The problem
Local split-screen co-op has been disappearing from PC games for a decade. Modern multiplayer titles like Path of Exile 2, Helldivers 2, Diablo 4, and Destiny ship online-only because supporting four simultaneous viewports on one machine is expensive. It taxes the GPU, complicates UI design, and requires dev work most studios can't justify.
But people still want to play together in person. The current options for "four friends, one living room, one online co-op game" are all bad:
- A console. The split-screen library is shrinking, and PS5/Xbox don't support these PC titles anyway.
- A LAN party. Works, but it's four monitors, four cables, four setups, and the room stops feeling like a couch.
- Remote Play Together. Solves "playing with remote friends," not "playing with friends already on the couch."
There's no clean answer for the case that matters most to people who actually own these games.
The idea
Each player brings their own SteamOS device. Four games render locally, in parallel, and composite into one split-screen on the host TV.
This works for any combination of SteamOS hardware. Four docked Steam Decks at a friend's place. Two Decks and two Steam Machines. Four Machines in a permanent LAN setup. The architecture treats every SteamOS device as an equivalent compute unit on the network.
In one sentence: each player's device renders its own viewport at full quality, streams the rendered frames over LAN to a designated host, which composites them into a quad-split output for the TV.
This is the architectural inversion of Remote Play Together. RPT has the host doing all the rendering and streaming out to clients. Additive Multiplayer has each client rendering locally and streaming in to a host that does only compositing.
Compared to alternatives:
- Single-machine split-screen. One GPU draws N viewports. Quality compromised, demanding games unplayable.
- Remote Play Together. Host renders everything for everyone. Same GPU bottleneck, just with distributed displays.
- Additive Multiplayer. Each device renders one viewport. Load scales linearly, no quality compromise per player.
Why Valve specifically
The technical pieces already exist in Valve's stack:
- Streaming codec. Steam Link / Remote Play, battle-tested and very low-latency on LAN. Works identically on Deck and Machine.
- Network discovery. Steam Networking already handles peer pairing.
- SteamOS. A controlled platform shared across all of Valve's hardware, where the compositing layer can live.
- Family Sharing. Partial answer to the multi-license question.
What Valve would need to build: a SteamOS-level "host mode" that accepts incoming viewport streams from other SteamOS devices on the LAN and composites them, plus a client mode that discovers a host and offers to join the couch session. Most of the engineering already exists. The new work is the compositor and the join UX.
Why this is a SteamOS hardware differentiator
Sony and Microsoft can't really do this. Their consoles aren't designed to be daisy-chained, and they don't have Valve's streaming-tech maturity. Valve's hardware lineup is the only living-room platform that can credibly pitch "stack them for couch co-op."
The interesting wrinkle for the existing Deck audience: this gives every Steam Deck owner an immediate use case that doesn't exist on any other platform. Throw your Deck in your bag, plug into a friend's LAN, controller pairs to your Deck, you're in the session. The Deck becomes a portable couch-co-op enabler in a way no handheld has ever been. No one else can ship this because no one else has a handheld and a console-form-factor PC running the same OS with the same streaming stack.
The economic story for new sales is also clean: each additional player needs their own SteamOS device. A four-friend session is four units that are either being used or being purchased to get used. The marginal gap between "what I'd buy for myself" and "what I need to play with my friends" disappears.
Why now
The 2026 Steam Machine launch is the window for the platform-feature pitch, but Deck owners get value from this on day one of any rollout. RAM shortages have pushed Steam Machine pricing toward premium territory, which makes "buy one beefy PC capable of 4-way split-screen" even more painful as an alternative. A modular pitch like "you and your friends already own most of the hardware between you" sidesteps the sticker shock and reframes the platform as inherently social rather than per-household.
Prior art
Nucleus Co-op already implements a community version of this on Windows: multiple game instances on multiple LAN-connected PCs, composited to one or two screens. It's per-game, hand-built, and unsupported, but it's proof the technical concept works in practice. Valve productising it would turn a hacker workaround into a platform capability and give the entire SteamOS hardware lineup a defining feature.
The ask
A SteamOS feature for Additive Multiplayer: any SteamOS devices on the same LAN converge into one TV with composited split-screen. Per-player headphones for audio, per-device controllers for input, full quality per viewport. Works with Steam Machines, docked Decks, future Frame, or any combination.
It's the only couch co-op architecture that scales linearly with player count, works for online-only games, and gives every existing SteamOS device a use case it doesn't have today.