OpenAI's First Hardware Device Is a $400 Hockey Puck That Wriggles. Jony Ive Designed It. Apple Is Suing Over It.

For over a year, the tech industry has known that OpenAI was building hardware. The company acquired Jony Ive's io design studio for $6.5 billion in 2025 — the largest acquisition in OpenAI's history. It hired Tang Tan, the Apple executive who oversaw the physical design of iPhone, AirPods, and Mac for nearly two decades, as its Chief Hardware Officer. It partnered with Foxconn for manufacturing. What it had not revealed, until Bloomberg's Mark Gurman published detailed design specifications on August 6, 2026, was what any of this was actually building.

The answer: a doughnut-shaped smart speaker roughly the size of a hockey puck, battery-powered, displayless, portable enough to carry around the home with one hand, priced between $300 and $400, and equipped with moving mechanical parts designed to give it personality. The device launches in 2027. It runs on GPT-Live — the full-duplex voice architecture OpenAI deployed to more than 150 million weekly voice users in July 2026. And it has already triggered a federal lawsuit from Apple, which filed a 41-page trade-secret complaint alleging OpenAI's hardware effort was built on stolen design knowledge.

This is not an Amazon Echo with ChatGPT bolted on. Whether it is the future of consumer AI interaction — or a very expensive mistake — is the central question of OpenAI's move into hardware.

What the Device Actually Is

OpenAI's upcoming device will be shaped like a doughnut, roughly the size of a hockey puck, easy to carry around the home in one hand, with moving parts that help give it personality, speaker grilles and microphones for voice-based chats, a camera system and sensors to feed information about the user's environment, and a battery so it can move from room to room. There is no display. No touchscreen. No physical buttons described in the reporting. The entire interaction model is conversational — voice in, voice out, with the device's mechanical motion and lighting system serving as the visual feedback layer.

The moving parts are the design element that has generated the most reaction. The animated movement is intended to help the device communicate emotional responses, acknowledge user requests, or provide visual cues during conversations — distinguishing it from existing voice assistants that primarily rely on audio responses alone. Think of the mechanical motion as a kind of body language for an AI that has no face: an acknowledgement nod, a listening lean, an animating response to a question. Whether that reads as charming and alive or uncanny and uncomfortable in a real home environment is something no amount of controlled demonstration can answer. That is what the 2027 launch will determine.

The construction is premium by design. The device will use high-quality metal components, with design features that position it apart from existing smart speakers on the market — more expensive than typical smart speaker offerings but less costly than most standard smartphones. This is deliberate pricing strategy: OpenAI cannot afford to subsidise hardware the way Amazon did with the Echo. Amazon lost more than $25 billion on the Echo ecosystem over its life, using the device as a loss-leader to drive Prime subscriptions and shopping. OpenAI has no retail ecosystem to cross-subsidise a cheap speaker. The $300–$400 price is the actual margin it needs to charge.

The Engine — GPT-Live Changes What a Smart Speaker Can Do

The most important thing to understand about OpenAI's device is that it is not a smarter Alexa. It is a fundamentally different product category enabled by a fundamentally different voice architecture. The device runs on GPT-Live, OpenAI's full-duplex voice architecture — the same engine that replaced ChatGPT's turn-based Advanced Voice Mode for more than 150 million weekly voice users in July 2026.

The distinction between full-duplex and turn-based voice interaction matters enormously for what a device can actually feel like in daily use. Turn-based voice assistants — Alexa, original Siri, Google Assistant — operate in a speak-then-wait pattern: you talk, the device processes, the device responds, then waits for you to talk again. The interaction is sequential, clearly delineated, and feels like operating a machine. Full-duplex voice, like GPT-Live, processes and responds simultaneously. You can interrupt mid-sentence. The AI can react to tone and hesitation. Conversations can overlap and flow the way human conversations do rather than alternating in turn. The AI model used will have human-like responses and will learn more about the user over time, making interactions feel more personal and relevant over extended use.

