Google Stadia: Leaked Documents Explain its Failure


As part of the Microsoft-Activision merger evaluation, a number of opinions were gathered across the industry and there was a leak of one document by Dov Zimring (now available on Reddit), who was leading the Project Stadia project since its beginnings. Through this document we get to see in a lot more details the reasons why Stadia has failed before it finally shut down, from a Google’s perspective.

The Stadia Formula

Google Stadia as a project started in 2014 (9 years before this letter). As for the details of the service, you should all be aware of them by now. As highly suspected, the whole go-to-market approach was not coming out someone’s arse, but was elaborated through careful market research principles:

Stadia established pricing, positioning, and content acquisition goals based on extensive market research. We conducted consumer surveys, qualitative and quantitative segmentation, and conjoint analysis to determine offerings we believed would be optimal for users, publishers, and Google. The strategy of offering a free and paid tier, as well as the price for the paid tier (Stadia Pro), was determined through quantitative analysis primarily from our conjoint analysis. Among other things, this research established that the optimal catalog would need both a large number of games (breadth) and the latest and most popular games (depth) to attract users to the platform.

Market researches do not guarantee market success, but the more you do it, the more you increase you chances to, at least, not completely fail right at launch of your service. There is very good evidence, across multiple companies, that market research is a sound way to assess and approach a market you are trying to create or approach.

The promise of Stadia was eventually to make it possible to play games that would be impractical on physical, consumer hardware:

We envisioned numerous opportunities for cloud-native games that would not be possible with the constraints of traditional consoles or PCs. We showcased some of these technologies when we announced Stadia at the Game Developers Conference in March 2019. We established SG&E to lead the industry in developing these new gaming experiences.

This failed to lead to tangible results, but it would have been interesting if they actually had such kind of experience available right at launch.

The Tech Stack

The choice of the platform to run games on was apparently a trade-off between two strategies: enable future experience, while making it possible to also play existing games.

When Stadia was selecting the technology stack, including graphics processing unit (GPU), operating system (OS), and graphics application programming interface (API), for the Stadia servers that would run games, we considered several alternatives. We faced a decision between, on the one hand, technologies that would provide the most flexibility and opportunity for us to realize our vision and, on the other hand, technologies with which third-party game developers were already familiar and for which they could more easily develop new games and port their existing catalog of games. We ultimately opted to use technologies, including the open-source Linux operating system and Vulkan graphics API, that provide the greatest flexibility and opportunity for customization, as we believed this would enable the creation of new types of games, lead to the best end user experience on Stadia, and the best streaming quality.

This led to increased costs for any game studio wanting to go on Stadia - as porting efforts cost time and money:

One negative consequence of this decision was that it increased the cost for external game studios to build games and port existing games to Stadia, despite Stadia’s considerable investment to ease these costs through developer tooling, middleware, and personnel to directly support game studios in Stadia game development.

Windows was apparently considered at Google’s, as an alternative, but licenses costs and lack of customization made the choice easy to make:

An alternative that we considered to the Linux and Vulkan technology stack that would have reduced the costs for external game studios was enabling games to run on Windows and Microsoft’s DirectX graphics library, but we ultimately determined that Windows licensing fees were cost prohibitive for the long term viability of the service and that use of Windows/DirectX would not allow for the customization needed to create the high-quality end user experience that Stadia envisioned.

Stadia had some key technical advantages as well:

Stadia was the most advanced cloud gaming service and generally outperformed competing cloud gaming services in key metrics including video quality, smoothness of performance, overall performance, latency, and audio performance. For example, Stadia Pro offered a graphics resolution of up to 4K and 10-bit color (HDR), features not available on competing cloud gaming platforms.

Market Test

Ultimately user feedback proved that among all the different cloud gaming options out there, the main drivers for adoption were the breadth of the catalog as well as network effects:

For Stadia to succeed, both consumers and publishers needed to find sufficient value in the Stadia platform. Stadia conducted user experience research on the reasons why gamers choose one platform over another. That research showed that the primary reasons why gamers choose a game platform are (1) content catalog (breadth and depth) and (2) network effects (where their friends play).

