The device that OpenAI has spent the past year building with Jony Ive is, by most accounts, a smart speaker — no screen, no wearable ambitions, nothing that needs to be strapped to a lapel. According to Bloomberg journalist Mark Gurman, it is shaped like a doughnut, roughly the size of a hockey puck, runs on a battery and is meant to be picked up with one hand and carried from the kitchen counter to the bedside table. It is expected to be priced somewhere between $300 and $400 and to reach buyers in 2027.
That description will disappoint anyone who expected the man behind the iPhone, iPad and Apple Watch to arrive with a wholly new object. The category OpenAI is stepping into is a decade old and thoroughly occupied by Amazon and Google. What the company appears to be betting on instead is behaviour: not what the device does, but how it makes people feel.
The clearest sign of that is the mechanics. The circular body will contain parts that move on their own, signalling when the device is listening, thinking or replying. Lights will indicate when the microphones are live. A camera system and an array of sensors will read the room and pass what they see to the model. The stated intent, per Gurman’s sources, is to make the object seem less like a stationary appliance and more like something with a pulse.
Underneath, it is expected to run an advanced version of ChatGPT’s voice mode — the one that listens and speaks at the same time, allowing a user to cut in mid-sentence without the system losing its place. Over time it is designed to accumulate a picture of the person using it and adjust its conversational manner accordingly. OpenAI is understood to be building it from high-grade metal, and the finish is meant to feel expensive.
Whether the market rewards any of this is a separate question. Smart speakers have been a commercially awkward business for years, sold cheaply in the hope of capturing habits that never quite materialised. Amazon’s own range starts around $40 and tops out near $240. The devices were pitched as conversational companions and ended up as timers, light switches and music players, largely because the underlying voice technology stalled well short of what the marketing promised.
Large language models have begun to change that, though not painlessly. Rebuilding an assistant around a model that is slow, costly to run and occasionally wrong — without breaking the millions of small functions users already depend on — has taken Amazon and Google the better part of two years. Both have now shipped upgrades, Alexa+ and Gemini for Home, that handle natural speech, follow-up questions and strung-together commands. Apple has yet to release its own overhauled Siri, though it is reported to be developing a home hub whose display sits on a robotic arm and can turn to face whoever is speaking.
So OpenAI arrives late rather than early, which is roughly the position Apple occupied when it entered a touchscreen phone market that already existed. The counter-argument is that the acquisition of Ive’s startup cost OpenAI $6.5 billion, and the company’s record with adjacent products — a browser, a video tool, a shopping feature — has been mixed enough to make a hardware misfire expensive.
Keeping the device indoors is the sensible half of the strategy. Startups including Friend and Limitless have tried the same idea as a pendant worn in public and run into the discomfort that sank Google Glass: talking to a machine in the street is odd, and a live microphone on someone else’s chest is unwelcome. A speaker on a shelf carries none of that. If it earns enough familiarity there, a pocketable version becomes plausible later.
The uncomfortable half is the one OpenAI has to address at some point. A device engineered to feel alive is, by design, a device engineered to be bonded with. A small proportion of ChatGPT users develop emotional dependence on the chatbot, and that the risk is concentrated among heavy users of advanced voice mode. Ive has spoken publicly about his regret over the compulsive quality of the smartphone. His current employer is building a product whose competitive advantage is attachment.
Amazon and Google are still selling usefulness. Apple has been explicit that Siri is a utility rather than a friend. OpenAI is going the other way — and the legal fight it is currently having with Apple over trade secrets, which OpenAI denies, is the smaller of its problems.
