The second AI glasses panel of Display Week 2026’s Wednesday morning program pulled in the broader ecosystem: microdisplay display and light engine developers, a continuous authentication company, an independent R&D institute, and a metrology specialist. The result was a richer, occasionally contrarian conversation about what it will actually take to turn AI glasses from a niche product into an everyday object. The panel consisted of:

• Moderator – Gary Jones, General Chair of the Display Week 2026 program
• Yi Liu, VP of Business Development, Hongshi Intelligent Technology (QD microLED and optical engine)
• Eddie Chong, CEO, Raysolve Optoelectronics (QD microLED)
• Robert Cloudt, Chief Business Officer, Touch Biometrix (TFT-based fingerprint sensing and continuous authentication)
• Olivier Prache, SVP Product Development, eMagin Corporation / Samsung Display (OLED-on-silicon microdisplay)
• Philip Whartenberg, Head of Dept. IC & System Design, Fraunhofer IPMS (microdisplay R&D; OLED, microLED, LCoS backplane development)
• Sonika Obheroi, Director, Gamma Scientific (XR display metrology)

The Recurring Theme: Be Great Glasses First

Before any discussion of AI capability, waveguide architecture, or display technology, Yi Liu of Hongshi framed the primary constraint in the most direct terms of the morning: his regular prescription glasses weigh approximately 15 grams, and his personal threshold for daily wearability in an AI device is around 20 grams. That is the bar. Everything else — AI capability, display quality, battery life — is secondary to whether the device passes the initial comfort filter.

This “great glasses first” framing recurred throughout the session. Sonika Obheroi of Gamma Scientific extended it beyond weight: the glasses also must be transparent, natural in appearance, and socially acceptable — perhaps even fashionable. She pointed to the Meta Ray-Ban partnership and Google’s collaboration with Gentle Monster as examples of the industry beginning to take the fashion dimension seriously. The challenge with waveguide combiners, she noted, is that depending on the architecture they can attenuate real-world light, introduce color tinting, or create visible distortions across the waveguide surface. Combined with prescription stacks, these effects can undermine the naturalness of the viewing experience in ways that break adoption.

Defining objective metrics for see-through transmittance, color shift, and distortion — and building metrology tools that can evaluate them at manufacturing scale — is therefore not a secondary concern but a prerequisite for getting to market with a product users will trust.

Olivier Prache of eMagin (now a Samsung Display company) added a practical corollary: the field-of-view obstruction problem. Looking at current AR glasses, the optical system demands significant effort — users must redirect their gaze rather than simply glancing, which is incompatible with all-day wear. For glasses intended as a persistent ambient computing platform, the optical system must impose essentially zero behavioral burden.

The Three Enablers: Weight, Power, and Useful AI

Liu offered the clearest distillation of the panel’s first-order requirements:

Weight — Sub-20 gram total system weight for daily wearability. The display module currently accounts for a disproportionate share of system weight, making microdisplay and optical engine miniaturization the most direct lever.

Low power consumption — Not merely a battery life concern, but an all-day wearability concern. A device that needs frequent recharging is functionally not an always-on assistant. Every milliwatt matters in a system with this form factor constraint.

Truly useful AI — Liu’s example was telling: if, while attending a panel session, a user could issue a simple command to the glasses to record a speaker, summarize the remarks, and forward them to email, that would constitute genuinely useful AI. The bar is not impressive demos; it is persistent, contextually aware, practically useful capability that changes behavior. The current translation and research capabilities already available in shipping products represent an early proof point, but not yet the threshold for mass adoption.

Eddie Chong of Raysolve echoed these priorities and added the platform dimension: every major consumer electronics ecosystem — iOS, Android, Meta’s own platform — depends on the display as the primary interface between user and content. AI glasses that lack a capable display are cut off from the ecosystem entirely, which limits both the application space and the commercial model. Display quality and price are therefore not just hardware metrics; they determine whether the glasses can function as a genuine platform device.

