TL;DR:

  • Meta Orion uses silicon carbide waveguides for its AR display — the highest-refractive-index optical material available, enabling a wider field of view in a glasses-shaped form factor
  • Input is handled via a neural EMG wristband that reads electrical signals from forearm muscles, allowing fine-grained hand gesture control without hand-tracking cameras on the glasses themselves
  • The developer preview units (limited distribution, not a consumer product) give a real signal about where practical AR eyewear is heading — and what problems remain genuinely unsolved

When Meta announced Orion at Connect 2024, the reaction split pretty cleanly. Hardware people were impressed: silicon carbide waveguides are an actual technical leap, not a marketing claim. Consumer tech people were sceptical: another AR device that doesn’t ship. Both reactions were basically correct.

Orion is a research-and-developer preview product. Meta has been explicit that it won’t sell these to the public — the units distributed are seeding the developer ecosystem and testing real-world use ahead of whatever comes next. But the technical choices baked into Orion tell you a lot about the hard problems that all AR glasses have to solve.

The Display Problem

Every pair of AR glasses is fundamentally a display problem in a form factor that punishes display technology. You need bright, high-contrast images, a wide enough field of view to be useful, and a lens that still looks like a normal pair of glasses. These requirements are in direct conflict with each other.

Most AR glasses to date (HoloLens, Magic Leap, early waveguide devices) have used glass-based waveguides. Glass has a refractive index around 1.9, which constrains how wide you can make the field of view before the waveguide has to get physically larger — and therefore heavier and more obviously not a normal pair of glasses.

Silicon carbide has a refractive index of around 2.65. Higher refractive index means you can bounce light through the waveguide more efficiently and achieve a wider field of view in a thinner lens. Meta claims approximately 70 degrees of field of view on Orion — substantially larger than anything comparable at a glasses weight.

The catch is that silicon carbide is extremely difficult and expensive to manufacture. This is a large part of why Orion isn’t a consumer product: the cost per unit is reportedly in the tens of thousands of dollars. But the existence of a working device at this FOV and form factor is the proof that the optical problem is solvable. Cost is an engineering problem. Fundamental physics constraints are not.

The Input Problem

AR glasses need input. Earlier devices defaulted to hand tracking using cameras mounted on the glasses — effective enough, but power-hungry, adds bulk, and requires deliberate gestures that look odd in public.

Meta acquired CTRL-labs in 2019 for a reported $500m to $1bn, and the EMG (electromyography) wristband that emerged from that acquisition is one of the more interesting pieces of the Orion system. The wristband reads electrical signals generated by forearm muscles when your fingers move — before the movement is fully visible to a camera. This means it can detect subtle finger micro-gestures: a small pinch, a scroll movement, a tap against your thumb.

The practical upshot is that AR gesture input doesn’t require your hand to be raised and visible. You can control an Orion interface with your hand at your side, making small movements that are essentially invisible to bystanders. That’s a qualitatively different interaction model from anything previous.

There are limits. The wristband has to be calibrated per user, and fine gesture recognition in varied conditions (different forearm positions, varying muscle tension) is still an active research area. But as an approach to always-available, socially acceptable AR input, it’s the most credible attempt anyone has made.

The Compute Architecture

Orion splits compute across three components: the glasses themselves, the wristband, and a separate compute puck that you carry in a pocket or bag. The puck handles the heavy lifting — AI processing, rendering, connectivity — and communicates wirelessly with the glasses.

This is a pragmatic acknowledgement that you cannot currently fit the compute needed for real-time AI-assisted AR into a glasses-sized battery. The puck adds friction (another thing to carry, charge, potentially lose) but removes the constraint that’s killed most previous AR devices: thermal throttling and battery life measured in under an hour.

Future generations will presumably absorb more compute into the glasses themselves as efficiency improves. The Apple Vision Pro chose a different tradeoff — everything in one device, at the cost of headset form factor and weight. Orion bets that distributed compute across wearables is the right architecture for glasses-shaped AR, at least for now.

Where Samsung and Others Sit

Meta isn’t alone. Samsung’s Galaxy Glasses programme — developed in partnership with Google and Qualcomm — is targeting a more near-term consumer release, with Android XR as the software platform. The early signals suggest a more modest FOV than Orion in exchange for actual manufacturability and a consumer price point.

There’s a reasonable argument that the right near-term AR glasses product is not the most technically ambitious one. A smaller FOV device that ships at £500 and integrates with Android apps could be more useful day-to-day than a developer preview with extraordinary optics that costs more than most cars to produce.

What Orion establishes is the ceiling: the technical problems have tractable solutions. Silicon carbide waveguides work. EMG wristband input works. Distributed compute works. The question for the next two to three years is how much of that ceiling you can ship at a viable price point — and which compromises users will and won’t accept.

The spatial computing transition isn’t going to happen with one device announcement. It’s going to happen when AR glasses are good enough at enough things that people actually keep wearing them. Orion is evidence that the hardware constraints shrinking fast. Whether the experience catches up is the harder question.