When analyzing the value of a young, pre revenue company, one of the things you want to look it is how established comparable companies are valued. For AI coding assistants, the field is too young to do that directly. However, they are competing in the space of "developer productivity tool for writing code". That space is an established market that is currently dominated by IDEs.
Seeing young companies which are pre-revenue, which are competing for an unproven yet crowded sub market (AI coding assistant) out value an established incumbent in the larger space does not compute.
Add to this the fact that there is very little moat for AI coding assistants. Assuming the market as a whole proves itself, there is a very good chance that the winners will be the established incumbent IDEs who can add AI assistance as a feature in their established products.
All of that is to say, current AI valuations in this space look a lot like a bubble.
This analysis framework you're providing would've missed YouTube (pre-revenue; no incumbent successes; crowded with competitors like Google Video, Metacafe, Vimeo, etc). It would've missed Instagram (pre-revenue; no massive photo-focused incumbents; tons of competing photo sharing apps in the App Store at the time; no moat against a big social app adding filters). It probably. would've missed WhatsApp. And many others.
Which suggests that your framework is lacking.
Here's where:
1. You're neglecting to look at the differences between the fast-rising stars and the comparable incumbents, and instead you're assuming that the incumbents automatically represent a ceiling. In this particular case, JetBrains obviously isn't the most ambitious company on the planet, and isn't focused on hyper growth. There are plenty of avenues for AI IDEs to grow and expand their revenue that have yet to be explored.
2. You're overestimating the importance of concrete moats. Google had no concrete moat either. Just because people can switch easily doesn't mean they necessarily will.
3. These companies aren't pre-revenue. I believe JetBrains is making something like $400-$500 million dollars a year, after 25 years. Cursor is at half of that in just 2 years. Windsurf is also doing big numbers.
4. Related to #3, you're underestimating growth trajectories.
5. You're leaving out the context. Companies that can afford to make $3B acquisitions (a) have tremendous war chests, and (b) have extremely ambitious goals. They're not looking to build the next JetBrains, they're looking to join the pantheon of $1T companies. Achieving massive 10x or 100x or 1000x growth as an investor/owner requires making asymmetrical bets -- bets where if you lose you're still okay, but if you win, you win big.
Reading a bit between the lines, it seems like the buyers either 1) think that ai assisted coding will get good enough that a lot more people will be doing it - that in the future companies in other fields will spend on it for their employees much the way they are paying for general ai assistants now. Or 2) more likely, they think they will get good enough to completely replace programmers, and the current coding assistant's role is mainly to gather information from developers to eventually replace them completely, by selling a spinoff product at a much higher price. They think they need spyware, and coding assistants are the best version available.
I believe the key thing you are missing here is that there is probably an expectation that AI based coding tools will be more mass market, rather than traditional developer market.
ie it's a mistake to estimate the potential upside by looking at the size of the current developer market.
As you move you're coding tools towards a less technical customer base - there are two synergistic effects - your potential customer base is much much larger, and they are simultaneously less technically competent on average ( and so less likely to build their own tools if you charge too much ).
Whether those assumptions are true - time will tell - but I definitely see more people thinking software development is now accessible to them.
Seeing young companies which are pre-revenue, which are competing for an unproven yet crowded sub market (AI coding assistant) out value an established incumbent in the larger space does not compute.
Add to this the fact that there is very little moat for AI coding assistants. Assuming the market as a whole proves itself, there is a very good chance that the winners will be the established incumbent IDEs who can add AI assistance as a feature in their established products.
All of that is to say, current AI valuations in this space look a lot like a bubble.