Jony Ive's First OpenAI Gadget Is a Hockey Puck-Sized Smart Speaker

OpenAI and former Apple design chief Jony Ive are building a hockey puck-sized, doughnut-shaped smart speaker — no screen, battery-powered, and aimed at an early 2027 release — priced at or above $300, according to reporting from Bloomberg's Mark Gurman. Multiple independent sources have been building toward this picture since at least November 2025, when Sam Altman confirmed a working prototype existed. What separates this device from every smart speaker that came before it is a detail that sounds almost theatrical: parts of it physically move, with lights, to signal when the AI is responding.

What's Converging

The bet behind this device is that the smartphone-plus-app interaction loop is the wrong interface for conversational AI — that unlocking a phone, opening an app, and typing or speaking a query is a workaround, not a destination. That thesis has been gaining traction for roughly two years, and several companies are now racing to own the physical space where ambient AI settles: a countertop, a nightstand, the corner of a home office desk.

Apple is the most direct comparison. The company is expected to ship a camera-equipped home hub this year — a fixed device built for video calls and smart-home control, with deep integration into the updated Siri. The timing of that product and the OpenAI speaker is not coincidental; both companies are staking out the same physical territory, just with different philosophies about what a home AI device is actually for. Apple's hub is a command center. The OpenAI device, based on everything reported so far, is designed to recede into the background and pay attention.

OpenAI moved to lock in design credibility early. In May 2025, the company acquired io, the hardware studio Ive had founded after leaving Apple. That acquisition brought Ive's entire team — and his reputation, built across decades of iPhone, iPad, and MacBook design — inside OpenAI's hardware ambitions. By November 2025, TechSpot reported that Altman and Ive appeared together at Emerson Collective's Demo Day alongside Laurene Powell Jobs, where Altman publicly confirmed the team had completed a first working prototype, describing it in terms of joy, simplicity, and a deliberately peaceful quality. That event was clearly a calibrated signal, not a casual disclosure.

The Specific Development

The current picture comes primarily from Bloomberg's Gurman, whose hardware sourcing is among the most reliable in the industry, corroborated by earlier reporting from The Information in February 2026. The form factor is compact — small enough to carry one-handed around a home — and battery-powered rather than tethered to a wall. The doughnut shape and moving parts are the details that stand out: rather than an LED ring or a static light bar, the device apparently has components that physically reorient to communicate the AI's state. That is an unusual engineering investment for what might otherwise be a commodity form factor, and it signals that the product team is thinking carefully about what "AI presence" means without a screen.

On price, the sources diverge slightly in a way worth noting. The Verge's recent reporting puts the figure above $300. The Information's February 2026 reporting, as covered by MacRumors, cited a range of $200 to $300 with a launch no earlier than February 2027. The gap between those two figures likely reflects either different stages of pricing discussions or costs rising as features were added — MacRumors' sourcing added a camera with facial recognition comparable to Face ID, plus the ability to make purchases through the device. Those are not budget features. Either way, every report points to a premium, early-adopter price tier rather than a mass-market entry.

The camera and sensing layer deserve attention. Multiple sources confirm the device is designed to learn who is using it and what is happening around it. Built In's coverage notes plans for the speaker to connect with existing computers and smartphones, maintaining a continuous memory across user activity rather than treating every interaction as stateless. TechSpot's earlier prototype reporting mentioned the possibility of a small projector for displaying information on nearby surfaces — though it is unclear whether that feature survived into the current design. What the voice interaction model runs is described as a more advanced, humanlike version of ChatGPT voice mode, purpose-built for ambient, persistent conversation rather than ported from the phone app experience.

Our read is that the purchase capability is the detail most worth watching. A screenless home device that can facilitate transactions is not just a smart speaker — it is a commerce endpoint sitting in your kitchen. That positions OpenAI as a potential layer between users and purchases in a way that Apple, Amazon, and Google have all tried and failed to make feel natural. Whether OpenAI can thread that needle depends entirely on whether the persistent-context layer actually works at scale across real households.

What's Likely Next

If a February 2027 target is real, production timelines would require OpenAI to lock specifications within the next few months. That makes a formal hardware announcement — likely a dedicated event featuring Altman and Ive together — probable before the end of Q4 2026. Watch for any indication of manufacturing partnerships, which have not surfaced in reporting yet but would be a necessary disclosure as production ramps.

Beyond the speaker, MacRumors' sourcing confirmed at least two additional products in the pipeline: a smart lamp and AI glasses, both described as targeting 2028 or later. Those products exist to turn the speaker from a standalone device into a platform — distributed physical nodes sharing the same AI memory and context layer. That is the actual strategic play here, and it is more significant than any single device's specs. For developers building on voice or ambient interfaces, the relevant question is not whether the first device ships on time but whether OpenAI builds a hardware API surface with the same openness as its software APIs. That answer will not come in 2026, but the shape of it should be visible by the time the speaker launches.

Sources