The Metaverse Died. Long Live XR.

In 2021, Facebook renamed itself Meta, forging towards a future built around virtual and augmented reality. Five years and roughly $77 billion in losses later, that wave never seemed to really take off. While it feels like a lost cause, the patents tell a different story about how engineers have been adapting to a new narrative.

The bet was that the next computing platform would be built around XR (Extended Reality), an umbrella term covering virtual reality (immersive headsets that replace what you see), augmented reality (digital overlays on the real world), and mixed reality (a blend of both). Meta said it would build the headsets, operating systems, and social spaces.

Five years later, the bet has gone badly. Meta's hardware division, called Reality Labs, builds and sells the company's VR and AR products: Quest headsets, Ray-Ban Meta glasses, software, and the social app Horizon Worlds. Reality Labs has lost money every year since 2020. The losses widened from $10.19 billion in 2021 to $19.19 billion in 2025. The five-year cumulative bill for the post-rebrand era now sits at roughly $77 billion. Revenue across the same period stayed pinned near $2 billion a year. On Meta's January 2026 earnings call, Zuckerberg did not say the word "metaverse" once.

XR as a market category (the broad commercial bucket containing all VR, AR, and MR products) may yet survive all of that. The original pitch has not.

XR has entered a promising product phase. 

Consumer headsets look weak. Smart glasses are promising. Industrial AR has found buyers. Simulation and training are useful in healthcare, defence, and skilled trades. Gaming platforms like Roblox and Pokémon GO built parts of the so-called metaverse without using the label. 

And the patent record shows that the engineering work never stopped. It moved away from world-building and toward smaller problems, such as making the optics work in eyewear, making the input methods less awkward, and making the devices comfortable enough to wear all day.

The grand unifying theory died. But patent filings are still alive, showing that the tech underneath XR is entering a fresh wave.

This is how technology categories tend to mature. The rhetoric expands first. Then the engineering work narrows into the refined, less glamorous problems that determine whether the category becomes useful and defensible. The internet and mobiles both went through it. XR appears to be in that phase now.

Meta has framed Reality Labs’ incredible losses as strategic patience. It can also be read as the company spending its way past a product-market fit problem it never solved.

The XR product reviews were rarely the problem.

IDC reported Meta's Quest VR headset line shipped 42.3% fewer units in 2025 than in 2024 even as Meta cut prices. Apple's Vision Pro, the device that was supposed to validate the premium consumer end of the market, shipped on the order of 390,000 units across its entire 2024 launch year and an estimated 45,000 units in the fourth quarter of 2025 per IDC figures. Meta laid off more than 1,000 Reality Labs employees in January 2026, shut down VR studios, and pushed Horizon Worlds toward mobile.

Quest 3 is competent hardware. The headsets did not fail technically. Vision Pro's optics are extraordinary. They failed because they imposed too much cost and friction in exchange for too little habitual value. The basic exchange the headsets asked of consumers (pay $500 to $3,500, strap a 500-gram visor to your face, withdraw visibly from the room) was never one most people were going to take. Meta lost because consumers were being asked to change too much of their behavior, all at once.

The patent record is one of the better places to watch a technology category mature, because it shows you what the engineers thought was worth protecting. For XR, the picture the patents paint is of a refocused industry.

Over 6,700 AR/VR patents were issued worldwide in Q4 2020 alone, up from roughly 2,300 in Q3. The peak quarter, Q3 2021, saw 2,735 patents granted globally, with Q4 2021 close behind at 2,564. As of 2024 commentary from City/IFI, approximately 390,000 AR, VR and XR-related patent applications were pending worldwide, with portfolios spread across Sony, Microsoft, Samsung, Apple, Meta, Qualcomm, LG, Snap, and major Chinese filers like Tencent and Baidu. Apple alone has filed over 5,000 applications related to Vision Pro and adjacent technologies.

The headline numbers come from law firm Finnegan and patent analytics firm IPlytics. 

The Vitek IP USPTO chart shows AR/VR application volumes climbing from below 1,000 in the early 2010s to between 3,000 and 4,000 a year by the early 2020s, then leveling off. Henry.law's 2024 overview describes the same shape. A leveling-off, not a collapse. The category ran its filing surge into the metaverse hype window and then settled at a much higher baseline than it had pre-2015.

The themes are where the refocus is revealed. Five problem buckets recur across the post-2023 patent record.

Each row names a problem the industry is still trying to solve, links a representative patent family, and explains what the engineers are after.

Patent records via Google Patents. Representative sample, not an exhaustive census.

The market story in 2021 was about a dreamy destination of virtual worlds where you slap on a headset and interact with friends. It’s a bit of fun, but it's not a realistic, consumer ready product category. The patent story in 2026 is about infrastructure. Companies still think XR is worth building for, but in the form of tools, interfaces, components, and specific workflows rather than as one universal metaverse.

That refocus shows up in the market data too. The next section is where the money has gone.

