OpenAI’s first consumer hardware device will be a donut-shaped, hockey-puck-sized smart speaker with mechanical moving parts, an onboard camera, and a price of $300 to $400 — and the engineering reason that price is not a mistake is the same reason Amazon lost more than $25 billion on the Echo. Bloomberg’s Mark Gurman published the first detailed description of the device on August 6, adding physical form, moving parts, and pricing specifics to earlier reporting that had established the device would run 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 TechCrunch device coverage confirmed what earlier coverage had outlined: this is not a conventional smart speaker with ChatGPT bolted on, but the first consumer hardware expression of an entirely new voice architecture.
The device is battery-powered, screenless, designed to move from room to room, and positioned as an AI-first computer rather than a smart speaker with bolt-on intelligence. Its ring-like shape is built for portability — small enough to carry in one hand between a kitchen counter and a nightstand. Mechanical components shift autonomously while the device listens and responds, giving it what OpenAI internally calls a sense of “aliveness.” LED indicators signal listening and response states. A camera system and environmental sensors feed real-time visual context into OpenAI’s multimodal models, allowing the AI to incorporate its surroundings into conversation.
The device is being developed in partnership with LoveFrom, the design studio founded by Jony Ive after he departed Apple in 2019. OpenAI acquired Ive’s hardware startup, io Products, in May 2025 in an all-stock deal valued at approximately $6.5 billion — the company’s largest acquisition to date. Tang Tan, a 24-year Apple veteran who most recently served as VP of product design for the iPhone and Apple Watch, is now OpenAI’s Chief Hardware Officer. Apple filed a federal trade-secret suit against him and OpenAI on July 10, 2026.
What GPT-Live Actually Does Inside the Device
The technical story of this speaker starts not with Jony Ive’s aesthetic choices but with a July 8, 2026 engineering announcement that most consumer coverage overlooked: the launch of GPT-Live, a rebuilt voice architecture that is structurally different from everything that came before it.
Smart speakers, including every Amazon Echo, Google Nest speaker, and Apple HomePod, have historically used what engineers call turn-based conversation: the device waits for one party to finish speaking before processing and responding — modeled on a walkie-talkie, not a phone call. ChatGPT’s own earlier Advanced Voice Mode worked the same way, chaining three separate models — speech-to-text, language model, text-to-speech — into a pipeline with latency at each handoff.
GPT-Live differs at the architecture level entirely: it uses full-duplex processing — the device listens and speaks simultaneously, continuously, without waiting for turn boundaries. The technical term telecom engineers use is full-duplex: the same bidirectional simultaneity a telephone call has had since the 20th century, now applied at scale to a large language model. OpenAI described the core design principle in its GPT-Live engineering blog: it built GPT-Live for continuous interaction using a full-duplex architecture, processing input while generating output rather than handling a sequence of separate messages.
GPT-Live also separates the conversational layer from the reasoning layer in a way prior systems did not. Two models run simultaneously: GPT-Live-1 handles the low-latency voice conversation, generating responses and acknowledgment cues (“mhmm,” “got it”) while the user is still talking. In the background, GPT-5.5 takes over when a question requires deeper reasoning — running asynchronously so the conversation does not pause while the hard thinking happens. This delegation architecture is the piece that could prove most durable: OpenAI can upgrade the intelligence layer without rebuilding the voice experience each time.
The transport infrastructure was rebuilt to match. OpenAI rewrote its media front-end in Go, replacing a Python asyncio implementation. Its proprietary WARP (WebRTC Abridged Roundtrip Protocol) reduces voice session startup from six network round-trips to a single UDP packet, and its Instant Connect system can begin a voice session with that single packet. According to OpenAI’s engineering blog on continuous voice interaction, the rewrite brought 95th-percentile frame-delivery performance to where the old system’s median had been.
Bloomberg confirmed the device runs advanced GPT-Live, a more capable version of ChatGPT Voice Mode, meaning the speaker is not a ChatGPT interface dropped into a puck-shaped body. It is a dedicated hardware expression of a full-duplex, dual-model, WebRTC-transported voice system, with a camera and environmental sensors feeding real-world visual context into what GPT-Live’s delegation layer sends to GPT-5.5.
What the Price Actually Buys — and What Amazon’s $25 Billion Teaches
The $300 to $400 price point for OpenAI’s speaker has generated predictable skepticism. For context, Amazon Echo prices range from $40 to $240, and the Apple HomePod — which lacks a camera — sits at $299. Against that backdrop, $400 reads as premium hubris.
