$350,000
20
Blended rate uses real-world ratio: 93% input, 7% output
Both approaches spend the same on tokens
$3,500,000/yr
Steve: 10 cut salaries redirected. Jensen: 50% of 20 salaries added on top.
Steve: replace half, fund the rest
Engineers 10 remain, 10 cut
Salary spend $3,500,000
Total cost (salary + tokens) $7,000,000
151.7M
tokens / engineer / day
Token budget per engineer $350,000/yr
Tasks per engineer / year 554K
Total team tokens / year 554B
Total team tasks / year 5.5M
Jensen: keep all, add 50% in tokens
Engineers 20 remain, 0 cut
Salary spend $7,000,000
Total cost (salary + tokens) $10,500,000
75.9M
tokens / engineer / day
Token budget per engineer $175,000/yr
Tasks per engineer / year 277K
Total team tokens / year 554B
Total team tasks / year 5.5M

The trade-off

Jensen costs more
+$3,500,000/yr
Jensen keeps more engineers
+10
Each Steve engineer gets
2.0x more tokens

At these defaults, both models generate the same $3.5M token budget: Steve redirects 10 cut salaries, Jensen adds 50% on top of 20 salaries. Same tokens purchased, but Steve's engineers each get twice as many because they're split across half the team. Jensen's total bill is 50% higher because all 20 salaries are still being paid. The question is whether 20 engineers with fewer tokens each outperform 10 engineers with more. Blended token rate: 93.4% input, 6.6% output. Tasks estimated at ~100K tokens each. Prices from OpenRouter, March 2026.