Cerebras CEO says we’re nearing a future where everyone has a personal AI assistant at home

Andrew Feldman, the CEO and co-founder of chipmaker Cerebras Systems, said on stage at TechCrunch Disrupt 2026 on Tuesday that a future in which every household has a personal AI assistant is not far off. Asked what would need to happen for AI to become a home appliance, he replied, "Actually, I don't think that's very far away."
Feldman's model for the product is a hybrid one. He expects personal AI assistants that run locally on hardware in the home, doing some work on device while sending other work back to the cloud, an arrangement he compared to how a cell phone operates today. In his description, these assistants get to know their owner well, sit inside a security perimeter, and hold access to credit cards so they can make purchases on the user's behalf. He said he expects them to make a meaningful impact in making our lives better.
He pointed to recent momentum in consumer AI agents as evidence that the category is finding its footing with ordinary users. He named Meta's Muse and Instinct, the startup he said is now valued at $10 billion, as examples. The reference to a $10 billion valuation is notable because it places an agent startup in the same financial tier as well-funded infrastructure companies, which is the kind of signal that tends to pull more capital and talent toward consumer-facing agents.
The vision is an obvious fit for a company that sells AI compute, and Feldman's own habits were part of his argument. He said he now uses a collection of agents to sweep his calendar and to pull up information about the people he is about to have a call with, among other tasks. Those are the sorts of chores that were previously done by hand, and he framed the current wave as already handling much of that work.
What interests him more, he said, are the possibilities that did not exist before, though he is content to leave the specifics to other founders. He described his preference for infrastructure work in terms of that division of labor, saying one reason he loves being in infrastructure is that he does not need the consumer-facing idea, and that he likes building the platforms that other people's ideas take flight on.
For developers and product teams, several practical questions follow from the hybrid design Feldman described. A split between on-device and cloud execution implies a routing layer that decides which requests stay local and which leave the home, and that decision carries latency, cost and privacy consequences. Work kept on device avoids a network round trip and keeps sensitive data off remote servers, but it is bounded by the memory, compute and power available in a consumer box. Work sent to the cloud can call on far larger models and fresher data, at the cost of bandwidth, recurring fees and a wider trust boundary. Systems that are granted credit card access also need explicit authorization flows, spending limits and audit logs, since an agent that can buy things is an agent that can be manipulated or make mistakes with real money. The security perimeter Feldman described is a reasonable design goal, but where it sits, and who holds the keys, is exactly the kind of choice that tends to determine whether users trust an assistant with payment credentials.
This suggests a product direction in which the assistant is less a single app and more persistent state about a person, their calendar, their contacts and their preferences. That framing raises its own questions: how that profile is stored, what happens when a user switches vendors, and how well small local models can perform the personal-context tasks that make such an assistant useful in the first place. The source does not offer timelines or product details, so the schedule for any of this remains a prediction rather than a roadmap.
Why it matters: Feldman's remarks are a signal about where consumer AI is heading from one of the people supplying compute to the industry, and they suggest that the home assistant may be built as a hybrid of local silicon and cloud services rather than purely one or the other. Developers building agents should expect that split execution, local data custody and delegated payment authority will be design constraints rather than afterthoughts, and infrastructure vendors stand to benefit if that architecture takes hold. How quickly it does, and whether users accept an assistant that can spend their money, is still an open question.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Cerebras CEO Andrew Feldman predicts a future where personal AI assistants become as commonplace as household appliances, running partly on local devices and partly in the cloud.