It is worth understanding why it is not.
Amazon’s devices division lost more than $25 billion between 2017 and 2021, according to internal documents obtained by the Wall Street Journal. The Alexa division alone was projected to lose approximately $10 billion in 2022, according to the same reporting on Amazon’s device losses. The failure had a structural cause, not a product quality cause: Amazon sold Echo devices at or below cost, on a theory that the devices would drive shopping, Prime subscriptions, and ecosystem spending. That theory failed. Most Echo users used Alexa to set timers and play music — tasks that cost Amazon compute money on every query and generated no revenue in return.
The underlying economics of a full-duplex, camera-equipped, always-contextual AI speaker are different from a wake-word speaker in a way that makes the Amazon model not just risky but arithmetically impossible. A full-duplex system processes continuous audio, not discrete queries. A camera feeding visual context into a multimodal model generates inference requests that are computationally heavier than “play music.” If OpenAI sold this device at $40 to gain market share, every minute of active use would generate cloud inference costs that no downstream shopping behavior could offset, because the ChatGPT subscription revenue model is separate from any retail commission structure Amazon was trying to build.
At $300 to $400, with the near-certain expectation of a required ChatGPT Plus subscription for full functionality, OpenAI is attempting something Amazon never tried: charge enough for the device and the service that the business model works. Whether consumers will pay that is the actual open question. Whether the price makes sense given the architecture is not.
Is the Smart Speaker Market Ready to Be Disrupted?
The reaction among tech observers has ranged from enthusiasm to pointed skepticism. The camera, the moving parts, the premium price, and the always-on architecture add up to a device that is genuinely different from existing products — but that difference is not uniformly appealing, as MacRumors reader forum reactions among consumers has made clear.
The privacy question is the most structurally important one. An always-listening, camera-equipped device that processes ambient conversation without a wake word is, in capability terms, a more comprehensive home sensor than an Echo or HomePod. OpenAI has not disclosed its specific data retention and processing policies for the device. What it sends to GPT-Live servers, what it logs, what it retains in user profiles, and whether it can be meaningfully audited by consumers are questions that will determine whether the device generates adoption or backlash on launch.
The competitive landscape is also not static. Google’s Gemini Live supports bidirectional voice with native audio-to-audio processing. Apple’s iOS 27 Siri overhaul includes conversational AI capabilities and third-party AI extensions. Amazon has been rebuilding Alexa with generative AI since 2024. OpenAI is entering a market where its three most direct hardware competitors have billions of dollars of installed base and years of manufacturing relationships.
What OpenAI has is the GPT-Live architecture, 150 million weekly voice users already on the platform, and the only full-duplex delegation design that ships at consumer scale. Whether those advantages translate from a cloud service into a $400 puck on a kitchen counter is the question the 2027 launch will answer.
The device is part of a broader hardware program. OpenAI envisions a lineup of AI-native devices beyond the donut speaker, with Foxconn 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, as covered in TechTimes’ supply chain analysis.
Apple’s Lawsuit and What It Actually Threatens
One element of the story that the August 6 report touched but did not resolve: 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 and that a former Apple engineer exploited an authentication vulnerability to download confidential files after leaving the company. The complaint specifically alleges that OpenAI approached a shared contract manufacturer using Apple’s proprietary metal-finishing processes — directly relevant given the speaker’s described premium metal construction.
The practical legal risk is not an immediate launch ban. California courts have explicitly rejected the “inevitable disclosure” doctrine that would allow Apple to seek an injunction simply because its former engineers now work at OpenAI. What Apple can realistically pursue is targeted preliminary relief — evidence preservation orders and use-cessation certifications — that would force an internal audit without stopping development. Bloomberg Intelligence assessed that targeted preliminary relief, rather than a product ban, is the more likely near-term outcome, as reported in coverage of the Apple v. OpenAI lawsuit.
The lawsuit’s most immediate damage is operational: a chilling effect on Apple-to-OpenAI recruiting, supply chain scrutiny for manufacturers with Apple relationships, and a material litigation-risk disclosure requirement in OpenAI’s forthcoming IPO prospectus. OpenAI filed a confidential S-1 with the SEC on June 8, 2026, targeting a public listing. Apple’s complaint language — calling OpenAI’s hardware division “rotten to its core” in its filing — is precisely the kind of allegation that must be addressed in an S-1’s risk factors section, as detailed in Apple’s federal trade-secret complaint.