Security and Continuous Authentication: An Underappreciated Requirement

Robert Cloudt of Touch Biometrix raised a dimension that has received little attention: trust and continuous authentication.

As AI glasses become more capable, they necessarily become more intimate — recording what users see, hear, purchase, and attend to. Today’s authentication paradigms (face recognition, PIN entry, multi-factor authentication) are one-time gate checks. Once a user passes the authentication event, the device is unlocked indefinitely, meaning a lost or borrowed device provides full access to an accumulating record of highly personal behavioral data.

Touch Biometrics’ approach is continuous fingerprint authentication using a thin layer of sensing material, manufacturable on standard IGZO active matrix TFT technology production lines using only the backplane portion of the process — no additional capital investment or process steps required. Applied to the temple bars of an AR frame, the system authenticates continuously as long as the glasses are being worn and immediately triggers re-authentication if someone else picks them up.

Cloudt’s broader design principle was important: the authentication component must not constrain the product design. The component should conform to the product, not the other way around. Thin, flexible, configurable fingerprint sensing that can cover arbitrary surface geometries — including curved temple arms — allows security to be integrated invisibly rather than as an afterthought.

Display Technology: No Clear Winner, and an Honest Technology Assessment

The most technically nuanced exchange of the session concerned display technology choice — and it surfaced a more balanced, less consensus-driven view than the microLED-optimistic narrative that dominates much of the industry conversation.

Prache made a pointed argument: microLED, while widely accepted as the long-term technology of choice, is not yet ready. More tellingly, he noted that Meta is currently using LCoS, which Prache characterized as “rather ancient in the display technology world.” His inference: the power and optical efficiency requirements of AR are severe enough that even OLED-on-silicon is not yet clearly superior to LCoS in shipping products.

While a microLED engine capable of producing 1 million nits at the panel level will lose 98-99% of that light through the waveguide. That energy must be dissipated — and it dissipates near the user’s head. In a system constrained to minimal form factor and battery life, the system-level power budget needs to account for the full optical chain, not just the emitter efficiency.

Prache’s suggestion: for a device that users will consult briefly and periodically rather than staring at continuously, OLED-on-silicon may be the pragmatically correct near-term choice. OLED microdisplays are available today, at scale, in monolithic single-substrate form (no wafer bonding), and at lower cost than current microLED implementations. The 20-year maturation horizon for both LCD and LED technology at large should give the industry pause about betting exclusively on microLED for near-term product roadmaps.

Philip Whartenberg took a technology-agnostic position rooted in R&D pragmatism: there is no single winning technology, and the right choice depends on the specific application. A projection application has different requirements than a near-to-eye display. Fraunhofer’s work spans OLED, microLED, and LCoS backplane development, and the institute actively helps customers navigate the choice rather than advocating for a particular approach.

Whartenberg also introduced one of the session’s more forward-looking concepts: a semi-transparent microdisplay that functions as its own optical combiner, rather than requiring a separate waveguide element. Still in R&D, the concept addresses the waveguide power loss problem directly by eliminating the waveguide from the optical path entirely and integrating the combining function into the display itself. How the display can create a far field image, was not addressed, however.

Fraunhofer’s work includes microlens arrays as part of the optical module to coordinate light coupling rather than relying on microcavity designs, which Prache noted remain difficult to execute at scale and with sufficient color fidelity.

On the microcavity question specifically — Jones asked both Prache and Whartenberg whether microdisplays with integrated microcavities need to be developed for optimized waveguide coupling. Prash’s view: the concept is elegant, but manufacturing tolerances are not yet there at production scale, and microlens arrays may be a more tractable near-term approach to improving waveguide coupling efficiency. Whartenberg agreed the challenges were familiar, which was part of what motivated Fraunhofer’s semi-transparent combiner concept as an alternative architecture.

