I’m always looking for the “why” behind my favourite services, and Antonina Batova’s recent look at Boosteroid’s infrastructure explains a lot about their performance. It hits on the core struggle of cloud gaming: you can’t just rent generic power and expect top-tier results.
Building a global cloud gaming platform is a massive undertaking that has tripped up even the biggest names in tech. While many companies look to hyperscalers like Amazon Web Services (AWS) or Microsoft Azure to host their workloads, Boosteroid takes a different path. Antonina Batova, the SVP of Infrastructure at Boosteroid, recently sat down for an interview with Authority Magazine to discuss why the company builds its own GPU clusters from the ground up rather than relying on existing public cloud providers.
The Limits of the Public Cloud
It’s easy to think renting server space is the fastest way to grow, but general-purpose cloud platforms aren’t always built for the specific demands of high-end gaming. Batova notes that infrastructure essentially defines the limits of a product. By architecting their own distributed cloud, Boosteroid handles the demanding processing of modern titles without the constraints often found in “one-size-fits-all” server environments.
“Public cloud platforms are excellent for general workloads,” Batova explained. “But cloud gaming is not a general workload.” Early in Boosteroid’s growth, the team faced a choice: scale quickly using GPU-as-a-Service from hyperscalers or invest in custom-built clusters. They chose the harder path. While this meant navigating hardware supply chains during global shortages and managing complex colocation power constraints, it allowed them to hit performance targets that actually matched user needs rather than just meeting a “good enough” baseline.
ASUS and AMD Provide the Hardware Advantage
Maintaining a stable gaming session requires much higher technical standards than standard enterprise data. When you’re playing a fast-paced shooter, every millisecond of delay matters. Batova highlights that Boosteroid pushed data centres into a “very different operating regime,” integrating power-hungry, high-density hardware into facilities originally designed for much lower-density enterprise use.
This isn’t just theory; we’ve seen this strategy in action with their previous hardware rollouts. Their “Ultra” tier doesn’t run on generic blades; it utilises custom-engineered ASUS server chassis packed with AMD EPYC Zen 4 CPUs and Radeon RX 7900 XT GPUs. We previously saw this tailored approach when Boosteroid deployed new GPU infrastructure in Prague. By controlling the hardware stack down to the chassis design, they optimize for heat, power, and latency.
This strategy is what allows the service to function across 29 data centres globally. Instead of pushing the burden of expensive hardware onto the player, the infrastructure handles the processing remotely. By reasoning from “first principles,” Batova and her team have turned high-performance computing into a utility that works regardless of your local setup.
Ground Truth in Global Expansion
Scaling a platform across three continents requires more than just a clever software stack. It requires intense logistical discipline. Batova’s legendary checklists exist because she respects the ‘ground truth’ of the last mile. She learned this the hard way during an international expansion when a simple lack of a truck tail lift nearly stopped the entire project.
“Your vision is only as strong as your smallest checklist item,” she noted. This focus on the “ground truth” is what separates a theoretical network from a functional one. We see this play out in their massive infrastructure projects, such as the Bielsko-Biała site in Poland. Converting a former Fiat engine factory into a data centre that starts at 82MW and aims for 1GW is a prime example of her philosophy. This massive project points to a much bigger infrastructure push than many realised.
This hands-on approach was also vital when Boosteroid expanded into Brazil. Rather than treating it as just another international location, the team studied local usage patterns and network pathways. Markets that some global companies treat as secondary become strategic powerhouses if approached with long-term commitment and a deep understanding of local infrastructure realities. As the company continues to grow, it’s clear they are now scaling across Europe, not just into it.
Boosteroid Shifts Toward a Neocloud Future
The work done to optimize gaming is now providing a foundation for new initiatives in AI workloads. This is a natural pivot, as cloud gaming acts as a high-pressure training ground for low-latency response. Backed by massive financing, Boosteroid is moving toward becoming a Neocloud provider: a specialized alternative to the Big Tech giants.
Batova points out a key shift in the industry: while most current AI environments prioritize throughput and raw data volume, she believes the next phase of infrastructure must prioritize responsiveness. This means focusing on how quickly a system reacts to a request rather than just how much data it can move at once.
“That shift requires a completely different architectural approach,” Batova said. By applying the lessons learned from keeping cloud games lag-free, Boosteroid is positioning its infrastructure to stay ahead of the next wave of technological demands. Whether it’s a 4K gaming session or a real-time AI application, the goal remains the same: technology performs consistently in real-world conditions, not only in ideal lab scenarios.
As always, remember to follow us on our social media platforms (e.g., Threads, X (Twitter), Bluesky, YouTube, and Facebook) to stay up-to-date with the latest news. This website contains affiliate links. We may receive a commission when you click on these links and make a purchase, at no extra cost to you. We are an independent site, and the opinions expressed here are our own.
