Tokenomics for Your 2027 R&D Budget

Insights from Jellyfish Research + a five-step exercise

An engineer costs about the same in month twelve as in month one. AI spend per engineer doesn’t. It compounds. It jumps. And what each extra dollar buys you depends on how big your team is. If your budget treats AI as a flat per-seat line, the data doesn’t back it up.

 

Jellyfish Research looked at AI usage, delivery and work allocation data from March through August 2026, including more than 878,000 developer-weeks. The guide shows how AI spend really behaves, so you can set your 2027 number with confidence instead of guessing.

 

What you'll learn
  • The compounding bill: why the typical company’s AI spend per active user tripled in six months, and why a flat monthly line is the riskiest assumption in your plan
  • The size penalty: why teams with 200+ engineers start getting less from extra AI spend at about half the token usage of smaller teams, and why their ceiling is lower
  • The adoption multiplier: why companies where every engineer uses AI daily merge nearly twice as many PRs per engineer
  • The allocation blind spot: why tripling AI spend barely changed what engineers work on, and what your plan needs to say about where the freed-up capacity goes
  • A five-step exercise: plot your team on the token curve and check whether a bigger budget actually pays off, in one afternoon

It’s built for CTOs, VPs of Engineering and R&D leaders who are setting headcount, investment areas and AI budgets for 2027, and need numbers that will hold up with finance.