OpenAI's first consumer hardware device will be a moveable, screenless, rechargeable speaker designed to function as a humanlike AI companion, Bloomberg reported July 14, citing people familiar with the project Bloomberg. Reuters wire copy corroborated the report the same day KFGO.

The device has no display. Instead, it relies on mechanical elements that move on their own, an approach intended to create what Bloomberg's sourcing describes as an illusion of being "alive" Engadget. It can talk naturally, control smart home hardware, and play audio. A built-in camera and additional sensors are meant to give the assistant situational awareness of a user's surroundings, informing more personalized responses.

Under the hood, the speaker will run a more advanced version of GPT-Live, the real-time voice model family OpenAI launched July 8 that can listen and speak simultaneously Reuters. That model already underpins the newer voice mode in ChatGPT; the hardware variant is described as a further iteration rather than a wholesale departure.

On pricing and timing, the picture is layered. Reuters reported in February that OpenAI's first device — then referred to generically as a smart speaker — was likely to land between $200 and $300, citing two people with knowledge of the matter Reuters. Bloomberg's more recent reporting puts a release window on the device for sometime in 2027, while cautioning that the timeline could slip Engadget.

Design work has run through io, the startup founded by former Apple chief design officer Jony Ive, which OpenAI acquired in 2025 for $6.5 billion. Paul Meade, who previously led development of the Apple Vision Pro, now runs OpenAI's hardware division.

That personnel trail sits at the center of an active lawsuit. Apple has sued OpenAI along with two former Apple employees, Chang Liu and Tang Yew Tan, alleging trade secret theft, and is seeking a court injunction against OpenAI's hardware products. Apple's complaint accuses the pair of downloading dozens of confidential files covering technical specifications and proprietary data tied to unreleased Apple products. The suit also names io as complicit in the alleged theft. Separately, Apple has claimed OpenAI has hired more than 400 of its former employees.

None of these legal claims have been adjudicated, and OpenAI has not been reported as conceding the underlying allegations. The injunction request, if granted, could directly affect the 2027 timeline Bloomberg's sources described, though nothing in the current reporting indicates a ruling is imminent.

The device sketched out here reads less like a smart speaker in the Alexa or Google Home mold and more like an attempt to build a physical anchor for the conversational voice interface OpenAI has been assembling since GPT-Live's launch. A camera and ambient sensors pushing context into a model that already handles duplex speech is a materially different proposition than a cylinder that wakes on a keyword and streams to a cloud backend. It's closer, in spirit, to the companion-robot demos that have circulated at CES for a decade than to a HomePod competitor — though whether the market wants an object with moving parts simulating aliveness, rather than a screen it can glance at, is untested territory.

Worth flagging: the hiring numbers and design pedigree involved here mirror, in reverse, exactly what made Apple's own hardware culture effective for two decades — deep industrial design talent paired with a tightly controlled first-party silicon-and-software stack. OpenAI assembling that same combination, with a former Vision Pro lead and Ive's studio, is presumably why Apple's legal filing frames the matter in such pointed terms rather than treating it as routine attrition.

The bigger open question is architectural. A screenless device that leans entirely on voice and ambient sensing to be useful is a bet that natural-language interaction has matured enough to replace glanceable UI for a meaningful set of daily tasks — timers, smart home control, ambient conversation. GPT-Live's simultaneous listen-and-speak capability is a necessary but not sufficient condition for that bet to pay off; latency, false-wake handling, and privacy posture around an always-on camera and microphone array will determine whether consumers treat this as a genuine companion or as another gadget that ends up in a drawer.

For now, what's confirmed is narrower than the ambition suggests: a named model, a named design partner, a rough price band from one report, an approximate 2027 window from another, and an unresolved legal dispute over how some of the underlying talent got there.