The taxonomy — “micro-lasers” / “laser displays” means several different things
Disambiguating these is the first DD move, because they fail in different ways:
Why lasers (the bull case), strongest first
- Étendue → waveguide coupling efficiency (the strongest, most real argument). AR pupil-replicating waveguides accept only near-collimated, low-étendue light. Incoherent emitters are Lambertian/high-étendue, so most of their light never couples in and is wasted. Lasers are near-diffraction-limited, so far more of their output reaches the eye. For a given nits-at-eye, the laser engine draws much less power. This is the “wall-plug efficiency” point, and it is genuine.
- Colour gamut. Narrowband lasers reach Rec.2020+.
- Outdoor brightness. AR needs very high nits to beat sunlight; laser + waveguide can get there where microLED (especially red) struggles.
- Collimation / optical simplicity at the engine.
R&D challenges still to overcome (what to know)
- A. Red is the universal bottleneck (“the red problem”). Blue/green come from GaN/InGaN; red needs AlGaInP — a different, more thermally-sensitive material system that is inefficient at micro-scale. No efficient monolithic RGB on one substrate, so you are forced into heterogeneous integration (transfer/bond three colours): low yield, alignment-critical. Same curse that has cost microLED billions; arguably worse for lasers (cavity + facets on top of emission). #1 technical risk for any direct-emission laser display.
- B. Lasing at display pixel pitch (micro-laser arrays). Putting a real laser cavity (gain, feedback, threshold, mode confinement) into a ~2–5 µm pixel is brutal. Threshold current density rises as cavities shrink; sidewall non-radiative recombination eats efficiency (the same droop microLED fights). Aurelia had demonstrated only multi-pixel parallel lasing, not single-pixel addressable — a good marker of how early this TRL is.
- C. Thermal sensitivity. Lasers are more temperature-sensitive than LEDs: wavelength drifts, threshold climbs, efficiency falls, and (for nonlinear) phase-matching detunes. On a wearable with ~1–2 W total temple budget, a dense array or pumped converter is a thermal problem. “Lasers fix microLED’s heat” is only half true — they trade thermal-power load for thermal-stability sensitivity.
- D. Speckle. Coherence produces grainy interference. Mitigation (moving diffusers, multimode, wavelength/angle diversity, vibrating elements) adds loss, bulk, and power — fighting the very form factor. Often under-weighted.
- E. Coherence artifacts in waveguides. The same coherence that aids coupling causes interference banding/non-uniformity through pupil-replicating waveguides. Engine and combiner must be co-designed; a great engine in isolation can still image badly.
- F. Eye-safety. Coherent, collimated, high-radiance light near the eye faces stricter MPE limits (IEC 60825 / ANSI Z136); pulsed light stricter still. For wavemixing, invisible residual IR (no blink reflex) is a retinal hazard and a fault mode (conversion fails → full IR through). A single retinal-hazard fault mode is a product killer.
- G. Modulation / addressing at video rate. Individually modulating millions of laser pixels (arrays) or scanning fast enough (LBS) at full res + frame rate — driver bandwidth, power, per-pixel threshold/uniformity matching.
- H. Manufacturing, yield, cost. Epitaxy, transfer/bonding, known-good-die, mass transfer — the exact decade-long, multi-billion-dollar microLED slog. Laser approaches add cavity formation (facets, DBRs, gratings) and tighter alignment. Unless an approach avoids mass transfer (e.g. wavemixing uses a few discrete pump sources, not a transferred array), it risks repeating the trap.
- I. Green diode efficiency (direct-diode approaches). The historical “green gap” at the diode level has narrowed but green laser diodes still trail red/blue; partly why frequency-doubling (532 nm from 1064 nm) persists.
Market structure (context for any deal here)
Manufacturing: the photolithography bottleneck (why the light-path choice is also a fab choice)
A point that recurs and is easy to miss: the AR light-path architecture dictates the manufacturing base, and that may matter more than the optics.
The DD reflex: ask what fab base an approach commits to, what it costs per unit at volume, and whether it can be curved and ground to prescription. A beautiful flat engine that needs a silicon foundry may lose to a worse engine you can laminate and grind.
Market structure: an M&A market, not a scale market (the “burger curve”)
- Few buyers. Google, Meta, Apple, Amazon, plus supply-chain majors (Corning, BOE, Samsung) and manufacturing partners (Sony, Goatech). Target ~$200M; downside ~$50M if the pool is thin. The spread turns almost entirely on how many real acquirers exist.
- The held-cheap risk. A dominant OEM (Meta) can keep multiple rivals alive on NRE contracts (Alpha-Lum, others) specifically to suppress the eventual acquisition price.
- Meta’s stated ARM-like / no-exclusivity ambition (enable an ecosystem, don’t make hardware) is real at Reality Labs but, per Corning’s CTO, may not survive corporate-centre / Zuckerberg scrutiny — the simpler move is to acquire team + IP and pick a manufacturer (Google↔Samsung is the template).
- Pool expansion 2029-31. As smart glasses become the next consumer-electronics battleground, Chinese/Indian OEMs (Xiaomi), Essilor, and watch makers could multiply the acquirer count — the bull case for the $200M+ outcome.
DD checklist for any laser / micro-laser display deal (reusable)
- Which colour is the bottleneck, and what is the red story specifically?
- TRL honesty: single-pixel vs multi-pixel-parallel vs panel vs product? Lab vs fielded?
- End-to-end wall-plug efficiency (electrical → photons at the eye), and étendue/waveguide-coupling efficiency specifically — not just source efficiency.
- Thermal budget on a real wearable (W at the temple); wavelength/threshold stability over temperature.
- Speckle mitigation, and its form-factor/power cost.
- Eye-safety class + fault-mode analysis (especially invisible IR for wavemixing).
- Modulation/addressing scheme and driver power at full res + frame rate.
- Manufacturing path: does it avoid microLED’s mass-transfer trap, or repeat it? Yield/cost.
- Why does this beat both microLED and the other laser approaches (direct diodes / LBS / laser-LCOS)?
- IP / freedom-to-operate vs Meta / Snap / Apple / Microsoft patent estates.