OpenAI paid $6.5 billion for a Jony Ive smart speaker. The real product is you.

THE MOST expensive design brief in history began with a vision of serenity. When Sam Altman and Jony Ive appeared together at an Emerson Collective event in November 2025, they described their forthcoming AI device as something that would bring "peace and calm," the antithesis of the smartphone's attention-shredding dopamine loop. Altman compared the experience to sitting in a cabin by a lake. The device, he declared, would be "the coolest piece of technology that the world will have ever seen." Then details leaked, and the cabin turned out to have a camera, a microphone, and, if you squint at OpenAI's recent moves, an advertising business model waiting to be assembled around it.

According to a February 2026 report from The Information, OpenAI's first hardware product will be a smart speaker with a built-in camera, priced between $200 and $300, with more than 200 employees dedicated to the project and a launch targeted for early 2027. The company is also exploring a smart lamp and AI glasses, though those won't be ready until 2028 or later. For a venture born from a $6.5 billion all-stock acquisition of Ive's startup io (internally codenamed "Gumdrop"), a smart speaker might seem anticlimactic. But fixating on the form factor misses what may be the more consequential play hiding inside the hardware: an identity and commerce layer that connects who you are, what you want, and what you buy, and that can be monetized in ways a chatbot window never could.

OpenAI's speaker will include a facial recognition system similar to Apple's Face ID, enabling users to authenticate purchases and verify their identity. In internal presentations, employees were told the device would observe users and proactively suggest actions, recommending an early bedtime ahead of a morning meeting, for instance, by using its camera to interpret its surroundings. This is not a speaker you talk to; it is a speaker that watches, listens, identifies, and acts. The privacy implications are obvious. Less obvious, and arguably more significant, is how neatly this dovetails with another Altman venture.

World, formerly Worldcoin, is the biometric identity network Altman co-founded through Tools for Humanity. Its core proposition is straightforward: users scan their irises with a proprietary device called the Orb, generating a cryptographic proof-of-personhood that can be used to verify they are a unique human online. As of 2025, Altman served as chairman of Tools for Humanity, and the network has grown aggressively, claiming more than 17 million verified users worldwide. The two companies are nominally independent. But in January 2026, Forbes reported that OpenAI was exploring a "biometric social network" to combat AI-generated bots, and had specifically considered using either Apple's Face ID or World's Orb technology to verify human users. No formal partnership has been confirmed, but the overlap in leadership, technology, and stated mission is difficult to ignore.

The timing of World's own product evolution is equally suggestive. In March 2026, World launched AgentKit, a toolkit that lets AI agents carry cryptographic proof they are backed by a verified human, integrating with Coinbase's x402 protocol for stablecoin micropayments. The pitch is that as AI agents increasingly transact autonomously (booking flights, purchasing goods, managing subscriptions) platforms will need a way to verify that a real person authorized the activity. AgentKit allows developers to link multiple agents to a single verified human using zero-knowledge proofs, enabling services to enforce per-person usage caps regardless of how many bots someone deploys. Now imagine that verification layer running natively on an OpenAI device with a camera already designed for facial recognition and purchase authentication. The architecture, at least in theory, connects.

It also connects to money. Specifically, advertising money. In January 2026, OpenAI began testing ads within ChatGPT for free and Go-tier users in the United States, a move Altman himself once called "uniquely unsettling" but which the company's cost structure has made unavoidable. The rollout has been deliberately conservative (ads have reached only about 5% of ChatGPT's mobile users as of March) but the strategic ambitions are not. Truist analysts estimate OpenAI will generate under $1 billion in ad revenue this year, growing to over $30 billion by 2030, a trajectory that would make it one of the largest advertising platforms in the world.

Currently, ChatGPT's ads are contextual, matched to the topic of a conversation. OpenAI has pledged that ads will never influence responses and that user conversations remain private from advertisers. But the company also promised users can opt into personalization based on chat history and past ad interactions, and ads are already being matched to users based on conversation topics, previous chats, and prior ad engagement. This is, at present, a relatively blunt instrument compared to the hyper-targeted machinery of Google or Meta. A device that sits in your home, watches your surroundings, recognizes your face, and listens to ambient conversation would sharpen that instrument considerably. A device that also verifies your unique identity through biometric authentication would let OpenAI do something neither Google nor Meta can do well today: tie an ad impression to a verified, unique human being, not just a browser cookie or a device ID, and follow that signal all the way through to a completed, identity-authenticated purchase.

