OpenAI’s rumored AI phone and why Apple should pay attention

OpenAI can build some of the most powerful AI models in the world—but today, most people still access them through phones controlled by Apple and Google. That means even the smartest AI assistant is trapped behind someone else’s lock screen and someone else’s rules. New reports suggest OpenAI wants to change that by building its own AI-first phone and hardware ecosystem.

Why OpenAI wants its own phone

Right now, ChatGPT is just another app on your home screen. It has to live inside Apple’s and Google’s rules: app permissions, sandboxing, privacy controls, and strict walls between apps. Even a simple task like “order my usual dinner, check my calendar, pay, and let my friend know” turns into a long chain of app hops and manual confirmations.

If OpenAI wants an AI agent that truly understands your life and can actually get things done, it needs deeper access to the device. That’s the core idea behind the rumored OpenAI phone: move from an app-centric world to an agent-centric one, where you tell your phone what you want and the AI handles the rest.

The phone as the ultimate AI context hub

Smart speakers, glasses, and headphones are useful, but the phone is still the most information-dense device in your life. It knows your:

• Location and movement
• Calendar and reminders
• Messages and contacts
• Payment habits and shopping history
• Health data, photos, and daily routines

For an AI agent, that context is gold. The more the system understands about your day, the better it can anticipate needs, automate tasks, and offer meaningful help. That’s why analyst Ming-Chi Kuo argues that OpenAI needs full control over both the operating system and the hardware if it wants to deliver “complete” AI services.

From app icons to AI instructions

The current smartphone model is built around apps. You search for an icon, tap it, navigate menus, type, copy, paste, and constantly switch between apps to get anything done.

An OpenAI phone would likely flip that logic. Instead of thinking, “Which app do I open?”, you’d just say or type what you want:

• “Plan a weekend trip under $500 with my partner.”
• “Summarize today’s messages and schedule anything important.”
• “Find the best deal on this product and order it to my home.”

The AI would decide which tools, apps, and services to use in the background. Apps might still exist, but you’d stop treating them as separate destinations you have to manage yourself. The AI becomes the main interface; the OS and apps become the toolkit it quietly orchestrates.

Custom chips built for AI agents

To make this work, the phone can’t just be a normal device with a chatbot slapped on top. It needs a processor designed around constant AI reasoning and context awareness—without destroying battery life.

According to Kuo, OpenAI is working with MediaTek and Qualcomm on mobile processors, while Luxshare Precision has reportedly won the exclusive system co-design and manufacturing contract for the phone. The timeline looks like real supply chain planning: final chip specs and supplier choices are expected by late 2026 or early 2027, with mass production in 2028.

Under the hood, the device would likely blend:

• On-device AI for lightweight, fast, and private tasks
• Cloud AI for heavy reasoning and complex workflows
• Careful power and memory management so the agent can stay “always aware” without draining the battery

This is the same direction Apple is moving with its own on-device AI push. If you’re curious how Apple is thinking about this, it’s worth reading how it’s betting big on local intelligence in its next wave of on-device AI features.

A massive new hardware business

Kuo compares the economics of AI chips to AI phone processors. One high-end AI chip (like a data center GPU or TPU) can generate as much revenue as 30–40 AI agent phone processors. But phones have scale: the global high-end smartphone market is around 300–400 million units per year.

Even a small slice of that market would be a huge new business line for OpenAI. For Luxshare, the rumored main manufacturer, this is also strategic. Inside Apple’s supply chain, Luxshare has struggled to surpass Hon Hai (Foxconn) in iPhone assembly. Becoming the lead builder of an AI-native phone could move it from “one of Apple’s suppliers” to a central player in a new category.

OpenAI’s broader hardware lineup

The phone is reportedly just one piece of a larger hardware push. OpenAI has quietly built a hardware team of around 200 people, with product design led by LoveFrom—the studio founded by former Apple design chief Jony Ive. The team reportedly includes a lot of Apple veterans, such as Tang Tan (iPhone and Apple Watch product design) and Evans Hankey (former head of industrial design after Ive).

Rumored products include:

Smart speaker (first product)
• Expected price: around $200–$300
• Target shipping: February 2027
• Likely positioned as an AI-native home assistant, not just a music speaker

AI headphones
• Code names: Dime or Sweet Pea
• Design: metal “cobblestone” body with capsule-like earphones behind the ears
• Powered by a 2 nm chip for strong on-device AI

Smart glasses
• Mass production expected in 2028
• Positioned against Meta Ray-Ban and Apple’s rumored N50 glasses

Other prototypes
• A smart lamp
• An AI pen or pocket device, something Sam Altman has hinted at several times

The vision is clear: a family of devices where AI is the core experience, not an add-on.

A calmer alternative to the modern smartphone

Sam Altman has compared today’s smartphones to Times Square: noisy, crowded, and full of attention-grabbing distractions. In contrast, he describes the kind of device OpenAI wants to build as more like a lakeside cottage—quiet, focused, and under your control.

That framing matters. OpenAI doesn’t just want to sell another smartphone. It wants to sell a calmer, more intelligent way of using technology, where the AI filters the noise and handles the busywork so you don’t have to constantly manage notifications, feeds, and apps yourself.

The Apple challenge

None of this will be easy. Apple has spent more than a decade perfecting the iPhone: hardware, software, ecosystem, developer tools, chips, supply chain, and a strong privacy story. Entering the premium smartphone market is one of the hardest moves in consumer tech.

