They have nice cost structures, once the factory is set up, the marginal cost to produce each cereal box is low. Operational complexity is low, factories run at high utilization rates and demand is predictable. Airlines have extremely high fixed costs and people value price the most in making their decision. Variable costs (fuel) are large and volatile.
AI Coding is similar to cereals, people use claude code, Cursor, github copilot, because of differences in personal preferences and brand. Costs are variable and demand is predictable. They all do the same thing at roughly the same level, not enough difference in quality to justify a winner taking all.