Earlier this week, we posted about Google's token count taking off across products. Looks like API demand is no different.

A year ago, it was 480T. Today it's 3.2Q! Tomorrow's chart might need a new unit of measurement.

Last time we ran this chart, OpenClaw looked like an outlier. Turns out it was just the new normal for next-gen agents!

Cash flow is transferring from hyperscalers to AI infrastructure.

Follow the bottlenecks.

Dispersion has been extreme, and even more dramatic within TMT!

We are watching the fastest revenue speedrun of all time.

What's the Chinese word for Quibi?

AI reprices software around consumption. Traditional SaaS still charges for seats.

One number tells the story: 3.8%. That's the share of Traditional SaaS spend that's consumption-based, eleven months into the AI era.

Data collaboration with Ara Kharazian of Ramp.

Token demand is entering overdrive.

It took 20 years for each internet user to consume 50Gb of memory annually. Could AI 5x that in just 5 years?