ExplorXR

Why Valve and Meta Are Shipping While XR's Best Studios Shut Down

· 15 min read

Three platform holders, three imminent launches

In one two week window, every company with real distribution in XR signaled that new hardware is about to land.

Valve’s Steam Frame is close enough that a Linus Tech Tips side channel, ShortCircuit, accidentally published a 17 minute unboxing and first impressions video titled “Our first 30 mins with the steam frame” before pulling it private. The video lines up with earlier preview coverage and adds one new hard number: an Ergonomics Accessories Kit at $60, which bundles an extra head strap and Knuckles style controller straps. It confirms nothing about price or release date, which is the part the industry actually wants.

The software side is further along. Valve released SteamVR 2.17 to all users on September 11, rebuilding the Dashboard with Steam Deck style Quick Access controls and pulling achievements, guides and notes into an overlay that behaves more like Steam than like a VR utility. Earlier, Valve published a “Great on Frame” store page listing titles certified for the headset, which at that point held exactly five entries.

Meta’s next headset also leaked, and this one is more revealing. UploadVR found clear visuals of Project Phoenix inside Horizon OS firmware, shipped inside a prescription lens package that Meta distributed as an ordinary ZIP. The renders show a slim open periphery headset with glasses style nose pads instead of a face interface, sensors along the temples, and a compute puck cable attached to the left temple. The Connect 2026 agenda includes a session on controller free interaction, which lines up with a device built around eye and hand tracking.

ByteDance is doing the same thing on a different calendar. After Pico cancelled its September 2 launch event, Pico Space Pro cleared FCC certification for two model numbers, with the launch pushed to Q4. The teased spec sheet is the most aggressive of the three: 4000 PPI micro OLED panels, a dual chip architecture with a self developed coprocessor for computer vision and image processing, and a main processor rated at twice the CPU and GPU performance of the Snapdragon XR2 Gen 2 inside Quest 3 and Pico 4 Ultra, running the visionOS influenced Pico OS 6 and a RealityKit style Pico Spatial Engine.

Notice what changed in the spec race. None of these devices is selling resolution. They are selling weight, wearability and gaze plus pinch input, with compute moved off the face entirely. That is a mature, product level competition, and it is genuinely good news for the platform business.

In the same stretch, the middle of the industry disappeared

Now the other list.

Lynx, the French startup that promised the widest field of view of any standalone headset, entered compulsory liquidation in March, and EssilorLuxottica acquired the company, including its IP, assets and most of its team. Founder Stan Larroque confirmed the sale in July, said the legal paperwork would take a few months, and then left for a leadership role at the drone maker Parrot rather than stay with the buyer. The Lynx-R2 was specced around the same XR2 Gen 2 as Quest 3 and Pico 4 Ultra, but paired with aspheric pancake lenses developed with Hypervision for a 126 degree horizontal and 103 degree vertical field of view. It was, on paper, the most interesting mid market headset in development. It is not shipping.

Polyarc Games, the studio behind Moss, shut down after nearly twelve years, following a March round of layoffs that cut two thirds of the team. The Moss franchise collected over 160 awards and nominations. The studio’s competitive VR title Glassbreakers sits at a Mixed rating from just 22 Steam reviews, and its flatscreen port of both Moss games, Moss: The Forgotten Relic, launched in July carrying more than 200,000 wishlists. The wishlists did not convert into enough sales to keep the lights on.

Triangle Factory shelved Breachers: Outbreak, a co-op PvE extraction shooter that opened June’s VR Games Showcase as the first big reveal, less than three months after it was announced. The same studio is shutting down its long running shooter Hyper Dash in November. The Binary Mill put Survive The Night on indefinite hiatus weeks after revealing it. Schell Games, a twenty plus year old studio that had publicly framed avoiding layoffs as a philosophical choice, confirmed its first ever layoffs, joining Mighty Coconut, Kluge Interactive and nDreams on the 2026 VR list.

These two lists look contradictory. They are the same story told from two ends of the value chain.

One mechanism behind both lists

Hardware and content run on different clocks, and only one of them is funded by platform scale.

Start with who pays for what. Valve is about to launch a headset while having confirmed it has no new first party VR game in development, which sounds absurd until you look at what Steam Frame actually sells. It runs SteamOS on a Snapdragon 8 Gen 3 with 16GB of unified memory, plays the x86 flatscreen library through a compatibility layer, and streams PC VR titles through a Wi-Fi 7 dongle. Valve does not need a flagship VR game, because Valve owns the store where everyone else’s games are sold, VR and flatscreen alike. The headset is a distribution endpoint, and the $60 accessory kit is margin on top.

Meta plays the same game with different money. It ships hardware, funds content strategically, and now routes a meaningful share of its XR future through glasses built with EssilorLuxottica, the world’s largest eyewear company, which owns Ray-Ban and Oakley, runs more than 18,000 stores, and in which Meta holds a 3% stake. When Meta needs optics expertise or a spatial computing team, it can buy the parts. When a startup needs to sell headsets, it needs retail, capital and a content ecosystem at once.

That asymmetry sets the ceiling for everyone else. Polyarc’s numbers are the cleanest illustration available: a two title franchise with over 160 awards, a competitive multiplayer bet with 22 reviews, and a ported compilation with 200,000 wishlists. Each of those is an above average outcome for an independent VR studio, and together they still did not cover the cost of staying open. Meanwhile the hardware side is not kinder. Lynx had the best announced field of view in the standalone category and still ended up selling its IP and team to an eyewear conglomerate, with no guarantee the R2 design ever ships. The R2 was not killed by a bad spec sheet. It was killed by the cost of getting a niche device to retail.

