CES Demo Failures: Ballie, Rabbit R1, and Jony Ive's $6.5B AI Device

The package sat on the kitchen counter for three days before Markus opened it. Not out of anticipation—out of dread. His partner had already asked, twice, what the small orange square was actually for. 'It's an AI assistant,' he'd said. 'Like a phone but without the screen.' She'd looked at the stack of unopened mail, the half-loaded dishwasher, and said nothing. That silence was the first honest review the Rabbit R1 ever received—and a better one than most of the tech press managed on launch day.

Six Years of Rolling Demo

Samsung's Ballie is the patron saint of the demo-to-delivery gap. The bowling-ball-sized household robot—round, yellow, with a built-in projector—first rolled across a CES stage in January 2020. It returned in 2021 (virtual), 2022, 2023, 2024, and again in January 2025, each time with slightly refined capabilities and the same implicit promise: this is coming to your home. As Jon Markman reported for Forbes in March 2026, it still has not shipped. Six years is long enough for a child to go from kindergarten to middle school. It is also long enough to run through roughly three generations of smartphone silicon, two major Wi-Fi standard upgrades, and the entire lifecycle of Amazon's Astro robot—which launched, limped along as an invite-only curiosity, and was quietly discontinued, all while Ballie remained a CES exhibit.

Ballie has become less a product than a mascot—a friendly face Samsung trots out to signal it's thinking about the home, without ever having to answer what happens when the projector hits a textured wall or the microphones try to parse a command over a running dishwasher.— Jon Markman, Forbes

The demo version of any hardware product lives in a room with calibrated lighting, a dedicated 6 GHz Wi-Fi channel, and precisely zero toddlers trying to grab the device. The shipping version lives in a one-bedroom in Södermalm with a partner who already thinks you bought too much tech. That gap cannot be closed by another year of engineering. It can only be closed by shipping something and letting it fail publicly enough to learn from.

When Shipping Makes It Worse

The Rabbit R1 and Humane AI Pin did ship, and that was somehow worse. Both arrived in spring 2024 to reviews that ranged from disappointed to eviscerating. The R1, pitched as a voice-first AI companion that would replace app grids with natural language, struggled with basic tasks like setting timers or playing music reliably—the sorts of shortcomings you only discover when someone tries to use it while cooking, one hand covered in flour. Humane's pin, a wearable projector meant to liberate users from screens, had battery life measured in single-digit hours and a laser display that was illegible in daylight. TechSpot reported in November 2025 that Rabbit employees hadn't been paid for months. Humane sold its assets to HP for $116 million in early 2025, a figure that represented less than half the venture capital it had raised.

I think they made bad products. The question isn't whether you can put an LLM in a box. The question is whether that box does something the phone in your pocket doesn't already do better.— Jony Ive, speaking to Bloomberg, January 2026

Every demo looks like magic when the Wi-Fi is perfect and the user is the CEO. The difference between a demo and a product is that a product has to work for someone who isn't being paid to use it.

The Next One's Different—Until It Isn't

Jony Ive knows something about the gap between a demo and a thing you can buy. His design firm, io, was acquired by OpenAI in a deal valued at approximately $6.5 billion in May 2025, with the explicit goal of building an AI-first consumer device—palm-sized, voice-driven, deliberately unlike a phone. A year later, the project faces a preliminary injunction from a U.S. federal court, issued in April 2026, over claims from a predecessor company called iyO Inc. that dispute the ownership of core intellectual property. Even setting the legal mess aside, engineers close to the project describe immense compute requirements that current mobile silicon cannot meet without generating enough heat to make a palm-sized device uncomfortable to hold—the sort of physics problem that no amount of minimalist industrial design can solve.

• The demo runs on a script. The product runs on a toddler's whim.
• The demo has one user. The product has to coexist with a partner, a roommate, a dog that barks at beeping things.
• The demo is filmed on a tripod. The product is used by someone with tremor, low vision, or a hand occupied by a grocery bag.
• The demo never has to justify its price on a Tuesday morning. The product arrives with a receipt and a return window.

None of this means companies should stop showing prototypes. The demo has a real function: it tests appetite, recruits early developer interest, and gives internal teams a target to aim at. But the industry's ritual of presenting stage demos as imminent products has a cost. It trains consumers to treat hardware announcements the way they treat movie trailers—entertaining, aspirational, and not especially correlated with anything that will actually show up in their living room. For every product that eventually ships, there is a Ballie: six consecutive years of applause, zero cardboard boxes on doorsteps. The difference between a demo and a thing you can buy is not a matter of technology. It's a matter of honesty. And right now, the honesty gap is wider than the engineering one.