10,000 Sessions in a Month: The Decentralized Cloud Gaming Experiment Nobody Is Watching

Cloud gaming has a history of promising the moon and delivering a buffering icon. Stadia died, GeForce Now got more expensive, and every new platform seems to rediscover the same lesson: streaming games is easy, making the economics work is not. A Dutch startup called YOM is now trying a genuinely different answer, and its first four months of real data just landed.

YOM reported 10,196 cloud gaming sessions on its decentralized network in September 2026, up from 864 in June. The sessions all flowed through Walter.gg, the company's free browser-based alpha, which streams games from independently operated gaming PCs rather than data center racks. The report, shared by COO Jeff Outlaw on October 8 and corrected on October 9 after the company caught errors in its June and July figures, shows usage more than doubling every month on average.

A cloud made of gaming PCs

Here is the idea. Instead of building giant server farms full of GPUs, YOM coordinates machines that already exist: gaming PCs owned by independent operators who join the network and earn for hosting sessions. When you play on Walter.gg, your game session runs on someone's PC, is encoded, and streams to your browser on your phone, laptop, or tablet.

That makes YOM a DePIN project, part of the decentralized physical infrastructure movement that uses tokens and protocols to coordinate real-world hardware. The company's $YOM token launched on the Avalanche network in June 2026, days before YOM was selected for the inaugural Discord Gaming Founders Circle alongside companies like Poki. But the business pitch is simpler than the crypto plumbing: cloud gaming without the cloud bill.

The traditional cloud gaming economics are brutal. Providers pay for racks of GPUs that sit idle whenever players are not logged in, then charge subscriptions or hourly rates to cover the overhead. YOM's bet is that a network of home machines, already bought and already powered, flips that equation. If the numbers hold up, the marginal cost of a session trends toward the operator's electricity and bandwidth, not a data center lease.

Three games, one free alpha, zero revenue. And still, the network keeps doubling.

The numbers, with the caveats

Let us give the figures their due, because they are real and they are unusual for this industry. June brought 864 sessions, July 2,082, August 5,943, and September 10,196. The first four days of October added 3,329 more, roughly 830 sessions a day versus about 340 a day in September. September alone accounts for about 52 percent of every session the network has hosted since it went live. Four months total: 19,085 sessions.

Now the honesty part, which YOM itself deserves credit for publishing. Every one of those sessions came from Walter's free alpha. None was paid. Sessions measure usage, not revenue, and the base is still small enough that doubling says as much about the starting point as the trajectory. As the company puts it, doubling every month is much easier at a thousand sessions than at a million, and the pace will not hold forever.

Then there is the catalog. All of this growth came from exactly three games: Counter-Strike 2, Warframe, and Marvel Rivals, all free to play, all streamed with the player's own Steam account. PUBG: Battlegrounds and Overwatch 2 are listed as in testing, with no announced dates. You cannot yet connect your full Steam library, Walter tests and approves each title individually. Three games produced this curve, but the curve only becomes a business when the catalog does too.

Why the model is interesting anyway

What makes YOM worth watching is not the session count, it is the structural bet on proximity. Latency is the tax every cloud gaming service pays, and the cheapest way to cut it is to put the machine close to the player. A network of thousands of home PCs distributed across cities is, in theory, closer to players than a handful of regional data centers. YOM's bring-your-own-game model adds another twist: the catalog grows from libraries players already own instead of requiring titles to be rebuilt for the cloud, which sidesteps one of the biggest content bottlenecks in the industry.

The company is also testing a second kind of capacity. A UK trial with Cornerstone, StonesThro, and Zadara ran 168 gaming sessions from mobile tower sites over 20 days. The trial reported no packet loss across the measured sessions but inconsistent frame delivery that needs work before the next phase, exactly the kind of honest technical result that suggests the engineering is real. If telco towers can host reliable sessions, YOM gets always-available capacity that complements home GPUs that may go offline when their owners are using them.

Compare that to the giants. Netflix is pushing cloud gaming through its streaming footprint, and Microsoft's cloud ambitions keep growing, but both are centralized bets: big infrastructure, big capex, big subscriptions. YOM is attempting the Airbnb version, turning idle hardware into capacity and skipping the data center construction entirely.

What has to happen next

The free alpha is a proof of demand, not a proof of business. YOM's own retrospective lays out the to-do list plainly: add more games, launch paid subscriptions to turn usage into revenue, start the YOM AI voice pilots on the same network, and bring telco capacity online. A native Walter app is reportedly in alpha testing with an unnamed TV manufacturer, which would take the service from browser tabs to living rooms.

The risks are structural. Paid subscriptions will test whether players value this enough to open their wallets, node reliability will determine whether sessions route to machines that stay online, and three games will never be enough against libraries that count in the hundreds. Independent analyses have also flagged the hard parts of any decentralized streaming stack: variable node quality, regional bandwidth constraints, and the perennial question of content licensing.

But there is something refreshing about a cloud gaming company publishing its numbers, correcting them when wrong, and naming its own weak spots. After a decade of moonshot promises in this space, a startup saying "demand is showing up, the network can serve it, and here is what we still have to fix" is a format worth encouraging. The next four months, when subscriptions are supposed to launch, will tell us whether the model works as a business or just as an experiment.