The clearest evidence is two lines in a patch note

The most useful data point of the week is buried a third of the way down the Rendering section of a No Man’s Sky patch notes list.

Hello Games shipped the Cosmos update, version 7.0, for the game’s tenth anniversary, and it added eye tracked dynamic foveated rendering on PC VR. The same feature arrived on PlayStation VR2 back in August 2023 with the Echoes update, when players were complaining about blurry visuals on Sony’s headset (UploadVR’s earlier coverage documented that jump in clarity).

Why this matters technically: human vision only resolves fine detail in a small central region, so dynamic foveated rendering keeps full resolution and full shading rate in that foveal zone and steps down sampling density in rings outward, using variable rate shading to render peripheral tiles at lower cost. The constraint is latency. The renderer needs a gaze direction before it shades the frame, which means predicting where the eye will be at photon time, not where it was when the sensor last reported. Standalone headsets faked this with a fixed peripheral map, which needs no eye tracking at all. Real adaptation requires a headset with eye tracking, an OS and driver layer that exposes gaze quickly enough to be useful, and an application that consumes it in the render loop. That makes it a platform feature in practice, not a graphics trick a developer can just add.

So why did Hello Games bother? No Man’s Sky is a heavy procedural game that fights for CPU headroom before it ever touches the GPU. UploadVR’s reporter described dropping resolution scale to around fifty percent and graphics to just above the lowest setting on an RTX 3070 with a Ryzen 5 5600X, then upgrading to a Ryzen 7 9850X3D and an RTX 5070 Ti and still having to lower settings from what the game runs at on a monitor. On a title like that, a rendering technique that recovers GPU time in the periphery is not a nice to have, it is the difference between playable and not.

But the timing is the tell. PC VR has had eye tracked headsets for years. The August 2026 Steam Hardware Survey puts the PlayStation VR2 at 2.24% of SteamVR users and the Quest Pro at 1.51%, the two highest shares among eye tracking capable hardware, against a platform that runs 1.8 to 3 million monthly users. Even at the top of that range, the combined audience is a little over 110,000 people, and there is no chance all of them play No Man’s Sky. Hello Games shipped the feature days before Steam Frame launches because the eye tracking install base is about to change shape, and it wants to be in the “great on Frame” conversation when it does.

That is rational behavior for a studio with money. It is also the mechanism that squeezes everyone else: developer effort flows toward whichever platform holder ships next, and studios without a platform to align with do not get that tailwind.

The catch: Steam Frame’s price is the whole thesis

If the eye tracking wave is what rescues content economics, Valve has to move units, and that is not guaranteed.

Valve said in February that the memory and storage shortage forced it to revisit both pricing and shipping schedule for Steam Frame, Steam Machine and the Steam Controller, with DDR5 pricing up roughly 300% according to PCPartPicker data cited by Road to VR’s coverage of the shortage. Valve also confirmed to a YouTuber that the Steam Machine will not be subsidized the way a console is, and that machine landed at $1,050. Valve had told Road to VR it expected Steam Frame to come in cheaper than the Index, whose full kit was $1,000 with the headset alone at $500. If Steam Frame follows Steam Machine upward instead, the influx of eye tracked users that No Man’s Sky is betting on becomes a few hundred thousand existing enthusiasts rather than a new cohort, and the squeeze on studios continues.

The second objection is that the middle is not uniformly dying, it is adapting. Superbright went quiet on In Death: Unchained for two years, cut to a smaller team, and is now bringing the game back with a new season every six months, describing the war in Ukraine and the broader games downturn as compounding pressures. Gunman Contracts, a Half-Life: Alyx mod turned standalone shooter, chose a hybrid VR and flatscreen Early Access launch instead of waiting for Steam Frame to set the date. Polyarc’s own port of Moss to flatscreen was the right diagnosis, delivered too late to save the studio.

So the honest reading is not that XR is dying. It is that XR revenue is narrowing toward hybrid titles and platform aligned products, and studios in the middle get squeezed from both sides.

What this means if you build XR

Three practical rules fall out of the week.

Align with platform gravity. If a platform holder is about to ship eye tracking or a new input model, implementing it now is worth more than a resolution bump, because both the hardware sales and the press attention will be pointed at that device, and the storefront will be looking for things to feature.

If you cannot align with a platform holder, hybridize. Gunman Contracts and Moss: The Forgotten Relic point the same direction: when the VR install base is a few million and the flatscreen market is two orders of magnitude larger, shipping a version that runs without a headset is not a betrayal of the medium, it is survival funding.

Treat expansions as products. Valve is selling a $60 strap kit for a headset that has not been priced. Superbright is reviving a six year old roguelike with seasonal content rather than building a sequel. The lowest risk revenue in VR right now is content for people who already own the hardware.

Two dates will decide how the next year goes for everyone in the middle: Steam Frame’s price and release date, and Meta Connect, where a session on controller free interaction will reveal whether Phoenix is a shipping product or a roadmap. If both land as hardware that people can actually buy at a price they can justify, the install base finally expands and the studio math improves. If both land as premium products for people who were already going to buy in, then the next twelve months of XR will be another spec race, and the list of studios that did not make it will get longer.


Sources

Related reading