A $300 steel doughnut is a ridiculous object on which to wager the way forward for computing. That’s why Silicon Valley could also be critical about it.
OpenAI’s first client gadget, developed with former Apple design chief Jony Ive, will reportedly resemble a doughnut-shaped hockey puck sufficiently small to hold round the home in a single hand. It’ll have microphones, audio system, lights, cameras, environmental sensors and mechanical parts that transfer because the gadget listens and responds.
There can be no show. The {hardware} will as an alternative provide a extra succesful model of ChatGPT’s voice expertise, studying about its proprietor over time and finally finishing duties on their behalf. OpenAI is contemplating a worth between $300 and $400, with an unveiling anticipated earlier than gross sales start in 2027, in response to Bloomberg’s Mark Gurman. OpenAI has not confirmed these specs, so they continue to be superior product plans fairly than completed info.
On the floor, that is an costly good speaker coming into a class Amazon invented greater than a decade in the past. Beneath, it’s OpenAI’s try to seize management of the interface by means of which individuals talk with synthetic intelligence. The corporate doesn’t need ChatGPT to stay an app trapped inside {hardware}, working techniques and distribution channels managed by Apple, Google and Microsoft.
NEW: OpenAI’s gadget – a human-like good speaker –
will appear like a doughnut and be the dimensions of a hockey puck. It meant to be held. It has a digicam, audio system, microphones, lights and, most notably, transferring elements to indicate interactivity. It’ll be $300-400. https://t.co/gDxQU6FB8E— Mark Gurman (@markgurman) August 6, 2026
What OpenAI seems to be constructing
The reported gadget is battery-powered and designed to work in a number of positions: held within the hand, positioned on a kitchen counter or left beside a mattress. A hoop-shaped steel physique will comprise speaker grilles and microphones, whereas lights point out when the machine is listening.
Cameras and extra sensors will permit it to understand its environment. Its transferring parts are supposed to sign consideration, response and persona, making the article really feel much less like a stationary equipment and extra like a small bodily presence.
That final element is the place this stops sounding like a HomePod and begins sounding distinctly unusual. OpenAI is seemingly attempting to present the gadget a type of physique language. A tilt, rotation or mechanical motion might present that it has heard anyone, is contemplating a request or is directing its consideration towards an individual within the room.
Exactly how these parts work is unknown. The ring might accommodate a 360-degree microphone array, directional audio system or cameras able to observing the room from a number of angles. Motion may orient these sensors, enhance acoustic efficiency or just create emotional expressiveness. These potentialities are knowledgeable hypothesis, not disclosed options.
What is obvious is that OpenAI needs the machine to really feel alive with out making it look human. That could be a troublesome design transient. Too little persona and the product is one other Alexa. An excessive amount of and it turns into the haunted doughnut from a low-budget science-fiction movie.
The $6.5 billion design rent
OpenAI didn’t recruit Ive to enhance the speaker grille. In 2025, it acquired io, the {hardware} startup based by Ive and a gaggle of former Apple designers and engineers, in an all-stock transaction valued at roughly $6.5 billion. OpenAI already owned about 23% of io, which means it paid roughly $5 billion for the rest. Round 55 product designers, {hardware} engineers and manufacturing specialists joined OpenAI, whereas Ive’s unbiased LoveFrom studio assumed broad artistic tasks throughout the corporate.
The deal introduced collectively an unusually completed assortment of Apple alumni, together with former industrial-design chief Evans Hankey, product government Tang Tan and {hardware} specialist Scott Cannon.
Ive’s collaboration with Steve Jobs helped produce the iMac, iPod, iPhone, iPad and Apple Watch, whereas his affect prolonged to Apple’s software program, packaging, retail environments and headquarters. When Ive left Apple in 2019, Tim Prepare dinner stated his contribution to the corporate’s revival “can’t be overstated.”
The partnership with OpenAI is Ive’s first alternative since leaving Apple to design a wholly new computing platform fairly than a automobile, constructing, brand or luxurious object.
“I’ve a rising sense that every little thing I’ve discovered over the past 30 years has led me to this second,” Ive stated when the io deal was introduced. OpenAI’s accompanying letter described current computer systems as interfaces inherited from an earlier technological period—machines that haven’t been redesigned round techniques able to seeing, reasoning and understanding.
That’s the central thesis of the mission. Synthetic intelligence has modified dramatically, however we nonetheless entry it by means of keyboards, glass rectangles and working techniques whose fundamental metaphors have been created many years in the past. OpenAI and Ive imagine the interface is now the bottleneck.
Ive’s conflict towards the smartphone he helped create
There’s a pleasing historic symmetry right here. Ive helped flip the smartphone into crucial client object of the previous 20 years. He’s now serving to to assemble the argument that we spend an excessive amount of time taking a look at it.
The gadget he and Altman describe is meant to be calmer than a cellphone. It won’t bombard customers with grids of apps, banners, notifications and crimson circles demanding consideration. As an alternative, the AI will supposedly perceive context, resolve what issues and interrupt solely when mandatory.
At an Emerson Collective event, Altman contrasted the supposed expertise with the sensory chaos of Occasions Sq., evaluating the gadget as an alternative to sitting peacefully beside a mountain lake. Ive stated he prefers options that “teeter on showing nearly naive of their simplicity”—objects which are subtle internally however really feel unintimidating and nearly inconsiderate to make use of.
That sounds enticing as a result of fashionable smartphones have develop into casinos we faux are submitting cupboards.
However eradicating the display doesn’t take away complexity. It hides it.
