Prelude to XR War II: Meta, PICO, VR Glasses, and Split-Architecture Headsets
X：@xuwu
Exploring spatial computing, XR & the metaverse. Angel Investor. Founder of Maoyan (HK: 1986). Venture Partner at ZhenFund. Early employee at Meituan & Baidu.
Email: realwuxu@gmail.com
Main Conclusions：
1. While the industry is still debating whether Apple has canceled the next Vision Pro, Meta and PICO are ready preparing to ship their answers.
2. Buiding on the Vision Pro approach, the industry has created a new product category built around split architecture.
3. Split architecture is a new trend because it solves the trade-off between weight and performance. For the past ten years, higher performance almost always meant more weight, while making a headset lighter usually meant cutting features and performance. For the first time, split architecture makes it possible to have both lighter weight and high performance. The headset weight problem that has held the industry back for more than a decade is finally being solved. Meta VR Glasses take this even further. They show that VR can start to look more like glasses.
4. Split architecture also opens up a much wider price range for VR. Like PCs, future VR devices could come in many different configurations, from a few hundred dollars to several thousand. This could bring VR to a much wider group of users.
1. It Looks Like VR Is in Trouble Again
If you only followed the news over the past six months, it would be easy to think VR was in trouble again.
Apple is rumored to have canceled the next Vision Pro, which was expected to be lighter and cheaper.
Meta has also been cutting jobs and projects at Reality Labs, while putting more resources into AI and glasses. Zuckerberg is also talking more about returns when he talks about VR.
At the same time, many VR game and content studios are laying off staff, canceling projects, or moving to other platforms.
Many people who once rushed into VR are now leaving.
2. In 1983, Video Games Looked Dead
Before we say VR is over, it is worth looking back at what happened to video games.
In 1977, Atari launched the Atari 2600.
For the first time, video games truly entered mainstream American homes. By the early 1980s, Atari had become one of the most visible technology and entertainment companies in the United States, and video games had quickly grown from a new kind of toy into a multibillion-dollar industry.
Then everyone rushed in. The market filled with low-quality games, and consumers no longer knew what was worth buying. The most famous example was Atari’s E.T., based on the movie. To make the Christmas selling season, The game was rushed out in just a few weeks. Atari had huge expectations for it, but ended up with large amounts of unsold inventory. Then came one of the most symbolic scenes of that era. Truck after truck of unsold game cartridges left Atari’s warehouses and were eventually buried in a landfill in New Mexico.
Starting in 1983, the North American video game market collapsed. Many game companies shut down or left the industry. Retailers lost confidence in game consoles. Consumers lost interest in what had recently been one of the hottest new forms of entertainment.
If you had been living in the United States in 1983, watching Atari lose huge amounts of money, game companies go out of business, consoles disappear from store shelves, and boxes of cartridges get thrown away as trash, it would have been very easy to reach a conclusion that sounds ridiculous today but seemed completely reasonable at the time: Video games might have been nothing more than a short-lived bubble.
But this is where history gets interesting. In the same year that the North American video game industry was collapsing, a Japanese company on the other side of the Pacific, best known at the time for playing cards and toys, was moving in exactly the opposite direction.
In July 1983, under Nintendo’s third president, Hiroshi Yamauchi, Nintendo launched a small console called the Family Computer. The world later came to know it as the Famicom. Nintendo then released games such as Super Mario Bros. and The Legend of Zelda, franchises that are still known around the world today.
Two years later, Nintendo brought a redesigned Famicom to the U.S., where people had nearly given up on video games. There it received another famous name: The Nintendo Entertainment System, or NES.
A few years later, video games were not dead at all. The industry entered a new golden age, far bigger than the Atari era. Nintendo became the leading company in the console market and remains one of its biggest players today.
Looking back, the timing is striking. In the same year, Atari was burying unsold games, while Nintendo was launching a new generation of game consoles. One era was ending. Another was already beginning. They were just happening on opposite sides of the Pacific.
3. Forty Years Later, History Starts to Rhyme Again
More than forty years later, history seems to be rhyming again..
During the first half of this year, I visited several teams that are still working on VR. They are going back to some of the most basic questions about the product. Why does a headset have to be this heavy? Why do the compute, battery, and cooling system all have to sit on your face? Why should the product form we have used for the past ten years be the final form of VR?
As a result, we are starting to see more aggressive experiments. Some teams are moving compute off the head. Some are redesigning the optics and displays. While others were losing faith in VR, these teams were ignoring the noise and quietly building the next generation.
the products I am most excited about are Meta VR Glasses and PICO Space Pro.
This may sound aggressive, but I believe that while the rest of the industry is still debating whether Apple canceled the next Vision Pro, Meta and PICO may already have built the next generation of VR.
