R
Roel Huis in ‘t Veld
Director, Kriya Materials · Week of September 14, 2026
Waveguides just became a manufacturing story. That is when materials companies matter.
Two announcements in one week say the AR optics bottleneck has moved from the lab to the line.(Finally)
On 9 September, Lumus announced that Zhejiang Crystal-Optech will become the third manufacturer of its geometric waveguides, alongside Quanta Computer and SCHOTT. Eight days earlier the company had signed a licensing agreement with Quanta for its next-generation thin waveguides, the Z-30 2.0 and the A-Lens, at 30 and 50 degrees field of view, the A-Lens at roughly half the weight and thickness of the previous generation. Crystal-Optech, for context, was founded in 2002, has shipped more than 30 billion components and runs plants in China and Vietnam. That is not a research partner. That is a volume partner.
One day later, on 10 September, NTT-AT and EV Group published the results of a joint evaluation of NTT-AT's high-refractive-index UV-curable resin, the NH700 series and specifically NH762, in EVG's SmartNIL nanoimprint process. The numbers are the kind that only matter to people who make things: faithful replication of 350-nanometre line-and-space structures, 45-degree slanted gratings and meta-optical elements with features down to 65 nanometres, and a pattern-height variation of about 2.2 percent across 25 consecutive imprints, with little change after damp-heat and light-exposure testing. They took it to CIOE in Shenzhen, a trade show, not an academic conference.
Put the two together and you get a picture I find more interesting than any headset launch. When a waveguide company signs its third contract manufacturer and a resin supplier publishes imprint-to-imprint variation figures, the question has changed. It is no longer “can we make it”. It is “can we make ten million of them with the same layer on every lens”. VR.org, which reported the Lumus deals, made the sober observation that these are capacity commitments against expected demand rather than confirmed purchase orders. Fair. But capacity is what gets built first, and it gets built by materials and process people, not by product marketing.
This is the moment my part of the industry exists for, and I say that with some relief. For years AR optics has been a demo business in which the glamorous parts were the light engine and the waveguide design, and the coatings were a footnote. They are not a footnote at scale. A waveguide sits behind a cover that has to be anti-reflective, hard enough to survive a set of keys, and clean enough that fingerprints do not turn a 50-degree field of view into a smear. Between the grating and the glass sit index-matching and planarising layers whose thickness tolerance is measured in tens of nanometres. Every one of those is a wet-chemical film that has to be coated with the same uniformity on the millionth part as on the first, and every one of them is decided by a formulation, a cure and a line, not by a slide.
There is a lesson in the NTT-AT and EVG work for any specialty materials company, mine included. The resin was not qualified in isolation. It was qualified inside a named process on a named tool, and the two companies published the pairing so that a customer can pick the resin and the process together and skip much of the material selection and early process development. In nanoimprint, the process is part of the product, and a material that has not been run on the tool the customer will use is, commercially speaking, not yet a material. We have always thought of ourselves as a chemistry company. Increasingly we have to think of ourselves as a chemistry-plus-process company, because that is what the buyer is actually buying.
This is also where the convergence argument from G-Force Disruption stops being abstract. AI glasses exist because several exponential curves crossed at roughly the same time: microdisplays got bright enough, compute got small enough, and nanoimprint got precise and cheap enough to make optical structures at 65 nanometres in a step-and-repeat process rather than as a cleanroom heroic. Materials are the slowest of those curves. The coating on the lens is often the last thing standing between a prototype and a product. That is a good place to be, provided you can hold a tolerance.
The conversations worth having this year are not about field of view. They are about yield.
Sources
VR.org, 10 Sep 2026 — Two Waveguide Deals in Eight Days, and Both Partners Are Lumus Investors (Quanta licence 1 Sep; Crystal-Optech manufacturing 9 Sep) — https://vr.org/articles/lumus-crystal-optech-quanta-waveguide-deals-eight-days-2026
LEDinside, 10 Sep 2026 — NTT-AT and EVG Demonstrate High-Refractive-Index Resins for AR Optical Devices (NH762 on SmartNIL) — https://ledinside.com/news/2026/9/2026_09_10_02