The post-metaverse market is at least four separate markets, each with different buyers, hardware needs, and economic logic.

One headline private-market number frames the rest. AR/VR startups raised roughly $3.28 billion across 117 deals from 2022 to 2025, per Landbase's aggregation. Three companies captured about 76% of the 2025 total. All three were building lightweight wearable computing, not social VR.

Smart glasses are gaining traction because they ask less of the wearer than a headset. Products that try to introduce entirely new behaviour fail more often than products that nudge existing behaviour. People already wear glasses, listen to music in earbuds, and ask voice assistants for directions. A device that combines those three things in one familiar form factor is a far easier sell than a 500-gram visor that asks the wearer to disappear into a separate world.

EssilorLuxottica, the world's largest eyewear company, sold over 7 million Ray-Ban and Oakley AI smart glasses in 2025, more than triple the 2 million sold across 2023 and 2024 combined. Production capacity is being scaled toward 10 million pairs a year by late 2026, and Counterpoint's smart-glasses tracker shows the category up 139% year-over-year in the second half of 2025.

Enterprise AR workflows make immediate economic sense in warehousing, field service, maintenance, guided picking, and remote support. The buyer is an operations lead trying to reduce errors, compress training time, and improve throughput.

The deal flow here is smaller per cheque than in consumer smart glasses, and broader. The pattern is fragmented SaaS-plus-hardware deployments where the ROI is measurable and the cheques are smaller but recurring.

Simulation and training is probably the strongest part of the entire stack. It spans healthcare, defence, skilled trades, and industrial training. The procurement question is whether competence improves faster and more repeatably than the status quo.

The commercial signal is concentrated in healthcare. The FDA had cleared 69 XR medical devices as of September 2024 per a Frontiers review, and the pace held in 2025. These are clinical-grade devices sold into procurement cycles that did not depend on the metaverse story.

The largest single XR cheques still clear in defence. The IVAS program (Integrated Visual Augmentation System) is the largest single AR procurement in history. Microsoft was willing to walk away because its enterprise narrative had moved on. Anduril, founded by Oculus creator Palmer Luckey and structured around defence procurement, was the natural inheritor.

Defence XR demand is the most opaque category in the data because procurement is closed and most public figures are contract caps rather than booked revenue. The buyer logic is durable. Soldiers do not have to choose to wear the headset, which is rare in this market.

Gaming and immersive entertainment still matter, but have pivoted. Roblox, Fortnite, Minecraft, and operators like Sandbox VR show people will spend time and money in persistent or curated digital spaces. Pokémon GO is an interesting case-study, as the largest commercially successful real-world AR product, showing there’s no need for hardware strapped to the player’s face to have a bit of fun.

The 2025 deal flow in this category was dominated by Niantic’s massive exit.

This is where the earlier metaverse forecasts went wrong. They bundled unlike things into one giant market and treated them as if they would converge onto a single platform. The opposite happened. Different buyers, different hardware, different economics, different measures of success. The surviving XR industry is smaller than the original forecasts suggested, and probably healthier than the obituaries imply.

The bottleneck has stopped being mainly about optics, battery life, or interface design.

It is whether people want these systems on their face, around their eyes, and inside their social lives.

The wearability problem has not gone away, and neither has privacy. An always-on AI assistant in glasses sounds useful until you are the person standing across from it. Industrial environments tolerate more friction because the return is measurable and the context is controlled. Public environments are harsher. The product has to clear social, aesthetic, and behavioral hurdles before technical progress matters much.

The privacy problem is already manifesting. In October 2024, two Harvard students built I-XRAY, a system combining Ray-Ban Meta glasses with the PimEyes facial recognition service and public records scrapers. In live subway tests, the system identified strangers' names, phone numbers, and home addresses in seconds. The students declined to release the code, but the demonstration showed what an always-on camera, a generalist AI assistant, and a facial recognition API can do when combined.

The regulatory response is now explicit. Reuters reported in December 2025 that Irish and Italian regulators were pressuring Meta and EssilorLuxottica to improve user-notification systems on Ray-Ban Meta glasses. In early 2026, more than 70 civil-society groups signed an open letter urging Meta to abandon rumoured facial-recognition features in the next generation of smart glasses.

And traditional problems still remain. A 2024 meta-analysis covering 3,016 participants found substantial cybersickness variation by content type, with gaming worst on average. A 2024 review on digital eye strain documented persistent visual fatigue concerns relevant to all near-eye displays. It’s a tricky wearability problem that has not improved at the rate the optics roadmap has.

For all of those reasons, the future of XR may be decided less by immersion than by tolerance.

How much hardware will people tolerate on their face? How much ambient capture will bystanders tolerate in public? How much friction will an enterprise tolerate before the productivity gain disappears?

Those questions are less grand than the metaverse story, and much closer to the economics that will decide what survives.

The metaverse was too broad, too early, and too unified to survive contact with reality.

What came out the other side is narrower:

Meta sold a world.

What survived was a promising toolbox.