Summary: On April 27, 2026, Ming-Chi Kuo reported that OpenAI may be developing a phone designed around agents rather than a traditional app ecosystem; OpenAI declined to comment. Kuo attributes the reported supply chain to MediaTek, Qualcomm, and Luxshare and places mass production in 2028 if the unconfirmed project proceeds. The report matters because it would move OpenAI closer to owning the device entry point, but the timeline and suppliers remain analyst claims.
TechCrunch broke the news on April 27, 2026: Ming-Chi Kuo — the analyst best known for accurate calls on Apple’s hardware plans — said in a post on X that OpenAI is building a smartphone of its own, one designed from the ground up for AI agents completing tasks in the user’s place rather than for a traditional app ecosystem. When TechCrunch asked OpenAI, the answer was no comment.
If the rumor holds, it marks OpenAI’s biggest leap yet from a model company into a model-plus-device company. ChatGPT is closing in on one billion weekly users; against that backdrop, a handset rethought for agents — one that keeps reading the user’s context — drops the model straight into the pocket’s default entry point, no longer bargaining with the platform rules of Apple and Google.
Kuo’s Note: The Chip, the Assembly Partner, and the Clock
The split of work, as Kuo lays it out: MediaTek and Qualcomm would jointly develop the phone’s chip with OpenAI, and Luxshare would take care of co-design and manufacturing. As for the clock, specs and component suppliers should be locked down between the end of 2026 and Q1 2027, with mass production not starting until 2028. Even with zero stumbles, shoppers remain close to two years away from holding the device.
The names on the list carry their own weight. Kuo’s record of accurate calls on Apple hardware plans means the note arrives with credibility attached, and Luxshare is no stranger to this story — it has been tagged as assembly partner for OpenAI’s first hardware more than once, and this leak stretches that pairing from a small device up to a full phone.
The Case for Skipping the App Ecosystem
TechCrunch reads the motives on two levels. Level one is platform restriction: third-party apps get only tightly rationed system access on Apple and Google, and an LLM that genuinely operates the phone on your behalf cannot help crossing those red lines — hardware of its own is how OpenAI stops asking permission. Level two is data and context: a phone built for agents can keep reading user context, capturing habit data no third-party app could ever obtain. On architecture, the expectation is a hybrid design — small models run on the device, while complex requests go to the cloud.
That picture matches what court filings exposed in June 2025: early on, OpenAI and iOS worked together on exploring an AI device. What changed is the living arrangement — back then OpenAI was a tenant on someone else’s platform; the device now rumored would be its own house.
Earbuds Take the First Slot: Debut Set for H2 2026
The phone is not the opening act — OpenAI’s hardware clock is already ticking. Chief Global Affairs Officer Chris Lehane has said the company’s “first hardware product” will be announced in the second half of 2026, and reporting has largely pointed to a uniquely designed pair of earbuds. Lined up in order, the plan reads: earbuds out in late 2026 while phone specs are locked, the supply chain in place through 2027, and the phone entering mass production in 2028 — a step-by-step route that lets the small device season the supply chain first.
That same day brought a second headline: OpenAI secured FedRAMP Moderate authorization, opening the door to the U.S. government market (our separate story has the details). Pushing the government front and consumer hardware at the same time is a deliberate move to lean less on any single platform or market.
Do Apps Have a Future? Two Camps Answer Differently
The agents-replace-apps argument is not short of heavyweight company. At SXSW in March, Nothing CEO Carl Pei said flatly that “apps will eventually go away,” with AI agents absorbing more and more smartphone tasks. Replit CEO Amjad Masad, for his part, has sketched a post-app shape of software — software no longer a static interface but a flow generated as demand calls for it.
The skeptics bring specifics of their own. Payments, notifications, background execution, privacy sandboxes: super-apps and operating systems spent over a decade welding these deep integrations into place, and agents still show visible gaps in reliability and latency before a seamless takeover is realistic. Then comes the tougher, more practical question about the developer ecosystem: with no app developers, who supplies the server-side capabilities agents need?
Reality Check: How Far a Leak Is from the Factory Line
Rare is the supply-chain rumor that reaches mass production without at least one reset along the way. A 2028 ramp asks OpenAI to land three things at once within two years: co-design with two chip partners, on-device inference for the hybrid model split, and an answer to the existential question — how a phone with no app ecosystem wins over its first ten million users.
Product teams should read all of this as a clock. If agent-first devices arrive in 2028, the question of how your service gets triggered and paid for in a world with no app icon to tap deserves an answer now — protocols like MCP and the emerging agent standards are the API layer of that world. The rumor can be wrong; the direction is far harder to get wrong: device makers, model companies, and operating systems have all put the same future on the table.