OpenAI eyes phone market, Kuo says

A phone is still the most important computer most people own. That is why this rumor matters more than another weird AI gadget. Ming-Chi Kuo now says OpenAI is working on a smartphone with Qualcomm, MediaTek, and Luxshare, aiming for mass production in 2028. If that is real, OpenAI is no longer just trying to live inside other companies’ operating systems — it is trying to build one of the main doors to the internet itself. (macrumors.com) ### What actually changed? The big change is simple — the story used to be “OpenAI is making hardware, but not a phone.” Kuo’s new supply-chain note says that is no longer true. He says OpenAI has picked Qualcomm and MediaTek as chip partners, with Luxshare as the exclusive co-design and manufacturing partner, and that the project has moved far enough along to have a rough production window. (9to5mac.com) ### Why would OpenAI want a phone? Because a phone sees almost everything. It knows where you are, who you talk to, what you search, what you buy, when you move, and what app or service you are trying to use. Kuo’s argument is that this makes the smartphone the best device for real-time AI agents — better than a speaker, glasses, or a desktop browser — because it has the richest stream of context. (macrumors.com) ### Why not just stay an app? That is the catch. An app can be powerful, but it is still a guest. Apple and Google control the operating system, the defaults, the permissions, the store, the payments, and a lot of the user flow. Kuo’s read is that OpenAI would need tighter control of hardware and software to build the kind of always-on, context-aware agent (macrumors.com)g a slot in someone else’s app store may not be enough. (macrumors.com) ### What would make this phone different? Probably not the slab of glass itself. The interesting part is the interface. Kuo says the device would be shaped by AI agents, with less emphasis on opening separate apps and more emphasis on telling the system what you want done. Sam Altman added fuel to that idea on April 26, writing that it feels like a good tim(macrumors.com)e and agents. That does not confirm a phone — but it fits the direction. (9to5mac.com) ### How does Jony Ive fit in? OpenAI has already been building a hardware arm with Jony Ive. The company said last year that Ive and LoveFrom had been collaborating with Sam Altman, and OpenAI later tied that work to io, the hardware effort built around new AI devices. Earlier reporting pointed to non-phone products first — things like a speaker or other ambient devices. A phone would not replace that strategy so much as escalate it. (openai.com) ### Is this close to shipping? No. The timeline is long by consumer-tech standards. Kuo says specs and supplier choices should be finalized by late 2026 or the first quarter of 2027, with mass production in 2028. That means this is still a supply-chain thesis, not a product launch. Plenty can change — partners shift, products get canceled, and “mass production” can slide. (macrumors.com)ai-smartphone/)) ### Why did markets care right away? Because even a rumor like this points to new winners and losers. Qualcomm shares jumped after the report, which tells you investors heard “possible new premium-device demand” and reacted fast. Luxshare also matters here — it has long wanted to be more than an Apple-adjacent assembler, and a new OpenAI hardware category would give it a different growth story. (cnbc.com) ### So what is the real stakes? The real fight is not handset market share in 2028. It is control over the interface layer. If AI agents become the thing that actually uses software on your behalf, then the company that owns the agent may gain leverage over the companies that own apps. That is why this rumor lands so hard. OpenAI may be lookin(cnbc.com)(macrumors.com) ### Bottom line? Right now, this is still one analyst’s supply-chain read, not an OpenAI product announcement. But the logic is coherent. If OpenAI really believes AI agents will become the main interface, a phone is the most obvious battlefield — and maybe the hardest one to avoid. (macrumors.com)