The Steam 'Teraleak' Exposes a Decade of Lost Gaming History
AI-generated image · US National Wire
A massive 12TB dump of defunct server data reveals prototypes and cut content, highlighting the precarious nature of digital preservation.
A massive data leak, dubbed the "teraleak," has released more than 12TB of content spanning the history of PC gaming on Steam between 2003 and 2013, as Ars Technica first reported. According to the outlet, the leak consists of a "content server dump" from Valve's now-defunct "Steam2" server architecture. The collection includes thousands of "depots" containing various versions of games, including public releases, prototypes, playtests, and pre-release builds from both Valve and third-party publishers.
The leak cuts off in 2013, the year Valve transitioned from the proprietary Steam2 format to the current SteamPipe system. Ars Technica reports that the SteamPipe update streamlined patches and removed update approval bottlenecks by moving to standard HTTP file trees. Prior to this leak, many of these early versions were considered lost by archivists, existing only as local downloads on old hardware.
Questions remain regarding how the data was accessed. Gabe Follower, a Valve dataminer, stated on social media that the content was obtained via a publicly accessible API endpoint, asserting that the lapse was "Valve's fault." Similarly, analyst and streamer Scolcer noted the content was available on a site that was "100% accessible to the public" without password protection. However, moderators from the Valve Cut Content (VCC) Discord, CrazyBubba and The One Epicplayer, suggested the leak might instead be a collection of files hoarded privately by individuals since the 2013 transition. A user claiming to be the original uploader provided screenshots of a 22TB server upload log on the VCC Discord, stating that releasing the data required "a significant amount of effort."
Gaming communities are currently mining the leak for lost assets. Ars Technica reports that playable pre-release versions of *Portal 2* have revealed deleted sequences, including scratch track dialogue from Cave Johnson and GLaDOS. These prototypes also feature circular portals, slow-motion guns, and a weapon model previously seen in Valve documentary footage of the unreleased *Half-Life 2: Episode 3*. While no playable *Half-Life 3* content was found, researchers have identified "ep3" data files. Other discoveries include early betas of *CS:GO* and *Left 4 Dead 2*, as well as assets from *F-Stop*, a canceled camera-based sequel to *Portal*.