Valve's Steam Machines: Was the Idea of Linux-Based Gaming Devices in Vain?


I can still picture it just like it was yesterday. My colleague, pointing one finger to our new product in front of us. Asking me a simple yet subtle question:

“What’s the value proposition?”

“What do you mean?” I was unsure of how to answer.

He continued: “You know… why would one even want to buy this product over something else? What is its reason to be? What problem does it solve?”

This was my introduction to Product Development, Branding and ultimately Marketing. Once you internalize this kind of framing, you cannot unsee it anymore. It becomes like a sixth sense you apply to everything around you. Ads, posters, products you use everyday, products you consider buying, and the list goes on.

It’s like when a friend (or a foe?) teaches you about font kerning just to mess with you. Every single sign, every printed paper that looked absolutely normal before, will seem off and wrong. Life changing knowledge (for the worst, in the case of font kerning).

All of this reminds me of the time when Valve started to entertain the idea of the Steam Machines initiative. For the Linux community, it was a time of great hopes and enthusiasm. Valve was putting its heavy weight behind our favorite OS, Zombies were getting faster, and soon there would be official hardware running on Linux on the market to play Steam games. The ultimate liberation from Windows to play PC games.

However, as more details trickled in, the less enthusiastic I became. The more it appeared as if the Steam Machines were a product in search of a problem to solve. Without a clear value proposition.

I wrote on this topic long before their actual launch. Valve ended up more and more hands-off as time passed after the initial announcements. There are many reasons that we now know about (Steam Machines being an insurance policy to force Microsoft to back down on their exclusivity plans), and others that we can only guess.

The point still stands today: what’s the actual value of having a PC running Linux only to play games, compared to the same hardware running Windows? For most PC gamers, probably none. They don’t mind the Windows 10 telemetry. The unexpected shutdowns for updates. They don’t care about having a proprietary system as long as they can run their favorite applications. And Windows still remains the best bet to run their (Windows) games without a hitch on a PC as soon as they release.

When the Steam Machines launched, they had almost nothing for themselves. Very few native games, with worse performance, and an OS with limited functionality in a world where even consoles offer Netflix and more. It did have a cool and innovative Steam Controller though. A great companion for playing mouse-based games from the couch. For games not compatible with SteamOS, you could stream them from another home-based Windows machine. But was it really worth it? Instead, you could just grab a Steam Link device that would do the job for the fraction of the cost.

Out of these three initiatives, two of them had an actual value proposition. The Steam Link brought your PC games to your living room without a second PC and without noticeable lag. The Steam Controller made it possible to play all sorts of games (Civ, FPS, RTS, etc.) while sitting on a couch far away from a keyboard.

SteamOS… SteamOS, also known as a Linux distro for gaming, was a product in search of a problem to solve.

And it’s very telling, once you look at the history of Google Trends, that we can use as a proxy for interest, how these three initiatives fared over time.

SteamOS only generated very short spikes of interest and quickly disappeared into oblivion even as the Steam Machines launched. This was despite the fact that fixes and improvements landed on SteamOS over time… albeit slowly. In truth, nobody ever cared about it besides a tiny fraction of Linux users (in which I count myself, as I used SteamOS for quite a while as a living room distro). It brought almost nothing on the table that a regular distribution could not do. It did have automatic updates, driver support from the get go, but anything beyond playing very few Steam games required messing with the configuration in the terminal. Might as well stick with a regular distro then. And that’s not even talking about fatal errors within SteamOS.

The Steam Controller and Steam Link devices continued to have a life of their own, well beyond the Steam Machines. While they are now both discontinued, their legacy lives on. The Steam Link turned into a mobile application to stream Steam games anywhere at home or even from the internet. The Steam Controllers have run out of stock and no new production is in sight. Yet it still has a fond following, and the Controller API it helped create now powers the controller support of numerous joypads out there in Steam. This is precisely why you can remap controls to your liking in virtually any game on Steam now, whether you use a Xbox controller, a DualShock 4 or a classic Wii controller.

By the count of what survived, Valve did fail with Steam Machines. If we however look at the conceptual level, some of the ideas they iterated on had value and helped shape Steam into a greater platform.

SteamOS may be the only exception. Too half-baked. Too limited.

