Spatial computing had a distribution problem, not a graphics problem.
For a decade the objection to browser-based immersive work was fidelity, and for a decade it was the wrong objection. Plenty of projects that could have run perfectly well in a browser were built as installed applications, and then quietly failed for reasons that had nothing to do with rendering.
They failed because a client could not install anything on a managed laptop. Because a headset app needed an organisational store account and a review cycle. Because the fifteen people who mattered were on four different device types. Because updating it meant redistributing it. Because the person who was going to charge the headsets left.
Almost every immersive project that dies, dies of distribution. Very few die of frame rate.
That is the frame worth holding while reading the rest of this. The technical changes below are genuinely significant, but their commercial value is that they remove the last honest reasons to accept the distribution penalty.
Four things that became true, and roughly when.
None of these is new on its own. The change is that all four now hold at the same time.
What the browser can now do
1. Reach a headset
WebXR ships in the Quest browser,
the Pico browser, and Safari on
Apple Vision Pro.
Not on iPhone or iPad Safari.
Off by default in desktop Safari
and Firefox.
2. Draw at modern GPU level
WebGPU: compute shaders, modern
pipeline, implemented against
D3D12 / Vulkan / Metal.
~85.6% of global usage.
Firefox is the big gap.
3. Show captured reality cheaply
3D Gaussian splats reconstruct real
places photorealistically and render
in real time -- and export to formats
web viewers read directly.
4. Receive a GPU stream, uninstalled
NVIDIA CloudXR.js, announced at GTC
on 31 March 2026: a browser client
for GPU-rendered OpenXR applications
on Quest, Pico and desktop.Take the fourth one seriously, because it is the one that breaks an old rule. Remote rendering used to be the fidelity option and the browser was the reach option, and you chose. CloudXR.js means a server-rendered, full-fidelity OpenXR application now arrives through the same channel as a web page: a link. Fidelity and reach stopped being opposites. What replaces the trade-off is a straightforward operating cost — a GPU per concurrent viewer — which we have worked through in CloudXR versus local rendering.
The signal that matters more than any benchmark.
If you only take one piece of evidence from this article, take this one: Meta built its own WebXR framework on three.js.
Meta sells headsets and runs a store. Its commercial interest is in native applications distributed through that store. A company in that position investing in the open web stack is not a marketing gesture — it is a read on where the audience is going, made by the party with the most to lose from being wrong.
The second signal is NVIDIA’s. CloudXR 6 could have remained a native client SDK. Shipping a JavaScript framework so the client is a web page is the same judgement from the opposite end of the market: the company whose business is high-end GPU rendering deciding the delivery surface should be a browser tab.
Benchmarks tell you what is possible. Where platform owners spend engineering budget tells you what they expect.
What the browser still cannot do.
An honest list, because a piece like this is worthless without one. None of these is close to being solved.
- Offline. A factory floor, a vessel, a tunnel, a remote site. If it must run without a connection, that is an installed application, and no amount of caching changes it.
- Full-fidelity CAD and BIM on-device. There is a hard ceiling on what a mobile chipset renders inside a browser tab. If the master model cannot be reduced without destroying its purpose, the answer is remote rendering or a workstation.
- Deep hardware access. Specialised tracking modes, some body and face tracking, certain device management hooks. The browser surface is deliberately conservative for privacy reasons and will stay that way.
- Long, heavy, physics-driven sessions. Sustained simulation with complex state is still more comfortable to build and run natively. Our own driver assessment work is a desktop simulation for exactly this reason.
- iPhone immersive. Safari on iPhone and iPad does not support WebXR. Vision Pro does — different Safari, different operating system — but a large mobile audience gets the flat view and nothing else.
Notice that none of those limits is about visual quality any more. They are about environment, scale and hardware. That is a meaningful change from five years ago, when the objection genuinely was that browser 3D looked poor.
What this changes for a three-year plan.
Four practical consequences, in rough order of how much money they move.
Procurement stops being the gate. If the people who need to see your work are clients, regulators, partners, press or a board, the ability to send a link is not a convenience. It is frequently the difference between the project having an audience and not having one.
One artefact instead of a matrix. A web build degrades to a flat 3D view where immersive mode is unavailable, so the same thing serves the headset audience and the much larger non-headset audience. The saving is not the second build. It is the second build, forever, every time anything changes.
Content ownership becomes realistic. Publishing a change and knowing everyone has it are the same action on the web. If your content has a clock on it — model years, site progress, course modules, product ranges — this dominates every other factor in the comparison.
Design flat first. This is the rule people resist and then confirm. Most sessions never enter immersive mode. The flat view is the product for the majority of your audience, and it is also the cheapest way to discover whether the idea works at all.
What does not change: content is still the expensive part, immersive work is only worth it when the spatial dimension carries information a flat presentation loses, and somebody still has to own it in a year. The delivery mechanism got much better. The discipline did not get any easier.
How to test the claim in ten minutes.
Rather than take the argument on trust, open things. All of these run in a browser with nothing installed, and all of them are ours, so you can judge the standard rather than a showreel.
- Simam Immerse — a headset-grade immersive media player that arrives as a link: a spatial lobby you walk through and a catalogue of 360° and 180° titles. Same build on Quest and Vision Pro, no store submission.
- The browser playground — a set of live experiences including the product configurator pattern and a spatial walkthrough of a captured environment.
- The splat editor and the gallery of published scenes — photoreal captured places, opened from a link, edited in a tab.
Do it in this order, because it is the order that makes the point. Open one on a laptop and note that it simply works. Then open the same URL in a headset if you have one to hand. The gap between those two experiences is the thing to design around, and it is the whole reason the flat view deserves the attention.
What we would build on the web today, and what we still would not.
On the web: product configurators, design and site review tools, digital twin front ends, training and assessment that is not physics-heavy, virtual venues and events, immersive media delivery, captured-place navigation, and anything at all whose main problem is getting in front of people who cannot install software.
Not on the web: offline field applications, full-fidelity engineering review of unreduced master data, heavy simulation with sustained physics, and anything depending on hardware capabilities the browser does not expose. For the first two of those, remote rendering is now a genuine third option rather than a research curiosity — the CloudXR guide covers when that makes commercial sense.
The honest summary of 2026 is not that the browser has won. It is that the browser has become the correct default, and native has become the considered exception. That is a reversal of where the industry sat three years ago, and it should change the first question in a brief from which headset to who needs to see this, and how do they get to it.
What this means for a buyer.
Start with the business decision, audience, and evidence the project must produce. Simam Digital can turn that into a focused discovery, prototype, MVP, or production roadmap across AI applications, SaaS platforms, digital twins, real-time 3D, XR, and interactive systems.