That’s likely why OpenAI is aggressively hiring Apple talent and tapping Apple’s suppliers. According to reports, OpenAI hired more than 20 hardware experts from Apple in a single year. Apple was reportedly so concerned about executive departures that it canceled a closed-door meeting in China. Meanwhile, companies like Luxshare and GoerTek—both experienced Apple suppliers—are reportedly lining up to assemble OpenAI devices and provide components like speaker modules.

In a sense, OpenAI is using Apple’s own talent and supply chain network to build something that could one day compete with Apple’s most important product.

China’s faster, riskier AI phone experiments

While OpenAI seems to be taking a slower, ground-up approach, China is already testing a more radical version of the AI phone through ByteDance and ZTE.

At the end of last year, ByteDance partnered with ZTE to launch the Dubao phone (Nubia M153). Engineering prototypes reportedly sold out immediately. The phone was priced around 3,500 yuan (about $480), but resale prices reportedly spiked as high as 36,000 yuan (around $5,000), and ZTE’s stock even hit its daily limit.

Dubao’s approach is aggressive: instead of waiting for every app to expose clean APIs for AI agents, its model reads the screen and simulates manual actions like a human user. That lets it:

• Compare prices across different apps
• Organize files and content
• Write more polished WeChat replies
• Book flights, order food, and handle everyday tasks by directly operating the UI

The upside is speed. People could actually use an AI agent phone in the real world at the end of last year. The downside is security and compatibility. Major apps like WeChat, Alipay, Taobao, and banking apps reportedly started blocking the Dubao phone, because from their perspective, an AI that can imitate user behavior and bypass normal app boundaries is a serious risk.

A wave of AI OS phones in China

Despite the pushback, Dubao phone 2.0 is already in development and expected in the second half of the second quarter this year. There are also reports that ByteDance wants to expand the concept to more phone makers.

Local media have reported that Honor was the first manufacturer ByteDance approached, though Honor has publicly denied any partnership so far. One source summarized Honor’s caution: Dubao can afford to be radical as an experimental engineering project, but a brand with hundreds of millions of users has to prioritize stability, compatibility, and security.

According to well-known tech blogger Digital Chat Station, Vivo is now in talks with Dubao, and other top-five domestic manufacturers are lining up. His take: a wave of AI OS and Dubao-style AI phones is coming.

So while OpenAI builds from scratch, China is trying to transform existing Android phones from the inside out.

Two paths to the same future

Both OpenAI and China’s Dubao ecosystem are converging on the same conclusion: AI at the app level isn’t enough. If the agent is just another feature inside someone else’s phone, it stays limited by that phone’s rules.

There are two main strategies emerging:

• Transform existing phones: Use techniques like screen reading and GUI automation to let AI operate current apps (the Dubao path). Faster to ship, but risky for security and app compatibility.
• Build new devices from scratch: Design the hardware, OS, and chips around AI agents from day one (the OpenAI path). Slower and harder, but potentially cleaner and more stable long term.

Whichever path wins—or if both coexist—the direction is clear: the AI agent is moving from “smart feature” to “soul of the device.”

OpenAI’s shifting stance on AGI

All of this is happening while OpenAI quietly reframes how it talks about AGI (artificial general intelligence). Sam Altman has publicly floated the idea that AGI could arrive by 2030 and has made provocative comments like “after AGI, no one is going to work and the economy is going to collapse,” while also saying GPT-5.5 and Codex are so useful he’s switching to polyphasic sleep to work more.

At the same time, other leaders in the field are pushing back on the AGI narrative. Peter Steinberger, creator of Multibot, has argued the industry should focus more on specialized intelligence rather than a single general system. Anthropic president Daniela Amodei has called AGI an outdated concept, and Google DeepMind’s CEO has said AGI will require robust world models.

Reflecting this shift, OpenAI updated its operating principles. In its 2018 charter, AGI was mentioned 12 times. In the 2026 version, it appears only twice. The focus has moved from one big AGI breakthrough to “iterative deployment”: society adapting step by step as each new level of AI capability arrives.

One notable change: the 2018 charter promised that if another safety-focused lab got close to AGI first, OpenAI would stop competing and help them instead. That promise is gone in the 2026 version, which now talks about trading off some “empowerment” for more resilience when needed.

In other words, OpenAI is positioning itself less as a pure research lab waiting for AGI and more as a company building powerful, practical systems and products—like an AI-native phone—that roll out gradually into everyday life.

What this means for users and builders

If OpenAI succeeds with its AI phone and hardware ecosystem, we may move from a world where we manage apps to a world where we manage goals. That would change how we design software, how we think about privacy, and even how we work.

For power users and developers, it’s another signal that the future isn’t just about smarter models, but about where those models live and how deeply they’re integrated into our devices. You can already see this shift in how people are choosing tools like OpenAI Codex over more traditional assistants, especially for work and automation-heavy use cases—something we broke down in detail in our look at why many power users are switching to OpenAI Codex.

For now, the OpenAI phone is still a rumor backed by serious supply chain signals and high-profile hires. But if it ships anything like what’s being described, Apple and Google won’t just be competing on cameras and screens—they’ll be competing on who controls the soul of the device: the AI agent that runs your digital life.