If Valve were ever to revist the idea tomorrow, SteamOS may fare better this time around. In what measure? Well, in an hypothetical SteamOS where Proton would be enabled by default, a lot more games would be available to play. A lot more, but not all, which is always ,always the issue. No real person actually cares about having 3568 shovelware titles to buy from if Cyberpunk 2077 does not work on launch day. Gotta have Keanu on Day 1. Anything short of 100% of big, huge titles will make it hard to take seriously. And let’s not even talk about others stores’ support: inexistant, at least in any official form.

But let’s suppose that in some remote future, 99.99999% of Windows games could run on SteamOS out of the box. I’m not sure it would actually change anything. The very same question would still linger: “Why not use Windows anyway?” This is where the actual execution matters a lot. Everything needs to be seamless, almost like magic: OS installation, driver updates, unbeknownst to the user. Companies like Microsoft with the Xbox, Sony with the PlayStation and Nintendo with the Switch spend a LOT of time, money and efforts to NOT SUCK at this kind of things. Valve cannot just improvise, take shortcuts and call it a day, as their team members quietly move their desks to their next side project of interest.

I use Steam’s BPM (the would-be interface of SteamOS) constantly from my couch. The lack of polish is glaring. The hideous “STORE” icon, courtesy of Fiverr. The non-existent improvements in the interface for years. The lack of useful options in the settings. The fact that reviews open up in a browser for the lack of a better UI.

It’s OK, I guess. I mean, it does the job. It launches games. Until one day, a game crashes and you are stuck with a black-screen and you need to connect a keyboard, or use SSH to take control of your now zombie PC because nobody at Valve ever thought this could fucking happen. Let’s say it quickly dispells the illusion of a console-like system, because you don’t want me to go on cursing.

Instead, Valve has recently spent a lot of effort to redesign the desktop client look and feel. Steam BPM is still stuck in another era, and mostly gets updated text menus rather than proper love and consideration. This says a lot.

Which is unfortunate. Despite the initial setback of SteamOS, there has been so much progress. Proton in itself is a absolute miracle. Mona Lisa on steroids. One click and boom, your Windows games (more often than not) launch as if they were native clients. To this day I still find it amazing.

It’s been a wonderful journey, seeing Baby Linux-Gaming make steps and finally walk.

But let’s not kid ourselves: It’s not enough to move the mark.

It was indeed a Herculean task to turn Linux into a viable gaming environment. Valve had to secure Nvidia and AMD’s resources to work on their drivers. They had to create tools and port most of their own games to show that it could be done. They went ahead and developed a custom distro that could be the basis for a full fledged Steam device. They convinced Alienware to try and sell it.

But we can now say it: SteamOS sucked. It was a 30% product in a world where consoles are 90% there. It felt more like a demo than the real thing. To show that it’s actually possible to make a fully autonomous Linux gaming environment, albeit imperfect.

In recent years, the focus on Proton for Linux has reversed the tide a little. Gaming on a Linux environment does not suck as much anymore: we now have recent titles working out of the box, without the need of expensive ports that would take a long time to hit the platform. Performance is sometimes very close to Windows. Close enough that you would not see the difference.

As unlikely as it seemed, Valve and their partners pulled it off: Linux has become a credible gaming platform. Hats off to everyone involved. And Champagne on the house, you all deserve it.

We however need to come to a new realization. We are just half-way.

Instead of reaching the top of the hill, we find ourselves in this strange uncanny valley. A place where Linux is now good enough to do a lot of gaming for a lot of games, with great performance, yet not good enough to compete, survive, and stand on its own legs as a gaming platform.

If Valve walks off tomorrow, who will pick it up?

For Linux Gaming to survive and thrive, it needs to become a central point of focus for more than just one company. It needs redundancy of interest. Just like the Linux kernel exists at the precise intersection of multiple private interests and commercial ventures.

The fact that everything that is produced by Valve for Linux Gaming is mostly done under a FOSS license hardly changes anything: anyone who has been around long enough has seen FOSS projects wither and be abandoned over time. Large scale efforts benefit from solid financial support or better, a business model to move beyond the hobbyist stage.

Valve foots the bill for now.

Are you confident they will do it forever? No need for a plan B?

“Never trust anyone, Daniel, especially the people you admire. Those are the ones who will make you suffer the worst blows.”