This is the commercial argument for a $400 price point: the device is not competing with a $50 Echo Dot. It is competing with a new category of AI-native interaction that happens to take physical form in your home. The question is whether consumers who already have ChatGPT Voice on their phone will pay $400 for a dedicated device to do the same thing — and whether the physical presence, the moving parts, the premium audio, and the always-available ambient awareness are worth the difference.

Jony Ive — The $6.5 Billion Design Bet

The hardware's design lineage runs directly through Jony Ive — the British designer responsible for the original iMac G3, the iPod, the iPhone, the MacBook Air, and effectively the entire physical identity of Apple from 1997 to 2019. The device is being developed in partnership with LoveFrom, the design studio Ive founded after leaving Apple, whose io hardware startup OpenAI acquired for $6.5 billion in 2025.

The io acquisition was the moment OpenAI's hardware ambitions became undeniable. A $6.5 billion acquisition for a design studio with no shipping products is not a talent hire. It is a statement of intent — a bet that physical design will be as important to AI's consumer future as model capability, and that the person who shaped how a billion people interact with technology through their hands should be the one designing how they interact with AI through their voice and their home.

Ive's involvement also explains the premium material choice and the unusual form factor. A hockey puck-sized donut with moving parts and metal construction is not a device designed by an algorithm optimising for minimum cost of goods. It is a device designed by someone who spent three decades believing that how something feels in your hand changes how you feel about the thing you're doing with it.

The Apple Lawsuit — 41 Pages of Trade Secret Allegations

The most legally significant dimension of OpenAI's hardware story is not the device itself. It is the lawsuit surrounding how that device came to exist. Apple filed a 41-page federal trade-secret complaint against OpenAI on July 10, 2026, alleging that Chief Hardware Officer Tang Tan directed Apple engineers to bring physical hardware components to job interviews at OpenAI — effectively using the hiring process to extract Apple's confidential design knowledge and manufacturing techniques.

Tang Tan spent nearly two decades at Apple overseeing the physical design and engineering of iPhone, AirPods, Apple Watch, and Mac hardware. He left Apple in 2025 to join OpenAI as its Chief Hardware Officer. Apple's complaint argues that the move was not merely a talent departure but a deliberate extraction of proprietary knowledge — that Tan and others recruited from Apple used confidential information about Apple's unreleased hardware, manufacturing processes, and supply chain relationships to accelerate OpenAI's device development.

OpenAI rejected the accusations, maintaining that their new device is fundamentally different from any product previously offered by Apple and that no proprietary Apple technology has been incorporated. The "not something Apple has come close to launching" line in Bloomberg's report — almost certainly placed by OpenAI's communications team — is the company's public counter-narrative: not only did we not steal Apple's ideas, but we built something Apple hasn't even conceived of.

The lawsuit remains active and unresolved. Its outcome could delay the device's 2027 launch if a court issues an injunction, or it could proceed to trial without affecting the launch timeline. What it cannot do is unring the bell: OpenAI is in the hardware business, the device exists, and the design fight with Apple has been filed in federal court for the public record.

Manufacturing — Foxconn, Vietnam or the US, and Federal Requirements

Foxconn has been tapped to handle manufacturing for the speaker in Vietnam or the United States — a supply chain decision driven partly by federal security requirements tied to OpenAI's government infrastructure contracts. The Vietnam-or-US decision reflects the same geopolitical pressure reshaping supply chains across the consumer electronics industry: the US government's expanding scrutiny of Chinese manufacturing in devices that interact with sensitive data and are connected to AI cloud services. An AI speaker with a camera, microphone array, and continuous cloud connectivity is exactly the kind of device that federal buyers would not want assembled in China. OpenAI's significant government contracts — including Stargate involvement and federal AI infrastructure partnerships — create real compliance pressure around manufacturing geography.

Foxconn's involvement also signals manufacturing volumes the company is confident it can achieve. Foxconn does not take on early-stage contract manufacturing for unproven consumer devices. Its willingness to partner here suggests OpenAI has demonstrated enough pre-production confidence — in the design, the supply chain, and the demand projections — to commit a serious manufacturing operation to the launch.