A display permits customers to examine data, examine selections, right errors and perceive what a machine is doing. A voice-first gadget should talk all of that by means of dialog, sound, mild and motion. It should know when to talk, when to stay silent and when to confess that it has no concept what is going on.
These aren’t minor user-interface particulars. They decide whether or not the product feels magical or infuriating.
Sensible audio system promised this as soon as earlier than
OpenAI is just not the primary firm to vow a pc that disappears into the room.
Amazon launched the primary Echo in 2014, creating the fashionable smart-speaker class. Google adopted with Google Dwelling in 2016, pitching a conversational assistant able to answering questions, managing family duties and controlling related gadgets.
The merchandise bought effectively, however their ambitions contracted. Alexa and Google Assistant turned helpful for timers, music, climate stories and turning lights off after anyone had already climbed into mattress.
They by no means turned general-purpose computer systems as a result of they may not reliably perceive ambiguous requests, retain significant context or full sophisticated duties. The underlying assistants have been command techniques pretending to be conversational companions.
Giant language fashions change that equation. ChatGPT can interpret imprecise directions, ask follow-up questions and preserve far richer conversations than the unique Alexa. Multimodal fashions may interpret pictures, voices and environmental data, whereas agent techniques are starting to function software program and execute longer sequences of actions.
OpenAI is betting that the good speaker was not a foul concept. It merely arrived earlier than the intelligence required to make it helpful.
Which may be right. It could even be a sublime excuse for constructing Alexa once more, this time in anodized aluminum.
Why OpenAI wants {hardware}
Frontier fashions have gotten more durable to defend as standalone merchandise. Rivals replicate options, open fashions enhance and benchmark leads expire inside months.
As Gilded Age has argued in its examination of the AI compute wars, the sturdy benefit might not be the mannequin itself. It could be distribution, infrastructure, proprietary context and management over the system surrounding the mannequin.
{Hardware} provides OpenAI all 4.
A profitable gadget would set up ChatGPT because the default assistant with out requiring customers to open an app. It could generate distinctive sensory and behavioral knowledge unavailable to rival laboratories. It might create a brand new subscription enterprise whereas lowering OpenAI’s dependence on Apple and Google’s working techniques.
It could additionally give OpenAI management over all the expertise: microphones, processors, reminiscence, sensors, privateness controls and the bodily indicators by means of which the AI communicates.
Apple understood this mannequin higher than anybody. Its dominance got here not from possessing one of the best particular person element however from controlling the interplay between {hardware}, software program, companies and distribution.
OpenAI needs the identical vertical integration for synthetic intelligence.
The more and more hostile relationship between the 2 firms makes the technique specific. Apple has sued OpenAI and several other former staff, alleging that confidential {hardware} data and manufacturing methods have been misappropriated. OpenAI denies the accusations and says it’s growing one thing essentially totally different from Apple’s merchandise.
Regardless of the courts resolve, the broader battle is apparent. Apple owns probably the most invaluable consumer-computing platform in historical past. OpenAI believes that platform was designed for an period that’s ending.
What stays unknown
For all the thrill, crucial info concerning the gadget stay lacking. OpenAI has not revealed its title, ultimate look, battery life, processor, launch international locations or subscription mannequin. It has not defined how a lot computation happens regionally, whether or not it can perform with out an web connection or which outdoors companies it will probably management.
The absence of a show additionally raises sensible questions. How will the gadget current maps, paperwork, pictures or detailed comparisons? Will it ship visible data to a cellphone, tv or laptop? Can a number of relations use it with out their conversations and recollections bleeding collectively?
The seemingly structure is a hybrid one. Fundamental wake-word detection, sensor administration and privateness filtering might occur on the gadget, whereas demanding reasoning duties are despatched to OpenAI’s cloud. Working frontier-level fashions totally inside a hockey puck could be troublesome given battery, warmth and reminiscence constraints.
A recurring subscription additionally seems believable as a result of advanced voice conversations and agentic duties devour ongoing inference capability. OpenAI has stated nothing publicly about such a charge, nonetheless, and the $300-to-$400 worth could also be supposed to scale back resistance to a further service cost later.
The transferring parts are one other thriller. They could show useful, pleasant or irritating. Client-electronics historical past incorporates loads of motorized improvements that appeared enchanting in an indication and have become costly restore tickets two years later.
The smallest railroad of the AI Gilded Age
The industrialists of the unique Gilded Age fought to manage railways, oil pipelines and electrical grids. The brand new barons are preventing over the infrastructure connecting human intention to machine motion.
Google has Search, Android and Workspace. Apple owns the iPhone and its working system. Amazon has Alexa and the business equipment behind it. Meta is putting AI inside glasses. Microsoft controls the office desktop.
OpenAI at the moment has intelligence and a focus, but it surely doesn’t personal the bodily doorway.
The doughnut is its try to construct one.
The primary gadget in all probability won’t exchange the smartphone. Altman has acknowledged as a lot, simply because the smartphone didn’t remove the laptop computer. A extra believable end result is that it makes the cellphone barely much less central: one thing retrieved for visible work fairly than consulted compulsively all through the day.
That might nonetheless be a profound shift. The corporate controlling the assistant individuals converse to earlier than touching a display would occupy a very highly effective place between want and motion.
OpenAI is subsequently not likely making ready to promote a wise speaker. It’s asking customers for permission to put ChatGPT within the room, give it eyes and ears, educate it their routines and permit it to behave on their behalf.
For $300, the steel doughnut might look small. The declare it intends to stake is gigantic.
Alex Chen Alex Chen Read More
