Blended rate computed from input/output ratio below
93% in / 7% out
100K
Each region under Jensen's model
What each region gets when you apply Jensen's formula (50% of salary in tokens) locally. Higher salary = more tokens. An Indian engineer's token budget is a fraction of a Bay Area engineer's, despite doing the same work.
Each region under Steve's model
The CFO does the math and fires half the team, redirecting their full salary to tokens for the remaining half. Each remaining engineer gets one fired engineer's salary as their token budget. Again, the Bay Area engineer gets dramatically more AI throughput simply because they cost more.
The relocation matrix
What if you take the budget you're currently spending in one region and hire in another instead? Rows are your current spend (source). Columns are where you hire (destination). Each cell shows tasks/engineer/day and total cost. Green cells give you more tasks than hiring locally; red cells give fewer.
Jensen's model (budget = source salary x 1.5)
Steve's model (budget = source salary)
Senior engineer total comp estimates from levels.fyi (2025-2026):
Bay Area $400K,
US avg $250K,
Canada $170K,
UK London $150K,
Germany $130K,
UK other $100K,
India $60K.
Jensen: salary + 50% in tokens. Steve: fire the engineer, redirect salary. In the matrix, savings from hiring in a cheaper region are redirected to tokens. Token prices are global. The arbitrage is real: same AI model, same quality engineer, dramatically different throughput depending on where you hire.