ChatGPT is processing 2.5 billion messages a day (www.businessinsider.com)

ChatGPT is now handling at least 2.5 billion messages per day — a figure reported from July when the service had about 700 million users — and OpenAI’s weekly active user count has since climbed to roughly 800 million, which Sam Altman says amounts to more than 10% of the world’s adult population. Researchers who studied usage patterns found that around 70% of queries are non-work-related, with the most common request types being practical guidance, information-seeking, and writing; among work-related uses, generating novel text is the dominant task. The rate of global adoption, the paper concludes, is historically unprecedented for a new technology. For the AI/ML community this scale is both an opportunity and a technical challenge: billions of daily messages imply massive inference load, increased infrastructure and cost pressures, and urgent needs for efficient model serving, latency optimization, and robust monitoring. High-volume consumer usage also accelerates distributional shifts in inputs, raising stakes for continual fine-tuning, safety filters, evaluation pipelines, and privacy controls. Product moves like ChatGPT’s Walmart integration underscore rapid commercialization and cross-industry embedding of LLMs, meaning researchers and engineers must prioritize real-world robustness, guardrails, and metrics that reflect diverse, non-work-driven user behavior.