en.Wedoany.com Reported - Shadow plans to redirect its idle GPU computing capacity toward the artificial intelligence sector, aiming to provide more sovereign computing infrastructure for European businesses and developers. At the 2026 VivaTech conference, the company—part of the Synfonium group (alongside Qwant)—showcased its new direction: leveraging its European GPU infrastructure for specialized purposes such as AI inference, without abandoning its traditional cloud gaming and cloud computing businesses.
Since 2016, Shadow has offered remote access to high-performance machines, allowing users to utilize them without investing in high-end computers. Initially targeting gamers, the service later expanded to creative professionals and businesses for virtualized workstations and graphics processing. This trajectory led to the establishment of a GPU infrastructure located in Europe, designed for demanding interactive sessions, but with highly uneven usage throughout the day. Cloud gaming peaks between 6 PM and 11 PM, while professional use is more concentrated during office hours and does not always consume full capacity. Consequently, Shadow aims to offer GPU resources during idle periods to businesses and developers requiring computing power for AI applications.
The primary use case is inference—using pre-trained models to analyze queries, process documents, generate results, or produce responses. Compared to training large models, inference workloads are lighter but still require GPU resources as data volumes increase. For Shadow, the GPUs are already in place and need to be better monetized.
This direction aligns with Synfonium's broader plans, which also include Qwant and its API for businesses and alternative search engines—Staan. For AI services, Staan must provide up-to-date and accessible web access, while Shadow supplies the computing power needed to execute queries. Thus, the group seeks to cover multiple technological dependencies simultaneously, rather than reducing the sovereignty issue to merely a matter of location.
These dependencies have proven capable of straining services. In 2025, Microsoft terminated the Bing Search API, forcing clients including Qwant to migrate to other solutions. More recently, Anthropic disabled access to its most advanced models within the EU due to decisions by the Trump administration. In both cases, European enterprises learned the same lesson: an API, model, or computing resource can quickly become unavailable if it depends on suppliers subject to other industrial or political decisions. Therefore, this is not merely about choosing local providers on principle, but about reducing service interruptions for tools now integrated into business operations—a way to build sustainable European alternatives on a service-by-service basis.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
