Serverless Game Server

A serverless game server is a dedicated server session that spins up on demand when players need it and is terminated when the session ends — you pay only for active play time, with no idle servers sitting unused.

The term borrows from cloud computing, where “serverless” means you write code and the platform handles provisioning — you never think about the underlying machine. Applied to game servers, it means a studio calls an API when players are matched, receives an IP and port, and the platform handles everything else: which machine runs the session, in which region, using which hardware. When the match ends, the session is torn down and you stop being charged.

What it solves

Traditional game server hosting required studios to reserve capacity in advance. You’d rent a block of servers for a month, hope your player estimates were accurate, and pay for every idle hour when servers sat empty between matches. Underestimate and you turn players away; overestimate and you burn budget on machines doing nothing.

Serverless game server hosting removes that bet. The platform maintains a global pool of capacity that many studios share. When your players need sessions, capacity is drawn from the pool. When they don’t, you’re not holding any — someone else’s players are using it. Your cost tracks your actual playtime rather than your worst-case estimate.

”Serverless” doesn’t mean there are no servers

This is a common point of confusion. Serverless game hosting still runs your game on a real dedicated server — a physical or virtual machine with CPU, RAM, and network. The “serverless” part means you don’t manage or reserve those machines. The platform does. Your game server binary still needs to be compiled, containerised, and pushed to a registry. Dedicated, authoritative server architecture still applies. The difference is purely in how infrastructure is provisioned and billed.

The cold-start tradeoff

Pure on-demand provisioning has a problem: starting a new container from scratch takes 20–60 seconds. Players grouped by a matchmaker can’t wait that long. Good serverless game server platforms solve this by maintaining a warm buffer — a small pool of pre-started sessions ready to receive players immediately — while still billing only from the moment a session is allocated, not from when the buffer was populated.

This warm buffer is the engineering challenge that distinguishes capable orchestration platforms from simple container-on-demand services. Gameye’s Smarthost feature is designed specifically to predict demand and maintain the right buffer size — enough to eliminate cold starts, not so large that idle sessions inflate your bill.

How billing works in practice

Serverless game server billing is typically calculated per second or per minute of active session time. A 20-minute match on a 2-vCPU, 4GB instance at $0.05/hour costs roughly $0.017. Scale that to 10,000 concurrent sessions and the numbers are significant — but you’re paying for 10,000 active matches, not 10,000 reserved machines running 24 hours a day regardless of demand.

Back to Glossary