Modal hits $1B run rate, raises $750M at $16B valuation

Modal's rise tracks a shift in how AI infrastructure is bought and sold. A typical capacity contract gives a customer a fixed number of GPUs for a fixed period, which Erik Bernhardsson describes as a rectangle when plotted against time. Real workloads grow, spike, and fall, so buying for the peak leaves idle hardware and buying less risks missing demand. Modal's bet is to pool demand across thousands of customers and rent capacity underneath in smaller, elastic slices. That model worked well when most capacity came from spot and one-year reservations across hyperscalers. But as Modal grew and the cloud market changed, it had to work with almost 25 clouds and accept longer deals with upfront payments, building a procurement desk and a team to manage bare-metal capacity worldwide. Bernhardsson also flags a financial risk: if a provider locks in capacity at today's prices and GPU prices later fall, it can be left holding expensive supply. He says he doesn't expect an imminent collapse, but the possibility matters when building the business. The sandbox business adds another layer, since agents may create environments, run code, inspect results, and retry many times, making startup and restore latency part of the experience.