Thursday, October 15, 2026

AI & Models

Cerebras CEO: home AI assistants are closer than people think

Cerebras Systems CEO Andrew Feldman said at TechCrunch Disrupt 2026 that AI assistants running locally in the home are closer to reality than most people expect.

A Google Home Mini smart speaker on a wooden shelf, blending technology with home decor.
Photo: Pexels / John (Giannis) Tekeridis

Asked on stage what would need to happen for AI to become a home appliance, Feldman said, “Actually, I don’t think that’s very far away.” He pointed to advances in consumer AI agents — citing Meta’s Muse and Instinct, the latter now valued at $10 billion — as evidence that AI is already finding traction with everyday users.

Feldman predicted personal AI assistants will run locally, splitting work between the device and the cloud the way a phone does today. “Just like your cell phone, they’ll do some work on device and some work going back to the cloud,” he said, adding that he expects these assistants to get to know users well, operate inside a defined security boundary, and eventually have access to credit cards to make purchases on a user’s behalf.

As the CEO of an AI chipmaker, Feldman has an obvious interest in both the hardware infrastructure for that vision and broader consumer adoption of agentic AI. He’s also a personal AI user himself, saying he relies on a collection of agents to sweep his calendar and pull up background on people before calls — small, already-possible tasks he says are being automated now, even as he’s most excited about use cases that weren’t possible before AI. He said he’s content leaving those new ideas to other founders: “One of the reasons I love being in infrastructure is because I don’t need the consumer-facing idea … I like building the platforms that other people’s ideas take flight on.”

Why it matters

A chip supplier predicting imminent home-AI adoption has a direct financial stake in that prediction coming true, but it also signals where infrastructure investment is already pointed — toward devices that do real inference locally, not just in the cloud.