Cerebras CEO Andrew Feldman thinks personal AI assistants running inside your home are closer than most people expect. Asked on stage at TechCrunch Disrupt 2026 what it would take for AI to become a home appliance, he said: “Actually, I don’t think that’s very far away.” TechCrunch AI reports that Feldman pictured a hybrid setup. Your assistant would do some work locally and some in the cloud, it would know you well, and it would be able to spend your money.
That last part matters more than it sounds.
🏠 What Feldman Actually Predicted
Feldman wasn’t describing a smarter speaker. He described an agent with real autonomy. Here’s what he said these assistants will have, according to TechCrunch AI:
- Split processing: “Just like your cell phone, they’ll do some work on device and some work going back to the cloud.”
- Deep personal context: “They will get to know you very well.”
- A security perimeter: He thinks the data and actions will be walled off around you.
- Purchasing power: “They will certainly have access to credit cards and be able to buy things for you.”
He pointed to consumer agent products like Meta’s Muse and Instinct as signs that agentic AI is gaining real traction with everyday users. TechCrunch notes that Instinct is now valued at $10 billion.
He also uses this stuff himself. Feldman runs “a collection of agents” that sweep his calendar and pull background on the people he’s about to call. That doesn’t prove the home-appliance future is coming. It does show that the executive workflow version already exists, and consumer products usually trail what power users rig up for themselves.
🧐 Consider the Source
Keep in mind who’s talking. Cerebras sells AI chips, so a world where every household runs AI workloads around the clock means a much bigger market for compute. Feldman wants this future to happen.
That doesn’t make him wrong. It does mean you should read his timeline as optimistic. Hardware CEOs have a long history of calling the consumer moment early. Smart home hubs, VR headsets, and voice assistants all got “right around the corner” predictions years before they reached the mainstream, and some never did.
Feldman was open about his role, too. “One of the reasons I love being in infrastructure is because I don’t need the consumer-facing idea,” he said. “I like building the platforms that other people’s ideas take flight on.” Translation: he’s betting on the category and leaving it to someone else to find the killer app.
🔮 The Next 1-3 Years
One part of his vision is already happening. Hybrid on-device and cloud processing is the direction Apple, Google, and Samsung have taken with their phones. Moving that model into a dedicated home device is a smaller step than it looks.
The harder parts are trust and liability. An agent that “knows you very well” and holds your credit card raises questions nobody has fully answered:
- Who pays when an agent makes a bad purchase? Card networks and banks will need new rules for authorizing agent transactions.
- What does the “security perimeter” actually mean? Local processing helps with privacy, but anything that touches the cloud can leak.
- Will people hand over control? Agents that check your calendar are low stakes. Agents that spend money are a different psychological leap.
My expectation: the first wave won’t be a new box on your counter. It’ll be agent features added to devices you already own, like phones, TVs, and smart displays, with spending limits and confirmation prompts. A real standalone “AI appliance” probably needs another hardware cycle.
🛠️ What to Do Now
- Builders: Design for hybrid inference from day one. Decide which tasks have to run locally for privacy or speed and which can go to the cloud.
- Commerce and fintech teams: Start planning for agent-initiated transactions. Your checkout flow may soon serve a bot acting on someone’s behalf.
- Everyone else: Try what Feldman does. Set up a couple of agents for calendar prep or research. You’ll learn quickly where the trust gaps are.
The infrastructure players are making their bets. Whether consumers actually let an AI hold their wallet is still the open question. You can find Feldman’s full comments in TechCrunch AI’s coverage from Disrupt 2026.