For a fuller account of the lawsuit’s legal framework and operational consequences, TechTimes has covered both the initial complaint and its downstream effects in depth.
How Altman’s Platform Strategy Drives the Hardware Push
Sam Altman’s testimony before the US Senate Judiciary Subcommittee on Privacy, Technology and the Law in May 2023 established him as a public advocate for AI licensing frameworks — a position he subsequently revised as regulatory proposals tightened. By late July 2026, Altman was telling investors in a podcast interview that AI development may need pacing to avoid outpacing society’s readiness, signaling a shift toward voluntary governance frameworks.
The hardware push exists inside this policy context. OpenAI’s hardware strategy is not primarily about speaker revenue — it is about owning the physical access point through which users interact with AI, rather than remaining dependent on Apple’s App Store, Google’s Android ecosystem, and the carrier networks those companies control. A speaker that runs GPT-Live natively, in a user’s home, without routing through any competitor’s platform, is the first step in a distribution strategy that does not require Apple or Google’s permission.
The irony that OpenAI’s two most prominent hardware hires — Ive and Tan — both came from the company now suing it is not lost on observers. Whether that legal entanglement delays, reshapes, or ultimately has no effect on the device’s launch will be the hardware story of 2027.
Frequently Asked QuestionsWhat is GPT-Live, and why does it matter for the OpenAI speaker?
GPT-Live is OpenAI’s full-duplex voice architecture, launched July 8, 2026, that replaced the turn-based Advanced Voice Mode in ChatGPT. Unlike prior voice systems — including every major smart speaker on the market — GPT-Live processes incoming and outgoing audio simultaneously, the way a telephone call works, rather than waiting for one side to finish before responding. It runs as a two-layer system: a low-latency conversational model (GPT-Live-1) handles the ongoing dialogue, while GPT-5.5 takes over in the background for complex reasoning tasks without pausing the conversation. Bloomberg confirmed the OpenAI device will run a more advanced version of this system, meaning the speaker is a dedicated hardware expression of GPT-Live’s architecture — not a conventional smart speaker with a chatbot added.
How does the OpenAI speaker’s $400 price compare to competitors, and is it justified?
Amazon’s Echo smart speaker pricing ranges from $40 to $240; Apple’s HomePod (without a camera) costs $299. OpenAI’s $300–$400 range is premium by smart speaker standards. But Amazon lost over $25 billion on its devices division between 2017 and 2021 selling hardware below cost, with Alexa projected to lose approximately $10 billion in 2022 alone — the largest loss of any Amazon unit. The architectural reason is that cloud inference costs for an always-on, full-duplex, camera-equipped AI system cannot be cross-subsidized by shopping commissions or ad revenue the way Amazon originally planned. OpenAI’s higher price is the industry’s response to a lesson Amazon spent $25 billion to teach: there is no free smart speaker when the smart part requires continuous cloud compute.
Will Apple’s lawsuit stop the OpenAI speaker from launching?
Almost certainly not in the near term. California courts have explicitly rejected the “inevitable disclosure” doctrine, meaning Apple cannot win an injunction simply by pointing to its former engineers now working at OpenAI. To win product-level injunctive relief, Apple must trace specific stolen materials to specific design decisions in OpenAI’s hardware — a high evidentiary bar that takes years to clear in discovery. What Apple can realistically seek, and what Bloomberg Intelligence assessed as likely, is targeted preliminary relief requiring evidence preservation and compliance certification. The more immediate effect is operational: the lawsuit complicates OpenAI’s recruiting from Apple, places supply chain relationships under legal scrutiny, and adds a material litigation risk disclosure to OpenAI’s forthcoming IPO prospectus, as detailed in the trade-secret complaint coverage.
What are the privacy implications of a camera-equipped, always-listening AI speaker?
OpenAI has not disclosed the specific data retention and processing architecture for the device. What the design suggests is significant: a full-duplex system processes continuous audio rather than discrete queries, and a camera feeding environmental context into a multimodal model generates ongoing data streams that go well beyond what a wake-word speaker captures. The key questions a buyer should ask before purchasing are: what audio and visual data is sent to OpenAI servers, how long it is retained, whether it is used to train future models, and what controls the device provides for limiting collection. OpenAI will need to answer these questions in its product documentation before the device ships. Consumers who prioritize privacy would be well served to evaluate Apple’s on-device processing approach — which Apple has committed to running AI wherever possible on the device itself — as a structural comparison point.