We’ve got a new vision for the next great piece of AI hardware—and it (allegedly) is going to look a lot like a donut.
Details about OpenAI’s first piece of hardware, designed by Jony Ive’s LoveFrom studio, are emerging. According to Bloomberg, the device will be a smart speaker with no display. It will be shaped like a donut roughly the size of a hockey puck. It’s battery powered, small enough to carry around your home with one hand, and made of high-quality metal. It’s loaded with speaker grills, microphones, lights, and a camera system that watches your environment and feeds what it sees into ChatGPT.
According to the report, the speaker will work similarly to ChatGPT’s voice mode on smartphone apps, just with more advanced models. It also has—here’s the new parlor trick—parts that move on their own to signal when it’s responding.
It’s meant to give the hardware “personality,” sources say. The goal is to make it feel “more alive” than today’s stationary speakers.
According to Bloomberg, this “AI-first computer” is the first step in a device family that—someday, the company hopes—could replace the smartphone.
The smart speaker landscape
If we’re to judge by the recent success of home smart speakers, it will be an uphill battle for OpenAI’s to break through. As a market, smart speakers peaked years ago and have never developed beyond their role as a kitchen sidekick. Global shipments hover around 150 million units a year, roughly an eighth of the 1.25 billion smartphones that ship annually.
And what do people actually do with them? Play music, ask about the weather, set timers, and flip smart lights on and off. The top command categories for both Alexa and Google Home have always been music, simple search, and internet of things (IoT) control. Anything complex—shopping, payments, real tasks—remains vanishingly rare: just 5% of owners have ever ordered products by voice.
The smartphone market that the OpenAI donut wants to start invading with a moving donut is a trillion-dollar ecosystem people use for everything from banking to filmmaking. Marginal doesn’t begin to cover it.
Déjà vu
I still believe artificial intelligence will deliver the fourth and final UX revolution. Over many decades, technology has evolved to increasingly remove layers of abstraction between human intent and machine action.
The first was the command line, which let humans control computers with typed language after the punch-card era. The second was the graphical user interface (GUI), which the 1984 Macintosh used to demolish the wall between people and computing with point-and-click metaphors. The third was the smartphone, which made the GUI universal through direct finger manipulation and one app per task. In the fourth, the interface will disappear and interaction will be indistinguishable from talking to another human.
Will that fourth revolution come in the shape of a donut? It’s unlikely, but impossible to say before we see what the donut is truly capable of. Still, it’s helpful to think about the Enterprise from Star Trek for a second. People on the Enterprise carry Tricorders (our phones) and tablets (we have those, too). But nobody on the Enterprise needs to carry around the AI itself.
Nobody docks it, charges it, or moves it from the nightstand to the kitchen counter. Nobody on the Enterprise ever forgot their AI in their quarters and wondered where they’d left it. The “Computer” is simply there—ever present, totally invisible, embedded in the environment itself. And you know you are talking to it because, first, it talks back to you and, second, your brain is 100% positive that the AI works perfectly fine every time, unlike current large language models.
But Star Trek is not real life, and in real life we’ll need stepping stones to the future that Star Trek envisions. I’m not so sure a smart speaker shaped like a donut is the first step toward invisible computing. But we’ll have to wait until it launches in 2027 to find out.