Introduction

Last month I attended and participated in MicroLED and AR/VR Connect in Eindhoven. This conference reminds me of the early days of SPIE’s AR/VR/MR conference in that it is drawing some excellent speakers and companies while still offering some excellent opportunities to meet and network with people. This conference has been growing by about 20% to 30% per year. Being in September, it times out well for me with CES in early January, AR/VR/MR in late January/early February, SID Display Week in May, and AWE in June.

This conference started as simply MicroLED Connect and later added the AR/VR component, as many MicroLED companies were targeting their products for AR. As a sign of the times, it seems many MicroLED companies are now putting an emphasis on optical interconnects and Co-Packaged Optics (CPO) for use in AI data centers. In addition to discussing my participation in the conference, I thought it important to discuss the implications of this trend with MicroLED companies and some other sobering implications for AR that were discussed.

Display Skeptics October Video Update and Upcoming Events

The October video update from Radu Reit and my Display Skeptics, covering Snap SPECS, Meta’s range of XR devices, and TDK’s acquisition of Qqmented’s laser beam scanning plus their novel reflective Meta-Optic. We also discuss MicroLEDs companies pursuing CPO.

MicroLED Connect + AR/VR Connect released a video clip from my Master Class at the conference. I’m working on getting a longer version of this video released on this blog.

I’m schedule to be on the AI/XR Podcast in November with hosts Charlie Fink (Disney legend, Forbes), Ted Schilowitz (Red Camera), Rony Abovitz (Magic Leap founder).

Looking further ahead, I’m planning on attending CES in early January and SPIE’s AR/VR/MR conference in early February.

Panel Discussion of MicroLED, LCOS, Laser, and OLED for OST Augmented Reality

The personal highlight of the conference for me was hosting a panel discussion on display devices for optical see-through (OST) AR. The panelists included Bernard Kress from Google and SPIE Fellow and past President (formerly with Microsoft’s HoloLens), Ed Tang from Avegant, which is working on LCOS light engines, Berthold Hahn from Meta working on MicroLEDs, and Barry Silverstein from the Center for Extended Reality at the University of Rochester (and until recently at Meta), who is an advocate for laser displays (both laser illuminated LCOS and laser beam scanning); and my Display Skeptic’s partner, Radu Reit, who represented Micro-OLEDs.

We didn’t settle anything, but we got to discuss the various perspectives on which technologies are likely to be used where and when in AR. It’s clear that, for now, green-only MicroLEDs are dominating the small and light ~30° FOV products. From what I am hearing, most of these green only MicroLED AR glasses are primarily being using in Asia in for aiding in English translation. MicroLEDs are still struggling to provide a size- and cost-effective full-color solution in the near term, at least. LCOS still seems to have size and cost advantages for full color. Micro-OLED is finding some applications with 1-D reflective (aka geometric) waveguides in glasses that are a bit bulkier. Laser Beam Scanning (LBS), which has been around at least as far back as the 1990s continues to have proponents, including Barry Silverstein, that say it will some day offer a smaller solution.

My Master Class on The Latest Trends in OST AR

I gave a “Master Class” on the first day of the conference. I tried to cram about 2 hours of content into my 1-hour presentation time. I presented on the trends I saw in OST AR in 2026.

I include my list of 18 key issues facing OST AR designs (right). A key point of this slide is that trying to solve any one of these issues usually has a negative impact on multiple other issues. For example if you want a larger FOV, you need more light, which takes more power, which means bigger batteries, which adds more weight. Additionally, with a larger FOV you want more pixels to fill the area which means more data, more processing, and more bandwidth with then requires more power, battery, and weight.

Below are a couple of slides from the presentation that outline the major trends and barriers to adoption I am seeing.

The more I have looked into the technology for addressing vision correction with display glasses, the more concerned I have become about the “friction” associated with adding prescriptions to AR glasses. The manufacturing “flow” for making AR glasses with a prescription is very much out of the flow of today’s ways of filling prescriptions and today’s approaches provides sub-optimal vision correction compared to normal prescription glasses.

My last point on the slide 7 above is that Meta’s Jason Hartlove’s $100/4/100 million which means $100 for the display, projector optics, and waveguide for (4) is necessary to reach 100 Million units, I feel is putting the cart before the horse. If we look at the history of the democratization of technology, first comes a product that solves a user’s problem results in early adopters to get initial volumes that will drive the cost down. While the process can be somewhat accelerated with investment, such as what Meta has done, but it also requires iterative cycles to gain from lesson’s learned to guide investment in the most productive directions (to use a another saying, “9 pregnant women won’t make a baby in one month.”).

1-2 MU/Year Market for audio to visual language translation glasses

As stated earlier, I heard at the conference and elsewhere that a large percentage of the current AR glasses are being used for audio-to-visual translation between Asian (mostly Chinese) and English, often in a school or university environment, but it also has obvious business applications. I know Radu Reit, who I partner with on Display Skeptics, uses his Even Reality glasses for spoken Chinese-to-visual English translation and finds it very helpful. It seems obvious that audio-to-visual would be better than simultaneous audio-to-audio translation. Visual translation is easier to grasp, particularly while the person is talking, and has a persistence that audio does not have, giving more time to grasp what was said.

There are reports out of China that as many as 1.6 to 2 million green-only MicroLED glasses will be shipped in 2026. If we assume they are binocular, then this means about 3 to 4 million green MicroLEDs per year. While a market of 2+ million units is a promising sign, as I’m about to discuss, it is still not large enough to have a healthy business ecosystem for component makers.

