Same technology, same 2 years, 625x apart.

The top 1% burns a median company's entire annual AI budget in a single afternoon.

Ramp's data puts the median company at $12 per employee per month on AI and the top 1% at $7,500, a spread of roughly 625 to 1.

Today's upper tail already demonstrates how much inference a company can economically absorb, while the median firm is barely participating.

The median doesn't need to approach the top 1% for inference demand to grow enormously over the next 2-5 years; even modest movement toward today's power-user intensity would multiply token consumption across a much larger base of companies.

Chart from @a16z and @tryramp