Chong presented the case for quantum dot color conversion as the most practical path to high-quality, low-cost full-color microLED microdisplay. The company’s approach uses standard lithography rather than inkjet printing to deposit red and green quantum dot layers on top of a blue microLED array — addressing the yield and uniformity limitations of inkjet approaches. Current full-color brightness is 850,000 nits at the panel level, with a target of breaking 1 million nits this year. The company’s thesis is that display quality and price together will determine whether AI glasses can establish a sustainable platform ecosystem.

Liupositioned Hongshi as the only company to have achieved mass production and volume shipment from an 8-inch silicon-based microLED wafer architecture, with both monochrome and RGB color products. Ten AR glasses brands currently use Hongshi light engines, and at least ten more are expected to launch this year using Hongshi optical engines — a claim that, if accurate, represents a meaningful production lead over competitors in the silicon-based microLED segment.

Metrology: You Can’t Improve What You Don’t Measure

Obheroi made the metrology case with a line that Jones noted he had used himself at the previous year’s Display Week: “You don’t improve what you don’t measure.”

Her argument was structural: regardless of which display technology, waveguide architecture, or optical design approach wins, the path from R&D to manufacturing to consumer product requires consistent, standardized measurement at every stage of the pipeline. This is not just a production quality control problem — it is a design enablement problem. Without objective metrics for see-through transmittance, color shift, distortion, eye-box performance across different user geometries, and image quality under varying real-world lighting conditions, device makers cannot reliably predict what range of user experiences their products will deliver.

Gamma Scientific’s “Reality Verified” ecosystem is designed to provide this pipeline, from lab-stage characterization through production qualification, with measurement approaches tied to both established IEC and ICDM standards and to the perceptual reality of diverse user populations — different ages, different face and head geometries, different prescription requirements, different eye dominance patterns.

Obheroi’s broader point on retention was worth highlighting: metrology that informs design during development is what enables the “super vision” use case — the moment when AI glasses go from correcting vision to augmenting it, making users more capable than they are with unassisted vision. Getting to that threshold requires knowing, in advance and with precision, what the device is actually delivering to the full range of people who will wear it.

The Smartwatch Analogy — and Its Limits

Jones closed the session with a question to the audience: how many were currently wearing AI glasses? One hand went up. How many wear a smartwatch or wearable that connects to a phone? Most hands in the room.

His implicit challenge to the industry: the barrier to adopting a wrist-worn connected device proved surmountable. What is different about eyewear, and what would have to change to close that gap?

The panel’s collective answer, assembled across the session: eyewear is more demanding than a watch in every dimension that matters for consumer adoption — weight, aesthetics, social acceptability, field-of-view intrusion, always-on security — and it has to compete with a $15 pair of glasses that most people already find adequate. The smartwatch succeeded when it found use cases — notifications, fitness tracking, payments — that were genuinely better on the wrist than on a phone screen. AI glasses will succeed when they find the analogous use cases that are genuinely better experienced as a persistent, hands-free, contextually aware overlay on the world.

That use case may already be closer than the adoption numbers suggest. As Cloudt observed, most XR glasses are still effectively in a proof-of-concept phase. The components, the software, and the form factor are converging — but they have not yet converged in a single product at a price and quality level that makes the category self-evidently worth it. That is the work that remains.

Bottom Line

The XR-2 panel added important texture to the morning’s earlier market and manufacturing discussion. The technology choice is genuinely open — OLED-on-silicon remains a pragmatically competitive near-term option, microLED is not yet delivering on its potential, and new architectures like semi-transparent combiners are still in R&D. The security layer has been underinvested relative to the capability of the devices being built. Metrology infrastructure needs to be built in parallel with, not after, the display technology decisions. And the killer application — the AI glasses equivalent of fitness tracking for the smartwatch — is not yet universally agreed upon.

What the panel made clear is that the ecosystem is present and engaged. Companies spanning light engines, waveguides, security, R&D, and metrology are all actively working the problem, and the collaborative tone of both morning sessions reflects a shared recognition that no single company or technology will get there alone.