The same report that is saying 1.6 to 2 AR glasses in 2026 goes on to forecast the “usual” hockey stick growth in the subsequent years (chart on the right). Still, I’m more than a little skeptical that this is achievable given the various technical and physics problems that have yet to be solved in making AR glasses a mass-market product. They may be seeing a large niche market and then extrapolating use by a broad market. Still, it is nice to see an AR application that can be measured in terms of millions of users.

Unhealthy Business Ecosystem for AR Component Makers

First, at Texas Instruments (TI) for 20 years and later for 7 years with Syndiant, I worked on semiconductor components that went into other companies’ systems. Generally, as a component maker, to make the business numbers work, you need a lot of volume. At TI, there was a saying: “There are three variables in semiconductors- price, volume, and time (as in what year). You can pick two of the variables, and I will tell you the third.”

In Bernard Kress’s MicroLED and AR/VR keynote speech, he discussed this issue in the slide below. He made some market estimates for MicroLED glasses that may be a little low based on what some reports from China, and then used Jason’s Hartlove’s $100/4/100MU $100 display (likely full color) & optics price, which assumes 100MU volume, but it is still a useful exercise. Besides, even if you improve Kress’s numbers by more than 4x, the numbers still don’t result in a healthy business.

If Ray-Neo green-only binocular glasses cost $499 in the US, they are probably considerably less in China, which, by the time you account for other parts of the glasses, assembly, shipping, etc., means the display plus optics probably has to cost less than $100 per eye. I’ve heard (rumor) that Jade Bird Display gets about $40 for their green-only monochrome display with a projector (but not including the waveguide), with the MicroLED dominating the cost. Let’s use the 3 MU times the $40 price, or $120M total market. But then, using a 20% profit margin, we get $24M of possible gross profit split between several companies shipping green MicroLEDs. Note: gross profit only considers the per-unit cost and not the R&D, Marketing, and overhead of the business.

As Kress slide points out, eyeglass makers, thanks in part to the oligopoly of eyeglass makers and suppliers led by Essilor Luxottica, are used to 70% to 80% margins with no electronics and the other complexities introduced by AR glasses. Most think, as Kress does, that display eyewear today has gross profit margins below 10%. For a business category to be sustainable in the long term, all the companies up and down the supply hierarchy have to be making enough money to stay in business.

AI Data Centers Causing MicroLED Companies to Focus on “Co-Packaged Optics” (CPO)

If there was one message that came out loud and clear from MicroLED Connect 2026, it is that MicroLED companies are putting massive attention on optical interconnects in general and specifically Co-Packaged Optics (CPO). Every MicroLED company was talking about it, and even Bernard Kress, perhaps the leading advocate for AR glasses, spent about half his presentation on CPO.

Kress, in his presentation, discussed that the CPO has different requirements for both MicroLEDs and Waveguides. For LEDs, not only is there no need for full color, but the wavelengths are not the same. For waveguides, the goal becomes to keep a tight light pattern rather than a spread out one to create a human eye box.

The MicroLED companies are seeing CPO applications driven by AI data centers as a vastly larger potential market in the near term compared to AR. This is likely causing many MicroLED companies to shift their R&D focus to the requirements of CPO rather than giving attention to AR. This could particularly impact the development of full color.

Kress, in his presentation, pointed out that CPO is still a mostly speculative market for MicroLEDs. Currently, the CPO market is dominated by various lasers (ELS and VCSELs in the chart on the right), with only tiny inroads to date being made by MicroLEDs.

Bernard Kress didn’t present anything that was not known about CPO, but it was a little shocking hearing it come from him, one of the leading advocates for AR.

Avicena Tech is a company making OPC transceivers with MicroLEDs and Sensor Arrays on a single chip. In the chart below, they outline where they expect MicroLED transducers will fit in data centers. Basically, it is in on-board and near-board-to-board communication, where there is considerable power and delay overhead in serializing data for laser transmission. The expectation is that MicroLED will, compared to lasers, use “wide and slow” with many channels (LEDs) and many paths (via a waveguide or other optical connection) when the distances are relatively short.

The Positive Spin for AR

The “positive spin” for AR is that CPO investment will help MicroLED development in general. My concern for companies wanting MicroLED for AR is that often the means to an end becomes the end itself.

Waveguide Companies Getting into CPO

All the money being spent on AI data centers and the potential market for CPUs is not lost on Waveguide companies. Vuzix spent about half of their presentation discussing the different requirements and said they were looking to address the CPO market and were looking for customers and partners. A couple of their slides are copied below:

Conclusion

Once again, I enjoyed my time at MicroLED and AR/VR Connect. The combination of presentations along with personal discussions helps

While I was aware that MicroLED companies were starting to make serious efforts in optical communication and CPO, my experience at MicroLED Connect suggests that CPO could eclipse AR in terms of their focus for their R&D. CPO is not yet a large market for MicroLEDs and related components such as Waveguides. As Kress pointed out, there are still risks in whether it will become a large market. Still, all the money flowing into data centers is a major attraction for MicroLED companies.

It’s exciting to hear that there are finally 1 million-plus applications for AR in the form of audio-to-visual translation. I want to temper that enthusiasm with the reality that there are many technical issues in AR that are yet to be solved and that I outlined in my Master’s course at the conference.