What 1,769 open roles reveal about the AI leaders

A clear pattern emerges from the data: these six companies behave less and less like research labs and more and more like industrial firms. Their scientific core remains, but value creation appears to be shifting toward sales, execution and scale.

1 · Hiring turns on distribution, not research

Go-to-market accounts for 29% of all openings and ranks first at four of the six companies, whereas research falls to 11%. What these companies now lack is not researchers, but customers and deployments. Two exceptions prove the rule: xAI puts 58% of its posts into infrastructure and distributes through X and Grok, while Cohere is the only one still betting on research (30% of its openings) as its commercial argument.

2 · Talent is re-concentrating around headquarters

Each company hires in its home country, up to 79% in the US for OpenAI and 63% in France for Mistral, and fully remote work has almost vanished. This is a deliberate design choice: where the knowledge that matters is tacit and speed of iteration decides the advantage, proximity beats flexibility. The one exception, xAI's 21% remote, is its annotation workforce, a task that can be split, measured and moved away.

3 · Researchers stay very well paid, apart from the rest

Researchers remain among the best paid, with a median of $370k, clearly above the corporate and support functions (around $249k), so a real gap separates the research core from the back office. We compare base ranges only here, and because variable pay weighs heavily in commercial roles, we deliberately leave sales out of this comparison.

4 · Hiring clearly favours experience

These companies recruit almost no juniors, 4% in all and zero at OpenAI, and usually ask for five to eight years of experience. The most striking point is that the back office, finance, legal and HR, is even more senior than the labs themselves, staffed with a handful of seasoned operators rather than junior teams.

5 · This senior hiring allows several readings

The fact is clear, but its explanation is not, and several hypotheses can coexist. It may be preparation for a listing, consistent with OpenAI and Anthropic filing to go public in June 2026. It may be that they have already built their junior base and now top up only from the top. Or, more structurally, they may no longer need as many juniors, if AI is absorbing execution work while seniors bring the expertise the machine does not replace. We cannot separate these readings from a single week of data. We will therefore run this analysis again in a few months, on the open roles, and the comparison over time should give us a firmer basis for these deductions.

Fundamentally, sales comes first, experience prevails, teams cluster around headquarters, and the public markets are in view. The scientific frontier remains the core, but value seems to be sliding toward sales, execution and scale. The open question sits at the bottom of the pyramid, where these companies barely hire juniors anymore.

Read on

The six exhibits that follow take each of these findings apart: who each company hires, where it builds its teams, what the work pays, and the profiles it looks for. The last two are interactive, so you can explore any company, or any function, yourself.