Recently, reports by Minchi Quo about OpenAI's smartphone development plans have been making waves. It’s said that they are collaborating with MediaTek and Qualcomm, with Luxshare Precision handling manufacturing, and mass production starting in 2028. Honestly, this seems more like a way to solve management crises than a technological breakthrough.

Looking at OpenAI’s current situation, on the surface, it’s impressive. ChatGPT has 500 million weekly active users, and annual operating profit has reached $20 billion. However, here’s the key point: only about 5% of users are paying, while the remaining 95% are free users, all consuming computing power and electricity. Sam Altman himself admits that even the $200/month Pro plan is operating at a loss.

Total expenditure in 2025 is projected at $9 billion, with 70% of revenue going toward server costs. What’s even more concerning is that growth among paid users has plateaued. According to a Deutsche Bank report, the number of paid users in Europe is nearly flat in late 2025. The limits of the subscription model are becoming apparent.

OpenAI is also experimenting with advertising and enterprise services, but none of these are fundamental solutions. Advertising directly competes with Google, and the enterprise market is being encroached upon by Anthropic. In the secondary market, Anthropic’s valuation surpasses that of OpenAI.

This has led to a hardware strategy. CFO Sarah Fryer commented, “Hardware will become the next layer of value creation for ChatGPT,” but essentially, they want a platform to convert free users into paying ones. Bundling ChatGPT Pro with smartphones would enable automatic monthly billing—similar to bundling iCloud with iPhone.

However, what’s concerning is whether a company with a $100 billion funding shortfall can truly succeed in hardware. Past failures offer lessons.

Humane AI Pin raised $230 million but was sold for $699, with fewer than 10k units shipped. By February 2025, it was sold to HP for $116 million, and all user devices became non-functional. Rabbit R1 sold 100k units but faced massive returns. Users complained about 10-second delays in voice responses, and some said it was just an Android app covered by a shell. Johnny Ive publicly called both products “terrible.”

What will the market look like in 2028? Apple will have integrated Gemini and ChatGPT into the iPhone, and Siri will undergo a major AI overhaul in 2026. Samsung’s Galaxy AI already covers flagship to mid-range devices. Google Pixel runs Gemini natively. By 2028, all major smartphones will be “AI-enabled,” and AI features will be standard.

In such an environment, what will differentiate OpenAI? Minchi Quo suggests, “AI agents need to continuously understand users’ situations, and only smartphones can achieve that,” but this logic is weak. Models can be provided via APIs, and OpenAI already sells models to Apple and Samsung. It would be more profitable and less risky to sell to all manufacturers rather than manufacturing in-house.

Google invested ten years in Pixel but holds less than 2% of the global market share. Microsoft has been in the red for years with Surface until it became profitable. These companies had hundreds of billions in cash flow, but OpenAI does not.

The fundamental issue is that OpenAI is aiming for an IPO in Q4 2026. They plan to go public with a valuation of $85.2 billion, but Wall Street needs a convincing story. Model performance is converging, and subscription revenue growth is stagnating. That’s where smartphones come in. Selling 100 million units at $20/month would generate $24 billion annually—an attractive figure.

But Humane and Rabbit saw the same numbers. The results were disastrous. Consumers won’t pay for smartphones without an app ecosystem. Without WeChat, TikTok, or Google Play, no AI agent can meet daily needs.

What Minchi Quo hints at is a model of selling hardware at a loss and recouping through subscriptions. Again, it’s a “pre-loss, post-profit” story. OpenAI has been talking about this for the past three years. But by 2028, they will have spent over $100 billion. If smartphones don’t sell, this positive cycle will reverse.

Sarah Fryer is skeptical about OpenAI’s readiness for IPO and warns about the planned $600 billion expenditure over the next five years. A Bloomberg survey found that no institutional investors wanted to buy OpenAI stock on the secondary market.

Ultimately, the most likely outcome for the smartphone business is not a redefinition of the telecom industry but merely the addition of new slides to the IPO roadshow. How far those slides will go depends on factors outside OpenAI’s control. The tension between technological ideals and economic survival often forces difficult choices.