Out of cloud: Can neural networks take physical form

Traders Union
All News
Editors' Picks
Can neural networks take physical form

OpenAI and the bet on its first physical AI device

In 2026, OpenAI plans to present its first physical device powered by artificial intelligence. This was stated in Davos by OpenAI’s Director of Global Affairs, Chris Lehane. According to him, hardware solutions will become one of the company’s key areas of development as early as next year.

OpenAI’s interest in physical AI-powered devices stopped being mere speculation after the company acquired a design studio founded by legendary former Apple designer Jony Ive in 2025. According to media reports, the deal amounted to around $6.5 billion, becoming one of the strongest signals of OpenAI’s serious intentions in the hardware space. OpenAI CEO Sam Altman has called Ive “the greatest designer in the world,” directly hinting at the ambition to create a new class of product rather than just another gadget.

Although OpenAI has not officially disclosed project details, several sources report that the company is working on small, screenless devices, likely wearables, built around voice-based interaction. Among the assumptions is a pocket-sized device resembling an iPod Shuffle or even a pen, equipped with cameras and microphones. Such a device could handle tasks like note-taking, environmental analysis, or contextual prompts without the need to constantly look at a screen.

Loading...

The tweet was deleted by the author.

But we saved everything 🙂.

Another, more futuristic version circulating in the industry is an egg-shaped device tentatively called Sweetpea. According to rumors, it could feature always-on ChatGPT for voice interaction and run on a 2 nm chip. Altman has previously described the future product as “more peaceful than a smartphone,” emphasizing simplicity and a non-intrusive user experience.

The key idea promoted by OpenAI’s leadership is a shift from the smartphone era to so-called ambient computing. This concept refers to lightweight, almost invisible devices that are constantly close to the user, analyzing the world in real time and processing images, sounds, and queries without keyboards, screens, or traditional interfaces.

But this is not the first attempt to give AI a body. Despite the hype around OpenAI, the industry already has considerable experience, not all of it successful.

The loudest attempts

One of the most high-profile and instructive attempts was the Humane AI Pin, released in 2024. The startup, founded by former Apple employees, proposed a radical new format: a wearable, screenless device clipped to clothing, controlled by voice, and projecting information onto the user’s palm using a laser. AI Pin was positioned as the first real step into a “post-smartphone era,” where interaction with technology happens naturally and without constantly looking at a screen.

However, real-world use quickly cooled enthusiasm. The device was slow, overheated, had limited functionality, and depended on cloud services far more than users expected. A high price and the lack of a clear advantage over smartphones further weakened its appeal. As a result, AI Pin became more of an example of how an ambitious idea can outpace both technological readiness and user expectations.

Loading...

The tweet was deleted by the author.

But we saved everything 🙂.

A similar, though less radical, story unfolded with Rabbit R1, a compact, bright-orange device with a small screen that launched in 2024. Its main promise was that AI would be able to “act” on behalf of the user by executing voice commands, managing services, ordering tasks, and taking over routine actions. In presentations, it looked like a personal agent in your pocket.

In practice, however, Rabbit R1 turned out to be more of an interface to existing platforms than an independent intelligent device. It lacked depth, contextual understanding, and autonomy, while the AI itself failed to deliver a qualitatively new experience. Interest in the gadget quickly faded, reinforcing a simple truth: physical form has little value without a truly powerful neural network behind it.

The most widespread examples of a “body” for AI remain smart speakers and smart glasses. Amazon Alexa, Google Assistant, and Meta Ray-Ban glasses have already taught AI to speak and “see” the world through cameras. Yet these devices have not become the center of users’ digital lives, remaining auxiliary tools with limited context and a relatively narrow range of use cases.

Why a “body” for AI is a complex challenge

These cases clearly illustrate why creating a physical body for a neural network is far more difficult than it may seem at first glance. A gadget can be stylish and technologically advanced, but without the ability to deeply understand context, operate autonomously, respect privacy, and offer an experience that is genuinely better than a smartphone, it is destined to remain a niche experiment.People still think visually, are not ready to fully abandon screens, and remain cautious about devices that constantly “listen” or “observe.”

A physical body for a neural network is not only an engineering solution but, above all, an act of trust. Perhaps that is why OpenAI is not rushing. The real breakthrough here depends not on chips or form factors but on whether people are willing to let AI leave the cloud and enter their personal space.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.