OpenAI Spent $6.5 Billion to Build the Thing Nobody Wanted — Except It's Not the Same Thing

The Bloomberg article calls it a "handheld device." It isn't handheld. It's a speaker that moves on its own. That's not a small correction. It's the first clue that nobody really knows what OpenAI is building.

A mobile, screenless smart speaker with cameras, environmental sensors, and mechanical elements that shift to make the device feel "alive." It talks to you via GPT-Live, reads your email, controls your lights, and learns your habits over time. OpenAI aims to unveil the device in 2026 and ship it in 2027. Court filings suggest it won't ship before February of that year. The timeline is already slipping.

This is the first of roughly five hardware products in development. Behind it is a team of over 400 former Apple engineers and executives, recruited after OpenAI spent $6.5 billion to acquire io Products, the hardware startup co-founded by Jony Ive. The acquisition was OpenAI's largest ever. Sam Altman has said it could deliver $1 trillion in value.

The spending is extraordinary. The team is extraordinary. The question nobody can answer yet is whether there's a market for a companion that sits on a table, watches you, and talks to you.

Here's the part that makes you pause: every company that tried to build a screenless personal AI device before OpenAI failed. Humane's AI Pin was discontinued and its assets sold to HP. Rabbit's R1 couldn't justify its existence against a smartphone. The Friend pendant sparked public protests and was effectively abandoned. Meta's Ray-Ban smart glasses are the closest to anything working, with about 2 million units sold — a modest number that's nowhere near mass adoption.

The difference OpenAI points to is the team. Jony Ive didn't just design products; he shipped billions of them. Evans Hankey, former head of Apple's industrial design. Tang Tan, co-founder of io and former head of iPhone product design, now OpenAI's chief hardware officer. Scott Cannon. The team includes people who built the iPhone, the Apple Watch, the entire supply chain that delivers those products to stores. Unlike Humane or Rabbit, this isn't a team of people guessing how to build hardware. This is the team that already did it, at planetary scale.

But building something well and building something people want are different problems. Jony Ive himself dismissed the Humane AI Pin and Rabbit R1 as "very poor products" — an absence of new thinking. The implication is that OpenAI's device will have better thinking. But good thinking about what? The form factor still isn't settled. Bloomberg calls it a handheld. Other reports describe it as pocket-sized. Leaks from February 2026 pointed to a behind-the-ear wearable called Sweetpea and a pen-shaped recorder called Gumdrop. Sam Altman said the first device isn't glasses. The description has shifted from pocket companion to mobile home speaker depending on who's talking. A product this close to unveiling shouldn't sound like it's still figuring out whether it lives in your ear or your kitchen.

Then there's the lawsuit. Four days before Bloomberg described the device in detail, Apple filed suit in the Northern District of California, accusing OpenAI and two former Apple engineers — Tang Tan and Chang Liu — of stealing trade secrets related to unreleased hardware. The complaint is specific: Tan allegedly emailed himself supplier information before leaving Apple, used Apple's confidential information to recruit other Apple employees, and misled a supplier into performing a proprietary metal-finishing technique for OpenAI. Apple is seeking an injunction. OpenAI says the complaint lacks merit.

The irony isn't hard to spot. OpenAI is building a physical product with people who designed Apple's physical products, and Apple is arguing that those people brought more than general expertise with them. If Apple gets an injunction, the 2027 ship date becomes a guess. If OpenAI wins, the story still carries a reputational shadow that consumers notice even if they don't read court filings.

But the deeper question isn't about lawsuits or launch dates. It's about whether a screenless AI companion is a real product category or an early version of something else.

The naive version of this story is that OpenAI is building a better Echo. That's not what they're building. They don't view competitors' products as speakers at all; internally, this device is a computer built for AI productivity. The interface philosophy Altman described to Jony Ive is "attention liberation" rather than attention capture. The device is supposed to reduce anxiety, not create it. It should disappear as an interface, handling tasks on your behalf without requiring screens or commands.

That's a beautiful idea. It's also the idea that Humane and Rabbit had. What those companies got wrong — and what makes the OpenAI version potentially different — is the model. Humane's AI was mediocre. Rabbit's was mediocre. The hardware didn't matter because the intelligence behind it wasn't good enough to make the form factor compelling. GPT-level models are orders of magnitude better. The question is whether that difference in intelligence is enough to overcome the fundamental uncertainty: do people want a device that watches them and talks to them at home?

I suspect the answer depends on what happens to the smartphone first. Altman has said the companion won't replace the phone, the way smartphones didn't replace laptops. That's the safe position. But if the screenless companion becomes genuinely useful — if it can anticipate needs, surfacing information proactively and building a genuine rapport — the phone's dominance becomes the real question. Not today. Maybe not even in three years. But the direction of travel matters.

The way to evaluate this isn't to speculate on features or pricing. It's to watch what OpenAI does with its own resources. A company that has already spent $6.5 billion on hardware talent, is issuing RFPs to US manufacturers for silicon and motors, and is developing five products simultaneously isn't treating hardware as a side bet. This is a structural commitment. The question is whether the market has matured enough to reward it.

Humane and Rabbit tried too early. OpenAI is trying later, with better models and a team that actually knows how to build products at scale. Whether that's enough, or whether the category itself is an early version of something we haven't named yet, will become clear not when the device ships but in the six months after.

The test is simple. Watch what people do with it once they have it in the room. If they move it from room to room, the battery thing worked — mobility matters. If it collects dust in the corner, the whole category was a designer's fantasy. If they can't stop talking to it, the intelligence closed the gap that killed Humane and Rabbit. The form factor doesn't matter. The behavior does.