The Ray-Ban Display is Meta’s step into Tier 2 — a monocular 600×600px Micro-OLED display projected at 42 pixels-per-degree with a 20° field of view, hitting 5,000 nits peak brightness. The optics story here is a surface-relief waveguide: a diffractive structure etched into glass that couples light from the display engine into your line of sight. At 20° FoV it’s narrow but usable for HUD-style overlays without the colour fringing artefacts that plagued earlier holographic waveguides.

The included Neural Band is the more technically interesting piece. It’s an EMG (electromyography) wristband that reads muscle activation signals before they reach your fingers, enabling gestural control with sub-10ms latency — effectively a neural interface without implants. The SoC splits inference across the frame (always-on VAD, image capture) and the Neural Band (gesture classification). The cloud handles heavy multimodal reasoning via Meta AI. Total system weight is undisclosed but the frame alone is comparable to standard Ray-Bans.

Architectural bet: Neural interface input + cloud-heavy AI inference offloads most complexity off-device. Works well today; dependent on connectivity for anything non-trivial.

XREAL’s X1 is a custom ASIC — not a general-purpose SoC, but a spatial computing chip purpose-built for 3DoF head-tracking and display stabilisation. The result is 3ms motion-to-photon (M2P) latency, which is the threshold below which most humans stop perceiving lag between head movement and display update. That’s a genuinely hard number to hit in a sub-40g form factor and it’s what separates this from competitors using off-the-shelf SoCs.

The display is a Sony 0.55-inch Micro-OLED paired with XREAL’s Optic Engine 4.0 — a geometric waveguide using a birdbath combiner geometry rather than surface-relief diffractive elements. The tradeoff: birdbath combiners are bulkier and have higher reflection loss than diffractive waveguides, but they’re cheaper to manufacture at scale and deliver more consistent colour fidelity. At 57° FoV and 1080p, you’re getting a virtual 171-inch screen at any working distance. For productivity — code editor, browser, document — this is the current best-in-class display experience under $1,000.

Architectural bet: Vertical silicon integration (custom ASIC) for display-first workloads. Trades AI inference capability for best-in-class optics performance at price point.

Brilliant Labs is making a distinctly different architectural bet to everyone else: put as much AI inference on-device as possible, avoid the cloud by default, and publish everything as open-source. The Halo uses an Alif Semiconductor MCU — a power-sipping microcontroller that can run quantised small language models locally using INT4/INT8 inference. Their Neuphonic voice AI (Noa) runs entirely on-chip for most tasks, with long-term episodic memory stored locally rather than in a data centre.

The optics are a geometric prism bonded to a MicroOLED — a simple but effective design that weighs under 40g total and delivers all-day battery life because the MCU’s idle power draw is measured in milliwatts. The FoV is approximately 20° diagonal, which is modest, but for the ambient data overlay use-case (time, notifications, Noa responses) it’s sufficient. The privacy architecture is the real differentiator: Gemini, GPT, and other cloud models are optional plugins, not requirements. Alif partnership in early 2026 upgraded the neural processing unit on-die.

Architectural bet: Edge-first inference with privacy guarantees. The right call if regulation tightens on biometric data from wearables — which it likely will.

Google’s play is an OS-level wedge, not a hardware product per se. Android XR is a unified platform across glasses, headsets, and spatial devices — and the monocular glasses tier (targeting mid-2026) is their Trojan horse into daily wear. The AI stack is Gemini-native: the glasses can see what you see via an outward-facing camera, run visual question-answering (“what restaurant is this?”), and surface contextual overlays via the monocular display. Critically, Google is building Gemini Nano inference into the on-device pipeline, meaning a subset of queries never leave the hardware.

The three-tier XR strategy signals a willingness to commoditise the hardware layer (multiple OEMs building Android XR glasses) while capturing value in the OS and AI services layer — the same playbook that made Android dominant in phones. The risk: OEM fragmentation in a form factor where millimetre-level fit and weight balance matter enormously. The reward: if it works, Google instantly owns the distribution layer for AR eyewear.

Architectural bet: Platform play — own the OS and AI services layer, commoditise the glass. Highest ceiling if it works; history suggests Google can abandon hardware at any moment.

Apple’s glasses are in late-prototype / high-volume supplier sampling as of Q1 2026 — cameras, microphones, speakers, with Apple Intelligence handling translation, calls, and navigation. No display in the first generation if current leaks are accurate, which is a deliberate Tier 1 entry matching Meta’s proven playbook. The differentiator will be the silicon: Apple’s W-series or a new custom SoC will bring the Neural Engine to the wrist-and-face form factor, enabling on-device inference for the full Apple Intelligence stack without the latency penalties of cloud round-trips.

What matters architecturally is Apple’s vertical integration lock: the glasses will interface natively with iPhone’s Secure Enclave for biometric data, share the CoreML inference graph with AirPods and Apple Watch, and leverage the existing sensor fusion stack from the AW Series 10 for motion estimation. Apple doesn’t need to win the optics war in round one — they need to establish the accessory ecosystem lock-in and then add display capability in Gen 2 or 3. Projected launch: late-2026 at earliest, more likely Q1 2027. Shipments of 3–5M units at launch.

Architectural bet: Ecosystem flywheel — glasses as the fifth Apple device. Doesn’t need to win on specs; wins by making the other four devices better.