Both products are expected to come to market within the next three to six months.
4. It Actually Started With Vision Pro
None of this happened suddenly. If we follow the path of the technology, the signals have been there for a long time.
The story still starts with Apple Vision Pro. After Vision Pro launched, most people focused on its problems: It was too heavy, too expensive, had limited battery life, and did not have enough content.
At more than 600 grams, many people saw Vision Pro as proof that high-performance VR headsets had hit a physical limit.
But people overlooked another important side of Vision Pro: It had already shown us part of the path toward the next generation of headsets.
At least on the hardware side, I believe two changes have been seriously underestimated.
The first is Micro-OLED.
Vision Pro’s Micro-OLED display system can fit a huge number of pixels into a very small display area. Higher pixel density means the same resolution can be achieved with a smaller display. That also makes more compact optical systems possible. As displays, optics, and mechanical design continue to improve together, headsets can become smaller and lighter without giving up image quality or field of view.
The second change is that Apple moved the battery off the headset.
When Vision Pro launched, many people actually saw this as a weakness. A cable running to a battery in your pocket did not look like something Apple should make. But looked at another way, Apple broke one of the basic assumptions behind standalone headsets: Why does everything have to be on your head?
If the battery can move off the head, why not the compute? And in the future, why not move even more parts out of the headset? Once you start thinking in this direction, the whole design space opens up.
5. Why Didn’t Vision Pro Do It?
So the question is: if Apple already had the technology to make Vision Pro smaller and lighter, why did it still end up so heavy?
I don’t think this is purely a technology problem. Apple has some of the best chip, display, optics, mechanical design, and supply chain capabilities in the world. If you look at Vision Pro part by part, almost everything is top-tier.
And that may be part of the problem. The best parts do not always add up to the best product.
In the Steve Jobs era, Apple was smaller and had fewer resources. Product decisions were more top-down. Leadership had a clear idea of what the final product should be, and every team had to work toward that goal.
Apple today is much bigger and has far more resources. But it also has far more to lose. That can make product decisions more conservative. Each team can optimize its own part: the chip, the display, the optics, the industrial design. But when every part is optimized on its own, the final product may not be the best overall solution.
Vision Pro is a good example. It wants the best display, high compute performance, premium materials, EyeSight, broad use cases, accessibility, and many other features. Each choice makes sense on its own. Put them all together, and you get a headset that weighs more than 600 grams.
A startup can make a very different choice. A company with only a few hundred people can decide on day one: We are building a 200-gram headset. To get there, we can move out the battery. We can move out the compute. We can cut features that are not essential. We can redesign the cooling system. A startup does not have to win on every dimension. It only needs to be much better at the one thing that matters most.
So after Vision Pro put pieces like Micro-OLED and an external battery on the table, Apple’s competitors started doing something very interesting. They did not keep copying Vision Pro. They started taking Vision Pro apart. They kept only the parts that truly needed to be on the face and moved everything else out, one piece at a time.
6. It Is Not About Giving Up Performance. It Is About Moving Weight Off the Head
Once you move the compute and battery off the head, the result can be much more dramatic than people expect. A VR headset can get down to around 100 to 200 grams on the head. The compute and battery move into an external unit. What stays on the headset is mainly the display, optics, cameras, sensors, and the electronics that actually need to be there.
The headset gets much lighter without giving up the core experience. That is very different from earlier attempts to make VR lighter. You do not have to lower the resolution. You do not have to cut the FOV. You do not have to give up compute performance. You do not have to reduce tracking. And you do not have to turn the headset into just a screen in front of your eyes.
This is a different path. The weight comes down, while the core VR and spatial computing experience stays.
7. This Is Bigger Than One Product
I am not talking about one single product. If only one company had built a 200-gram headset, we could simply call it a brilliant piece of product design or engineering. But over the past six months, as I looked at these products together, I realized something: This is not one product. It is a very clear technology trend.
First, there is Meta VR Glasses, at only around 100 grams. Of course, Meta made trade-offs to achieve the glasses-like form factor, including a horizontal FOV of only around 70 degrees.
Then there is PICO Space Pro, with an FOV above 100 degrees and a weight of a little over 200 grams.
There is also the ultra-light headset demo shown by Play for Dream at CES, with an FOV above 100 degrees and a weight below 200 grams.
Another example is Pimax Dream Air, with an FOV above 100 degrees and a weight of around 170 grams.
And even earlier, there was Bigscreen Beyond 2, with an FOV above 100 degrees and a weight of just over 100 grams.
These products are not built the same way. But they are all moving toward the same idea: Less and less hardware needs to stay on your head.
