On April 27, 2026, TechCrunch reported that Ming-Chi Kuo, the analyst best known for accurate leaks of Apple’s hardware plans, posted on X that OpenAI is developing its own smartphone — and that the device is designed not around a traditional app ecosystem but around AI agents completing tasks on the user’s behalf. OpenAI declined to comment when asked.
If true, this would be OpenAI’s biggest step yet from model company to model-plus-device company. With ChatGPT approaching one billion weekly users, a phone rebuilt for agents — one that continuously understands the user’s context — would put the model directly into the default pocket entry point, no longer negotiating with Apple’s and Google’s platform rules.
What Kuo’s Note Says: Silicon, Assembly, and Timeline
Per Kuo’s note, the division of labor looks like this: MediaTek and Qualcomm would co-develop the phone’s chip with OpenAI, while Luxshare handles co-design and manufacturing. On timing, specs and component suppliers are expected to be finalized between the end of 2026 and Q1 2027, with mass production starting in 2028. In other words, even if everything goes to plan, consumers are nearly two years from holding this device.
The signal is in who got named. Kuo’s track record on Apple hardware gives the note built-in credibility, and Luxshare has repeatedly been linked to OpenAI’s first hardware as assembly partner — this rumor extends that relationship from a small device to a full phone.
Why Bypass the App Ecosystem
TechCrunch’s analysis points to two layers of motive. The first is platform restriction: Apple and Google tightly limit how much system access third-party apps get, and an LLM that truly operates the phone on your behalf inevitably hits those red lines. Owning the hardware means OpenAI stops asking permission. The second is data and context: an agent-first phone can continuously understand user context, capturing habit data no third-party app could ever see. The expected architecture is hybrid — small on-device models for local requests, cloud models for the heavy lifting.
This lines up with what court filings revealed in June 2025: OpenAI and iOS had collaborated early on an AI device. The difference is that the earlier version lived on someone else’s platform; the rumored one builds its own house.
Earbuds First: The First Device Lands in H2 2026
Beyond the phone, OpenAI’s hardware clock is already running. Chief Global Affairs Officer Chris Lehane has said the company’s “first hardware product” will be announced in the second half of 2026, widely reported to be a uniquely designed pair of earbuds. Read together, the sequence is: earbuds ship in late 2026 while phone specs get locked, supply chain comes online through 2027, and the phone enters mass production in 2028 — a gradual path that uses a small device to season the supply chain.
The same day, OpenAI also secured FedRAMP Moderate authorization, opening the door to the U.S. government market (see our separate coverage). Pushing the government front and consumer hardware at once reflects a deliberate effort to reduce dependence on any single platform or market.
“Will Apps Disappear?”: Two Camps
The agents-replace-apps thesis has heavyweight backers. Nothing CEO Carl Pei said at SXSW in March that “apps will eventually go away,” with AI agents taking over more and more smartphone tasks. Replit CEO Amjad Masad has sketched a post-app form of software, where software is not a static interface but a flow generated on demand.
Skeptics have concrete counterarguments. Super-apps and operating systems spent over a decade building deep integrations — payments, notifications, background execution, privacy sandboxes — and agents still have visible gaps in reliability and latency before they can seamlessly take over. The more uncomfortable question is ecosystem: without app developers, who provides the server-side capabilities agents need?
Reality Check: The Distance from Rumor to Mass Production
Between a supply-chain note and mass production usually sits more than one reset. A 2028 ramp means OpenAI must, within two years, simultaneously land co-design with two chip partners, on-device inference for the hybrid model split, and the answer to an existential question: how a phone with no app ecosystem convinces its first ten million users.
For product teams, the takeaway is timing. If agent-first devices arrive in 2028, now is the time to ask how your service gets triggered and paid for in a world without an app icon — protocols like MCP and the emerging agent standards are the API layer of that world. Rumors can be wrong; the direction is harder to get wrong: device makers, model companies, and operating systems are all pushing the same future onto the table.
Sources
- OpenAI could be making a phone with AI agents replacing apps — TechCrunch
- Ming-Chi Kuo on X
- OpenAI aims to ship its first device in 2026, and it could be earbuds — TechCrunch
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.