OpenAI’s first piece of hardware will arrive before the year is out. Chief Financial Officer Sarah Friar said this week that she has already used the company’s self-developed AI device and that it will launch by the end of 2026, pulling the release forward from the February 2027 date the company had previously indicated in documents.
Friar made the comments during a public appearance, describing the device as a product she has personally experienced in testing. She gave no details on its form or features, saying only that the company is “very excited” about it. The disclosure, brief as it was, represents the first time an OpenAI executive has confirmed both that the device exists and that it is shipping soon.
The accelerated timeline signals confidence inside the company. OpenAI’s hardware program has been one of its worst-kept secrets, with reports of a dedicated team and design partnerships circulating for over a year. A launch this year, ahead of the February 2027 date in earlier filings, suggests the project reached a state the company considers shippable faster than planned, a person familiar with the program said.
The strategic direction is clear even if the product is not. OpenAI is moving from a software company that rents access to its models to a company that sells the machines those models live in. The device, whatever it looks like, is designed to put ChatGPT and its successors directly in people’s hands, without a browser, without an app store and without another company’s platform in between.
The comparison points are obvious. Google pairs its Pixel phones with its own Tensor chips to control both the hardware and the software experience. Apple has spent decades building the A-series silicon that powers its products. OpenAI’s device is an attempt to claim the same vertical integration for AI, controlling the model, the device and the relationship with the customer, rather than depending on Apple, Google or a phone maker to deliver its technology.
The timing puts pressure on rivals. A launch this year would land OpenAI’s hardware before Apple’s glasses strategy produces a consumer product and while Google is still folding its AI into phones designed for a different era. The people who have seen the device describe it as a companion product rather than a phone replacement, but they caution that the final form could change before the launch.
The economics are the hard part. Hardware is a low-margin business compared with software subscriptions, and the device market is unforgiving to newcomers without distribution. OpenAI has no retail stores, no supply chain relationships and no track record in consumer electronics. The company will need partners for manufacturing, and analysts expect the device to be priced to sell, possibly at a loss, to seed a market for its services.
There is also the question of focus. OpenAI is simultaneously training frontier models, running a massive cloud business, fighting to keep customers and preparing an IPO-driven financial structure. Adding a hardware launch to that list is a bet that the company can execute across fronts that have historically tripped up software companies. Apple’s success with silicon took decades; OpenAI is trying to compress that into months.
The device has been rumored for months, and the pieces are coming together. Reports have described a hardware team staffed with veterans of consumer electronics companies, and OpenAI has filed patents covering a device that could handle voice, vision and touch, according to people who have reviewed the filings. The company has also been hiring supply-chain and manufacturing executives, the kind of talent a software company does not need unless it plans to ship physical products at volume.
The graveyard of AI hardware is instructive. Devices like the Rabbit R1 and the Humane Ai Pin, launched to hype and reviews, failed within months because their software was not good enough to justify a new form factor. OpenAI’s advantage is that its software is the reason anyone would buy the device: the model is the product, and the hardware exists to put it in people’s hands. But the same logic cuts both ways, and a device that merely repeats what a phone can do will face the same fate as its predecessors.
The competitive response is already forming. Google, which pairs its models with its own phones and silicon, has the infrastructure to counter, and Apple has signaled it will bring its own AI to its devices. OpenAI’s move compresses the timeline: if the company ships this year, it establishes the category’s expectations before the incumbents fully mobilize. The people who have used the device say the company knows the bar it must clear, and that the internal discussions have been blunt about the risks of rushing.
The industry has noticed a pattern: every major AI model company now wants its own hardware. Anthropic has explored devices, Google builds its own silicon, and Meta designs for glasses. The model, the thinking goes, is the operating system of the next decade, and the company that owns the device owns the customer. OpenAI’s confirmation that its device ships this year makes the race explicit. The product, when it appears, will be judged against a standard the company itself set: not just a gadget, but the first real test of whether an AI company can become a hardware company.
