Can OpenAI Pave the Way for AI Smartphones?

OpenAI Makes Its Most 'Bold' Move Yet.

Recently, Ming-Chi Kuo, an analyst at TF International Securities, disclosed that OpenAI is in the process of developing its own smartphone. The company plans to collaborate with MediaTek and Qualcomm on processor development, with Luxshare Precision serving as the sole partner for system design and manufacturing. Mass production is anticipated by 2028.

As hardware innovations fail to bring about transformative user experiences, the smartphone industry is increasingly viewing 'AI' as the key driver for the next wave of device upgrades. However, it's not just smartphone manufacturers that are eyeing the AI smartphone market. Similar to how ByteDance partnered with smartphone makers for its Doubao phone, OpenAI is not content with merely being a model provider.

So, what will OpenAI's AI smartphone look like? By opting for self-development, can OpenAI redefine human-computer interaction in the AI era, just as Apple redefined smartphones in the past? Compared to ByteDance, which is entering the market with fewer constraints, OpenAI faces a more challenging path ahead.

01. OpenAI's Hardware Gamble

OpenAI's ambition to venture into hardware development is not a recent development. Over the past year, there have been persistent rumors about OpenAI developing its own earbuds, smart speakers, and glasses. Industry insiders suggest that OpenAI aims to establish a dominant presence across all potential next-generation computing platforms.

This time, OpenAI is setting its sights on smartphones—the most demanding, high-threshold, and unforgiving sector of the hardware market. According to Kuo, OpenAI's decision is driven by two primary factors.

Firstly, while sales of smart glasses, earbuds, and 'AI Pins' are on the rise, smartphones are expected to remain the most prevalent and suitable devices for running AI agents in the future.

Future AI smartphones will transcend the limitations of traditional screens. A former Baidu product manager noted that AI smartphones will not only execute tasks for users but also proactively do so by sensing user states without explicit instructions.

For instance, if a user needs to travel for business after a meeting, the smartphone could proactively book flights and hotels upon hearing relevant conversations, eliminating the need for users to frequently check their phones. Interaction will shift from touchscreens to primarily voice-based.

Kuo shared a conceptual diagram of AI smartphones in his vision. In this envisioned form (which could be translated as 'form factor' or retained as 'form' for conceptual emphasis), apps will become obsolete, replaced by a series of task-executing agents on the frontend.

For agents to seamlessly integrate into daily life, a native operating system and hardware-software synergy are crucial. If OpenAI remains a model provider, it will always be reliant on other ecosystems, akin to a 'plugin'.

Current AI applications are confined to chatboxes by existing smartphone systems, susceptible to restrictions similar to how smartphone makers can replace pre-installed apps. Only by developing hardware in-house—controlling chips, sensor scheduling, system permissions, and user data—can OpenAI truly lead the next generation of interaction.

Moreover, OpenAI's push into hardware is motivated by a critical issue: the internet's data is no longer sufficient.

Models rely on online text data for training, lacking real-world perception and understanding. For example, when a user asked Doubao to 'make my hand look more bony,' it returned a photo resembling a CT scan—a humorous yet telling example of AI's inability to grasp real-world physics and human semantics.

By controlling hardware terminals, model providers can lawfully collect vast amounts of multidimensional data from the physical world, enabling AI to better comprehend reality. This is the fundamental reason why software giants like OpenAI and ByteDance are heavily investing in this crossover.

However, unlike ByteDance, which partnered with ZTE to develop the second-generation Doubao phone, OpenAI has opted for a pure self-development path. While ByteDance collaborates with manufacturers, OpenAI faces a steeper challenge. In overseas markets, the smartphone industry is dominated by Apple and Samsung. Finding a manufacturer with both production capabilities and a willingness to cede system control is nearly impossible. OpenAI has no choice but to pursue pure self-development.

02. OpenAI's Technological Hurdles

For model providers like OpenAI, self-developing any hardware means starting from scratch, with numerous technical challenges.

Firstly, at the software level, how can agents become truly practical? This is a challenge that domestic smartphone makers are also grappling with.

Domestic smartphone AI emerged in 2023. Today, Huawei, OPPO, Vivo, Xiaomi, and Honor have integrated large models into their assistants. However, current smartphone AI falls short of proactive perception. Even passive task execution often yields subpar results. Users frequently complain about their smart assistants failing to set alarms or perform simple calculations.

Only the pre-ban Doubao phone achieved cross-app, fully autonomous execution, such as 'order takeout with one sentence' or 'plan my trip with one sentence.'

While the Doubao phone offered unparalleled convenience and represented the highest AI intelligence level accessible to the public, its GUI agent approach posed higher security risks. Most domestic manufacturers opted for the safer 'API call' route—deep collaboration with mainstream apps, enabling AI assistants to invoke specific functions via pre-set interfaces rather than simulating human screen taps like GUI agents.

To further ensure privacy and data security while reducing latency, manufacturers are exploring on-device AI. This involves embedding large models directly into smartphones, minimizing data uploads to the cloud. AI thinking, instruction judgment, and task processing occur locally.

Stringent data collection and privacy regulations in Western markets compel OpenAI to adopt a robust on-device approach. At a minimum, it must run small models locally to continuously understand and remember user routines, delegating complex tasks to the cloud.

This places immense demands on OpenAI's hardware-software synergy. With no prior hardware industry experience, OpenAI must develop models that run efficiently on smartphones, design chips tailored to model computation, and create a native AI OS enabling agents to freely coordinate apps. Starting from scratch across chips, systems, and AI models explains why mass production isn't expected until 2028.

To expedite R&D, OpenAI is leveraging China's supply chain strengths to compensate for its lack of hardware experience.

According to Kuo, OpenAI's supply chain is taking shape: MediaTek and Qualcomm are handling processor development; Luxshare Precision is the exclusive system design and manufacturing partner. Luxshare Precision has extensive experience assembling AirPods, Apple Watch, and even iPhones, possessing end-to-end capabilities from components to system-level packaging. MediaTek and Qualcomm, global chip leaders, excel in flagship platform NPUs and power efficiency, respectively.

However, creating a smartphone involves far more than just three core suppliers. OpenAI has also recruited executives from Apple.

'Quijie Business' learned that OpenAI's hardware team now exceeds 200 members, including key Apple figures like former Chief Design Officer Jony Ive and former Product Design Head Tang Tan (who oversaw iPhone and Apple Watch launches).

Yet, while talent brings technical expertise, it cannot replace Apple's decades of organizational experience and supplier networks. A high-end smartphone relies on hundreds of suppliers and thousands of precision-coordinated processes. While OpenAI has sketched the supply chain 'skeleton,' filling in the 'flesh' will be its true test.

Notably, domestic manufacturers are accelerating AI smartphone iterations. By 2028, could Huawei, Xiaomi, or Honor preemptively achieve full-scenario AI agent interaction? The future of AI smartphone dominance is worth closely observing.