8. How Light Can It Get?
So how light can a VR headset actually get?
Much lighter than most people think.
Get Wu Xu’s stories in your inbox
Join Medium for free to get updates from this writer.
A Chinese startup called Gravity has already built a prototype that weighs just over 90 grams. According to the team, it has an 80-degree horizontal FOV, a 90-degree diagonal FOV, and full 6DoF tracking. With custom fans, custom display modules, and lighter materials, it could get close to 80 grams. The team believes it could go even lower over the next five years.
The current prototype gives up some FOV to get a more glasses-like design. But if you stop trying to make it look like glasses and build it more like a normal VR headset, the team says you can still get a 100-degree FOV, full 6DoF tracking, and keep the weight around 100 grams.
These are not lab experiments. These are engineering prototypes that are close to production.
So the weight problem that has held VR back for more than a decade may finally be getting solved.
Of course, there are still trade-offs. Some products are lighter. Some keep a wider FOV. Some focus on price. But after a few more generations, I think they will end up in roughly the same place: Around 100 grams and around a 100-degree FOV.
That may also be why Meta calls its new product Meta VR Glasses, not a headset. The message is simple: VR does not have to look like a big headset anymore. It can start to look like glasses.
9. What Split Architecture Really Solves Is the Trade-Off Between Weight and Performance
If we zoom out, I believe the importance of split architecture goes far beyond making today’s VR lighter. It solves one of the most basic problems in VR hardware over the past decade: The trade-off between weight and performance.
With a standalone headset, you are working inside a very hard physical constraint. If you want more performance, you need a more powerful chip, a larger battery, and a more complex cooling system. If all of those parts have to sit on your head, the headset will inevitably become heavier.
If you want the headset to become very light, you have to reduce power, use a smaller battery, cut display performance, and remove features. Eventually, performance suffers.
For years, the industry has basically had two choices: High performance, high weight. Or: Low weight, low performance. Split architecture finally makes a third option practical: Low weight, high performance.
Once the compute moves off the head, many of the physical limits that used to constrain VR performance suddenly disappear. Why does an external compute unit have to use a special low-power XR chip? It can use a higher-power mobile chip. It can use a laptop-class chip. It could even use PC-class CPUs and GPUs.
10. A Wider Price Range, Like PCs
Today there is a common assumption that a VR headset has to cost less than $500 or nobody will buy it.
I believe this assumption mainly comes from the design of earlier standalone headsets.
Split architecture allows much more flexible configurations for different users, opening the market to a much broader range of people.
The idea is similar to a PC. The headset is like the monitor. The external unit is like the computer.
The headset keeps the displays, optics, cameras, and sensors. The external unit handles the CPU, GPU, memory, storage, battery, and cooling. And the two parts do not have to be upgraded together. You could buy a headset with a better display. Or you could buy a more powerful compute unit. Two or three years later, if the compute starts to feel old, you may only need to replace that part instead of buying a whole new headset.
This also opens up a much wider price range.
Nobody asks what a PC should cost. You can buy one for a few hundred dollars. You can also spend $1,000, $3,000, or $5,000. Different users want different levels of performance.
VR could work the same way. Entry-level products could use cheaper displays and chips and sell for a few hundred dollars. Mid-range products could use better Micro-OLED displays and faster processors. High-end users could use laptop-class or even desktop-class compute and get performance far beyond today’s standalone headsets.
There is often a question of whether PICO Space Pro and Meta VR Glasses are priced too high compared with Quest. I think that is the wrong question. Different products can have different specs and different prices. Just like PCs. The real point is that split architecture makes all of these options possible.
It does not just make VR lighter. It also raises the performance ceiling and opens up a much wider price range.
11. What About OST AR Devices?
如果分体架构VR头显能降到100到200克，另一个问题就来了：
What does that mean for OST AR?
To answer that, we first need to understand the main advantage of waveguide displays. Today, waveguides are one of the optical technologies with the best chance of fitting into something that really looks like normal glasses.
They allow the display system to fit into a form much closer to normal lenses and frames. Users can see the real world directly through the lenses while digital information is overlaid on top.
So the full value of OST appears when the device really becomes a pair of glasses.
But that is also where the problem is. If the device is not yet truly glasses, the cost of using OST may not be worth it.
Take Snap’s latest AR glasses, SPECS, as an example. It is one of the most complete OST devices in the world today. It uses a waveguide display, supports hand tracking and standalone compute, weighs around 130 grams, and has an FOV of only around 50 degrees. Putting all of this into a 130-gram standalone AR device is already an impressive engineering achievement. But as a product, it sits in a difficult position. As glasses, it is too heavy. As a headset, it is not immersive enough. Its real competitor is actually the headset. And when it is compared with a 100-to-200-gram VR headset with a 100-degree FOV, the question becomes obvious: Why should users give up half the FOV and better image quality just to get something that looks slightly more like normal glasses?
