Cerebras CEO sees personal AI assistants moving into the home
Cerebras CEO Andrew Feldman predicts that personal AI assistants will run partly on local devices and partly in the cloud, but the idea remains a forecast.
Cerebras CEO Andrew Feldman believes personal AI assistants could soon become household appliances, with systems splitting work between local devices and cloud services. He made the prediction while speaking at TechCrunch Disrupt 2026.1
From phones to home AI
Feldman said the shift is not far away, pointing to the progress of consumer AI agents as evidence that the technology is finding a place in everyday life. His vision is not necessarily a single box that does everything: like a smartphone, a home assistant could process some tasks locally while sending others to the cloud.1
That split would allow an assistant to respond quickly on the device while still drawing on larger remote systems when needed. Feldman also suggested that such assistants could learn a great deal about their owners and sit behind a security perimeter, with access to sensitive capabilities such as payments.1
The infrastructure question
Cerebras makes AI computing hardware, so the company has an obvious interest in a future that requires more inference. Feldman said he already uses a collection of agents to manage his calendar and prepare information about people he is due to meet.1
The comments are a forecast, not a product announcement. No Cerebras home assistant was unveiled, and Feldman left the consumer-facing ideas to other founders. The company’s current focus remains the infrastructure that lets other products run.1
Our opinion
The most useful part of Feldman’s prediction is the emphasis on a boundary between local and cloud processing. A home assistant that can act without constantly sending everything away could be more responsive and easier to trust, but access to calendars, personal data and payment cards raises a much higher bar than a chatbot answering questions. The hardware may arrive before the rules and user controls are ready.
For now, this is a credible direction of travel rather than a timetable. The interesting story is whether future assistants earn enough trust to become part of the home’s infrastructure, not simply whether they can run a larger model.