In order to increase their catalog, Stadia secured partnership with several key actors:

Stadia vigorously competed with other gaming services and partnered with several publishers to bring their games to Stadia. For example, Stadia partnered with the video game publisher Electronic Arts to bring its popular franchises including FIFA, Madden NFL, and Star Wars: Jedi Fallen Order to Stadia. In addition, the video game developer and publisher Ubisoft was an early supporter of Stadia and was Stadia’s most prolific publisher with over 30 games brought to the service.

Google even invested in numerous tools to make it easier to port games to Stadia:

Google attempted to make it easier for video game developers to bring their games to Stadia. For example, Google created low-change porting, which is a tool kit for developers that includes: (a) partnerships with Unity and Epic Games (developers of two of the most popular game engines, software development programs that allow video game developers to create new games more quickly and easily) for their game engines to support Stadia, (b) the creation of libraries that auto-translate DirectX to Vulkan, and © cloud native play testing and quality assurance options. Google also helped video game developers by covering some or all of the costs of porting their games to Stadia, thereby investing a significant amount of capital to bring the popular games that players want to (1 line redacted here) of funding to subsidize developers’ porting costs.

But the problem that it was never enough compared to what other cloud gaming services could offer:

However, Stadia never had access to the extensive library of games available on Xbox, PlayStation, and Steam. More importantly, these competing services offered a wider selection of AAA games than Stadia. Our internal research confirmed that players expect to have access to many of these games on their chosen platforms.

And most developers default to Windows anyway, so you are fighting against a cultural mindset:

This default preference for developers to use Windows made it easier for cloud gaming services based on Windows to bring existing games to their services from a technical perspective even if it creates additional licensing costs. For these same reasons, developers are likely to develop future titles for Windows first, requiring those titles to undergo porting to non- Windows platforms as well.

Giving up on First Party titles

This was the beginning of the end, when Google dropped the idea of developing games in-house for Stadia:

However, on February 1, 2021, Google announced the closure of SG&E. The decision to close Stadia’s internal game studios after less than 14 months of operation was in large part a result of the increasing costs of creating best-in-class video games. These games require many years of development time, specialized expertise, and significant investment.

At the same time, Google was trying to pivot Stadia into a B2B business by selling their Stadia tech to other companies interesting in implementing it, with their offer called Immersive Stream for Games:

In February 2021, at the same time that Google decided to close SG&E, Google revised its business plan for Stadia in an attempt to keep the service in operation. In the face of Google’s inability to create the blockbuster video games that gamers demand and the growing cost of securing AAA games on Stadia, Google lowered Stadia’s subscriber (1 line redacted here) showcase for Immersive Stream for Games.

This did not change the trajectory for Stadia:

Even with this drastic reduction in expectation and the scope of the service, Stadia was still unable to attract enough subscribers to maintain a viable business. Without any clear path to reaching a critical mass of subscribers, Google announced on September 29, 2022 that Stadia would shut down in January 2023.

Lessons for Linux Gaming

This is a clear second data point that we can use to showcase that asking developers to port games to Linux just does NOT WORK. Valve tried it first when they first launched their Steam machines, by relying on dev’s goodwill, and it ultimately failed to get any traction. Then Google went back at it, this time on the cloud gaming front, and despite having very deep pockets they were completely unable to get developers and publishers to care about porting games to their Linux-based platform:

While Stadia enjoyed success in supporting smaller developers’ efforts to port their games to Stadia, we were unable to reach agreement with the developers of AAA games who complained about the cost of porting their games to work on Stadia given the lack of users on Stadia even with the subsidies that we provided.

In any case, this is a very interesting and important document that confirms what many of us suspected. This kind of post-mortem business evaluation does not usually go public, so this is a great piece of insight from a major player in this field.