Uber and OpenAI Retool Rate Limiting Systems

Uber and OpenAI are replacing static rate limits with adaptive, infrastructure-level platforms. Uber’s Global Rate Limiter utilizes probabilistic shedding to manage 80M RPS, while OpenAI’s Access Engine implements a credit waterfall to prevent user interruptions. Both architectures utilize distributed enforcement and soft controls to maintain system stability and service continuity at scale.

By Patrick Farry