So I respect the progress this path has made. But as a product category, I do not think it is very competitive yet.
In the future, XR may move in two very different directions. On one side are real glasses. They are light, natural, and comfortable enough to wear all day. They do not need a 100-degree FOV or the best image quality. On the other side are real headsets. Of course, smaller and lighter is always better. But they do not need to look like normal glasses. If users are already willing to wear a headset, then the product should push FOV, resolution, and performance as far as possible.
Products caught in the middle may struggle. They are not light enough to feel like real glasses, but they also cannot match the experience of a real headset.
12. What About AI Glasses?
This leads to another question.
How should we think about AI glasses?
AI glasses are doing something very different from VR headsets.
Today, AI glasses are more like a smartphone accessory and an AI interface than a new computing platform.
If you strip away all the hype and buzzwords, AI glasses are basically three things: A camera, audio, and a display.
The audio overlaps with open-ear headphones. The camera overlaps with action cameras. And the display overlaps with smartwatches.
AI has made all of this much more useful. But glasses face a basic limit: power and heat.
A phone can have a large cooling system. A computer can have fans. An external compute unit can have fans too. But glasses only have a frame, lenses, and thin temples.
There is very little space for a battery or cooling. So the chips inside glasses cannot run at high power for long. That makes full spatial computing, SLAM, and high-quality graphics very hard to run in this form factor. Even if most of the compute happens on another device and is streamed to the glasses wirelessly, the glasses still have to deal with their own power and heat limits. There is simply not enough room for a large battery or a serious cooling system. This is not something you can solve just by spending more money. It is a physics problem.
That does not mean AI glasses are not useful. AI glasses do solve real problems, and I expect this to become a big and active market. AI companies, phone makers, and big tech all have good reasons to enter. For them, glasses are not a standalone hardware business. They can be part of the smartphone ecosystem and also serve as a new interface for AI.
So I see AI glasses as an extension of today’s computing platforms, rather than a new platform of their own.
13. So What Is the Second XR War?
Now we can come back to the title.
Why do I think split architecture is where the second XR war begins?
If the question is which products will sell the most over the next two or three years, the answer may not be split-architecture devices. Standalone headsets and AI glasses may still sell far more.
But selling more units today is not the same as becoming the next computing platform. A true next-generation platform still needs to answer a few basic questions:
Can it deliver a real step forward in computing?
Can it change how we interact with computers?
Can it enable experiences that phones and PCs cannot?
And can it become something people actually want to use?
If we use these questions to compare the different paths in XR today, I think the answer becomes clearer.
Right now, split architecture gives us the best chance to deliver the next computing platform.
We have a chance to get both a much lighter device and a true next-generation experience.
That is why, after comparing all these different paths, I keep coming back to split architecture. I think this will be the main story of the second XR war.
With Meta VR Glasses and PICO Space Pro coming to market, we are about to enter the product validation stage.
At this stage, sales are not the most important thing. The bigger questions are:
Will a lighter headset bring new users into XR?
Will people use it longer? Will retention improve?
And what new use cases become possible when the headset gets much lighter?
If the answers are yes, then this generation will have done its job.
After that, more companies will come in. Phone makers. PC makers. Consumer electronics companies. That is when the second XR war really begins.
14. Looking Back at VR
If we look back at how VR has developed, one thing is clear. It has never followed a straight path. The PC followed a more familiar path. First came desktops. Then laptops. As chips and batteries improved, more and more computing power moved into smaller devices, and eventually into smartphones.
VR took a different route. It started with PC VR. Then, to reach more users, the industry moved to wireless standalone headsets. Devices like Quest put the display, compute, battery, and cooling all on the head. That made VR much easier to use. But it also made the headset heavier. Smart glasses made the opposite trade-off. To look and feel more like normal glasses, they gave up a lot of performance.
Now split architecture may look like a step backward. The cable is back. The external computer is back. It can feel like going from a smartphone back to a laptop, or even a desktop. But this may be exactly the step VR needs right now.
Maybe one day we will have a fully wireless standalone VR headset that weighs less than 100 grams and delivers PC-level performance. When that happens, we may look back and think this is what VR was always supposed to be.
But we cannot jump there in one step. We may need to go through this generation of split architecture first.
The second XR war starts with a reset. For hardware companies that fell behind in the last cycle, this could be a new chance. For developers, the playing field itself may already be changing.
A sea change is underway.
The best is yet to come.
