ENGINEERING · 2026-06-17
By Alexander Strandberg · Founder, Aevs
Unity ended Multiplay Game Server Hosting in March 2026. If your game depends on Multiplay for dedicated server orchestration, you need to migrate. The good news: the alternatives are mature and the migration path is straightforward — provided you validate the new backend before cutting over production traffic.
This guide covers what happened, your main alternatives, how to evaluate them, and — critically — how to verify your replacement works before players see it.
Unity Gaming Services announced the end of Multiplay Game Server Hosting (rebranded as Unity Game Server Hosting) effective March 2026. Studios using Multiplay for fleet management, server allocation, and matchmaking integration needed to migrate. Unity recommended its own Relay and Lobby services for certain topologies, but neither covers the dedicated server hosting use case that Multiplay served — leaving studios running persistent authoritative servers needing a third-party replacement.
The shutdown follows a broader pattern of Unity's fragmented service offerings. Multiplay was acquired by Unity in 2020 alongside Vivox (voice) and ChilliConnect (backend). GameSparks (acquired from Amazon) was sunset in 2023. Studios that built on these managed services have repeatedly found themselves migrating, often under time pressure.
| Provider | Pricing model | Unity SDK | Regions | Best for | Watch out for |
|---|---|---|---|---|---|
| Hathora | Pay-per-CCU | Yes | 14+ | Indie / mid-size, cost-efficient scaling | No reserved capacity; cold-start latency on first allocation |
| Edgegap | Pay-per-minute | Yes | 550+ PoPs | Latency-sensitive games requiring edge proximity | Largest edge network; pricing can be unpredictable at scale |
| AWS GameLift | Reserved + on-demand | Yes (via C++ / C# SDK) | All AWS regions | Enterprise / AAA with existing AWS infrastructure | Highest ops overhead; best SLA guarantees |
| Nakama (self-hosted) | Infrastructure cost only | Yes (open source) | Your infrastructure | Studios with DevOps capacity wanting full control | Significant ops burden; no vendor SLA |
| AccelByte | Per-MAU | Yes | AWS-backed multi-region | Full-featured BaaS replacing multiple services | Broadest feature set; highest cost at scale |
Not all alternatives are equivalent for all game types. Evaluate on five axes:
Server allocation latency
How long does it take from "player clicked Play" to "authoritative server is ready"? For real-time games, >3 seconds is a UX failure. Test at your expected peak, not average, load.
Region coverage vs. player distribution
Where are your players? Edgegap's 550+ PoP edge network is irrelevant if your entire player base is in Southeast Asia and the closest Edgegap node is 40ms away. Match region availability to your analytics.
Unity SDK maturity
Check the SDK's GitHub issues and release cadence. A Unity SDK with 6-month-old PRs sitting unmerged is a support risk. Test the SDK against your Unity version before committing.
Fleet management UI and API
Can your DevOps team manage fleet scaling without writing provider-specific scripts? Does the provider expose a REST API for programmatic fleet control? Multiplay had a reasonably complete UI — evaluate whether the replacement matches it.
SLA and incident history
What is the provider's uptime SLA? Is there a public status page with historical uptime? A 99.5% SLA is 3.6 hours downtime per month — potentially during your peak weekend event window.
The most common migration failure mode: the new provider works correctly in isolation, but breaks under the specific interaction patterns of your game's topology. Session management, matchmaking callbacks, and reconnect logic all behave slightly differently across providers. These differences only surface when the full stack is exercised together.
The safest migration path runs a declarative validation plan against the new provider in an isolated sandbox before touching production. Write the plan first, against Multiplay, while it still works — it becomes your baseline. Then run the same plan against the replacement. If the plan passes on the replacement, the migration is behaviorally equivalent.
This plan covers the four scenarios that most commonly break in migrations: allocation success, callback latency, initial session establishment, and reconnect. It takes under 30 minutes to write and can be run on any provider with an HTTP API.
Audit your Multiplay dependency surface
Document every API call your game makes to the Multiplay SDK: server allocation, deallocation, match assignment callbacks, fleet configuration. These are the integration points your replacement must cover.
Write your validation plan before migrating
Define "working correctly" as a set of steps your game must execute successfully. Do this before touching any infrastructure — it becomes your migration success criterion.
Provision a sandbox on the replacement provider
Stand up an isolated environment running your game server stack connected to the new provider. This should not be production traffic.
Run the validation plan against the sandbox
Execute every step in the validation plan. A passing plan on the new provider means the migration is behaviorally equivalent to Multiplay for the tested scenarios.
Migrate production and run validation as smoke test
Cut over production traffic. Immediately run the validation plan against production as a post-migration smoke test. Monitor allocation latency, session error rates, and reconnect failures for 24 hours.
It depends on how deeply you used the Multiplay SDK. Standard server allocation and deallocation calls have direct equivalents in all major alternatives. Multiplay-specific features — custom fleet configuration scripts, the Multiplay QoS system, or the Unity Dashboard fleet management API — will require provider-specific replacements.
Unity Relay (P2P relay through Unity's network) and Unity Lobby are viable for games that do not require authoritative dedicated servers — typically turn-based games, casual titles, or games where latency requirements are relaxed. For real-time competitive games (FPS, MOBA, battle royale) that require authoritative server-side physics and anti-cheat, dedicated servers remain the only viable architecture. Relay is not a substitute.
Your validation plan is complete when it covers the failure modes that would be visible to players: failed matchmaking, broken reconnect, and session state loss under concurrent load. A useful heuristic: if your plan passes and a player then reports one of those failures in production, the plan has a gap. Start with allocation, session, and reconnect — add more steps when you hit production issues.