why would openai build its own smartphone?

ming-chi kuo (the analyst who called basically every apple product before it launched) says openai is working with qualcomm + mediatek on a custom processor, with luxshare as manufacturing partner. mass production: 2028. no official confirmation yet.

qualcomm is up +12% premarket on the news. apple is down −1.29%. the market is already pricing in what an openai phone would mean for the iphone.

here's the part people are missing – it's actually a business model story.

openai is burning $25b in 2026. its gross margin is 33%, crushed by inference costs projected to hit $14b this year. it pays microsoft 20% of total revenue through 2032. the subscription model – $20/month chatgpt plus – has a ceiling.

hardware changes that equation entirely

apple makes ~$1,000 gross margin per iphone. the services layer on top (app store, icloud, apple intelligence) is pure leverage on that installed base. openai knows this math. a phone isn't just a device – it's a recurring revenue machine with a locked-in distribution channel that no app store can tax or remove.

but there's a more specific threat openai is responding to

right now, chatgpt lives inside apple and google's ecosystems. that means:

• apple can deprioritise openai on device in favour of other models
• google can do the same on android
• both companies control what data openai can access about user behaviour

this already happened.

apple just announced that with ios 27, siri will open up to any third-party ai via a new "extensions" system. claude, gemini, perplexity, grok – all plugging directly into siri. chatgpt's exclusive partnership, launched just over a year ago, is over.

and on the backend, apple signed a $1 billion deal with google to power siri with gemini models – so google's models anchor the native experience while everyone else competes as a third-party extension, one layer removed.

google baked gemini into android as the default assistant. openai is watching its distribution get slowly walled off by the two companies whose hardware it depends on.

so the phone isn't just about ux philosophy. it's an insurance policy

the talent side tells the same story

openai's chief hardware officer tang tan spent 25 years at apple and reported directly to john ternus. he's pulled 2 dozen apple engineers in 2025 alone – camera systems, audio, wearables, manufacturing.

you don't build that team to ship a novelty device. you build it to compete.

the humane pin is the obvious cautionary tale here. great concept, terrible execution, $700 device that became a brick when the company ran out of money. openai ended 2025 with ~$40b in cash, then raised another $122b in april 2026. and it has 910m weekly active users and a consumer brand humane never had. different situation.

still – 2028 production, specs not final until late 2026/q1 2027. this is a long game. and the screenless io device ("sweetpea") launching later this year is probably the real first test of whether openai can sell hardware at all.

if that lands, the phone becomes a lot more credible. if it doesn't, the phone stays a slide in a strategy deck.