OpenAI has already launched an "Instant Checkout" feature that lets users buy items from retailers such as Walmart and Etsy directly through ChatGPT. Pair that commerce layer with a device that knows who is in the room, what they have been discussing, and can authenticate a transaction via facial recognition, and you have something that looks less like an Echo competitor and more like the most vertically integrated advertising platform ever conceived. The model generates the recommendation. The device identifies the buyer. The identity layer authenticates the purchase. The commerce system closes the sale. All without leaving OpenAI's ecosystem.

This is where the competitive calculus gets interesting. OpenAI's application chief Fidji Simo recently described Anthropic's rapid enterprise growth as a "wake-up call," telling employees that the rival gained ground precisely because it placed fewer, more disciplined product bets. The strategic pivot that followed (cutting side quests, refocusing on coding and enterprise) reads as an admission that OpenAI overextended in 2025.

But there is a version of the OpenAI story in which hardware is not a side quest at all, but the foundation of a structural advantage Anthropic cannot replicate. Anthropic is, by design, a model company. It sells intelligence as a service: API calls, enterprise seats, developer tools. It does this extremely well; by early 2026, Anthropic had reached $19 billion in annualized revenue, up from roughly $4 billion in mid-2025, largely through business contracts with cloud providers. But it has no hardware, no identity layer, no consumer distribution channel beyond the chatbot itself, and, pointedly, no advertising revenue. Anthropic ran Super Bowl ads mocking OpenAI's decision to introduce advertising, with the tagline: "Ads are coming to AI. But not to Claude." The moral high ground is real, but it is also expensive. Without an ad-supported tier, Anthropic's path to profitability runs almost exclusively through enterprise contracts and subscriptions, a narrower base on which to build a public-market story.

If the future of AI involves agents acting autonomously in the physical world (making purchases, accessing services, authenticating identity) then the company that controls the device, the model, the identity verification, and the advertising infrastructure occupies a fundamentally different position from one that controls only the model. Apple understood this decades ago: owning the hardware, the software, and the services creates compounding lock-in that no individual layer can replicate. OpenAI, with 910 million weekly ChatGPT users, a Foxconn-manufactured device pipeline, a nascent advertising platform, and a co-founder who runs the world's largest proof-of-personhood network, is assembling the raw ingredients for something similar.

Those ingredients will soon face their most consequential audience. OpenAI could debut on the public markets as soon as the fourth quarter of 2026, and executives have expressed concern about Anthropic listing first and dampening retail demand for OpenAI shares. The company's financial position is imposing but precarious: its record funding round has swelled beyond $120 billion at a valuation approaching $850 billion, yet projected losses for 2026 alone could reach $14 billion.

In that context, hardware, identity, and advertising become narrative weapons as much as product categories. A model company going public is selling investors a margin story, one that must answer uncomfortable questions about commoditization, open-source competition, and inference costs that refuse to fall fast enough. A platform company going public is selling an ecosystem story: hardware driving distribution, distribution generating data, data training better models, identity enabling commerce, advertising monetizing the whole stack, and commerce justifying the hardware. It is the difference between pitching Wall Street a sophisticated API business and pitching it the next Apple-meets-Google. The design sensibility and device lock-in of one; the advertising engine of the other.

The ghosts of AI hardware past suggest the execution risk is severe. Humane's AI Pin was bricked in February 2025 after HP acquired its remains for $116 million. Rabbit's R1 retained just 5% of its initial users. But OpenAI is not Humane. It has the distribution, the capital, and the models. What it also has is internal tension: Ive's LoveFrom design studio operates with Apple-like secrecy, while OpenAI engineers have complained about slow revisions and restricted access. Reconciling the methodical perfectionism of an Ive-led hardware program with a company that just told its workforce to stop indulging side quests, and that simultaneously needs to demonstrate advertising scale to IPO investors, will test even Altman's formidable narrative abilities.

Altman once told employees the device would "know everything you've ever thought about, read, said." At the time it sounded like poetry. Paired with a biometric identity system that verifies you are who you say you are, and an advertising platform hungry for the richest user signal in the industry, it starts to sound more like a business plan. One that, if it works, would make the speaker's $200 price tag look less like a product and more like an invitation to the most valuable data relationship in tech. Whether that is a utopian cabin by the lake or something rather less serene depends, as it so often does with Altman's ventures, on which version of the future you choose to believe. ■

For more, join 75,000 subscribers getting tech's favorite brief here