A Phone Built Around AI Agents, Not Apps
New leaks suggest OpenAI is planning a radical smartphone that pivots away from the traditional app grid toward an “AI agents” model. Instead of opening separate apps for tasks, users could rely on a persistent assistant that understands context, coordinates services and continuously senses the environment. Analyst reports indicate OpenAI is accelerating development, aiming for a debut in the first half of 2027, roughly aligned with other flagship launches from established players. The company is reportedly positioning this device as a direct rival to future top-tier smartphones, using ChatGPT’s massive existing user base as a springboard. Crucially, the hardware is being designed from the ground up for AI-driven interaction, rather than retrofitting AI features onto a conventional phone. That strategic choice explains why OpenAI appears to be investing in custom silicon, advanced memory and secure virtualization, all tuned for always-on, agent-like behaviour.
Custom 2nm MediaTek Dimensity 9600: What It Means
At the heart of the rumoured OpenAI phone is a custom-tweaked MediaTek Dimensity 9600 built on TSMC’s N2P 2nm process, making it one of the most advanced 2nm smartphone chips proposed so far. The standout feature is a dual-NPU architecture, with two dedicated neural processing units handling different AI workloads in parallel. One NPU could specialise in vision-based sensing while the other focuses on language reasoning, enabling simultaneous tasks like real-time scene analysis and conversational interaction without overwhelming the system. The 2nm node should significantly improve power efficiency, allowing more on-device AI inference without draining the battery as quickly as older process technologies. This design underscores that the AI phone processor is being optimised less for synthetic benchmark scores and more for sustained, mixed AI usage—continuous listening, context tracking, and multimodal understanding that run quietly in the background throughout the day.
LPDDR6 RAM and UFS 5.0: Why Memory and Storage Matter for AI
The leaked OpenAI phone specs also point to LPDDR6 RAM and UFS 5.0 storage, both next-generation standards aimed at high bandwidth and low latency. For AI-heavy phones, memory speed is critical: models, context windows, and sensor data must move rapidly between the CPU, GPU, and NPUs. LPDDR6 RAM is expected to deliver significantly higher throughput than current mobile memory, helping the device sustain continuous AI-driven tasks and persistent background processing without frequent slowdowns. UFS 5.0 storage complements this by boosting read/write speeds, which matters when loading large AI models, caching conversation histories, or streaming multimodal data. While exact capacities are not yet known, analysts anticipate substantial RAM and storage to support advanced features, especially if some form of localised ChatGPT or vision model runs on-device. Combined, these components lay the groundwork for smoother, more responsive AI experiences that feel less like “loading” and more like real-time assistance.
Security, Visual Perception and the Role of pKVM
Reportedly, the OpenAI phone won’t just run powerful AI; it will be designed to see and understand the world more reliably and securely. An upgraded HDR image signal processor (ISP) pipeline is said to improve “visual perception,” enabling the device’s AI agents to continuously analyse surroundings—lighting, objects, people—in real time. That could power features like smart reminders based on what the camera sees, context-aware answers, or live assistance for navigation and tasks. To keep this always-on intelligence from becoming a privacy liability, the phone is expected to include a protected Kernel-based Virtual Machine (pKVM). This creates isolated environments where sensitive user data and AI-related processes can run with stronger separation from the rest of the system. In practice, pKVM could help ensure that experimental AI features don’t compromise core phone functions, while also adding another layer of defence for personal information handled by the agent.
Ambitious Production Targets and Market Implications
Analyst Ming-Chi Kuo reports that OpenAI is targeting production of nearly 30 million AI smartphones across 2027 and 2028, signalling ambitions far beyond a niche developer device. The plan appears to be leveraging ChatGPT’s hundreds of millions of weekly active users to bootstrap a hardware ecosystem, potentially bundling the phone with subscription-based services. Such a strategy mirrors how traditional tech giants blend devices, cloud services and app ecosystems, but with an AI-first twist: the phone would be a dedicated vessel for OpenAI’s agents rather than a generic platform for third-party apps. The scale goal is bold, especially given the competitive flagship market and the likely premium cost of 2nm chips, LPDDR6 RAM and UFS 5.0 storage. Whether OpenAI can convert software enthusiasm into hardware sales will depend on how compelling its AI experiences are in everyday life—and whether users are willing to reshape their smartphone habits around agents.