Terabit

For studios and publishers

The host you can
put your name on.

Official servers, an authorised community program, or both — on a 29-location anycast network with in-house DDoS mitigation, builds protected under NDA, and named engineers on your launch window.

See what a partner gets

Official partner since 2024

SCP: Secret Laboratory

29

Edge locations

Across five regions worldwide

500+ Tbps

Volumetric capacity

Absorbed at the edge, upstream of you

13+ Tbps

In-house stateful

Our filters, tuned by our engineers

4.2M+

Attacks absorbed

30 Tbps largest single flood

99.9%

Uptime SLA

Across the whole footprint

Two ways in

Which problem are you solving?

Studios arrive here for one of two reasons, and they are not the same conversation. Start with the one that matches yours — most partnerships end up covering both.

Official servers

What a partner gets today.

Everything below is running now, on the network the rest of the site describes. The things we would build for you are listed separately at the end — labelled, not blended in.

Build security

Your unreleased build stays yours

A partnership starts with an NDA, and the panel enforces it afterwards. Server files and any paths you nominate are hidden from the people operating the server — they get the console and the config surface you choose to expose, and nothing underneath it.

NDA before handoverRestricted file pathsIsolated environments

Patch day

Steam titles update themselves on restart

Anything distributed through SteamCMD pulls the current build when the server restarts, so a day-one patch reaches the fleet without a deploy queue or a ticket. Schedule the restart and the rollout happens; pin the version when you need a fleet held back.

SteamCMD auto-updateNo manual deploy stepVersion locking

Launch night

Named engineers, not a ticket queue

For a launch you get a direct line to the engineers who actually run the network, arranged ahead of time rather than discovered at 3am. Outside a launch window, support is staffed around the clock by people who understand game workloads instead of reading from a script.

Direct escalation channelAgreed before the window24/7 human support

DDoS mitigation

Filters written for your protocol

Filtering is always on, with no detection threshold to cross before it engages, and custom filters can be written for your game's traffic rather than generic L3/L4 rules. When a novel pattern leaks through, our own engineers tune it — that work does not sit with an upstream vendor.

500+ Tbps volumetric13+ Tbps in-house stateful0ms added latency

Footprint

29 locations behind one anycast IP

One address advertised from every live location, with each connection routed to the nearest edge automatically. Regions can be brought up ahead of a launch and scaled back as the curve settles, rather than being fixed at the shape of your busiest week.

5 regionsTier-1 transit100-400 Gbps ports

Hardware

Specified for clock speed, not core count

Server tick rate is bottlenecked on a single thread and on storage latency, so that is what the fleet is bought for — across every plan, not reserved for a premium tier that costs extra to reach.

Ryzen 9DDR5 ECCNVMe Gen4

Not shipping today — built with you

These are scoped against your release plan as part of a partnership rather than sold as finished product. If one of them is a hard requirement for day one, say so on the first call and we will be straight with you about the timeline.

  • Provisioning API — Scoped to how your title actually allocates sessions, rather than a generic REST surface you would have to work around.

  • Custom build pipeline — A deploy path for your own build artefacts, shaped around your release process and the way your QA gates a patch.

  • White-label panel — Community server owners manage their servers without leaving your ecosystem or learning someone else's product.

  • In-game rental — The storefront placement, the flow and the support path all agreed with you before anything ships to players.

Launch window

The part everyone worries about.

Nobody signs a hosting partner because of a feature list. They sign one because they believe the first weekend will hold. Here is how that gets planned rather than hoped for.

  1. 01Before

    Capacity planned against your curve

    We size concurrency against the numbers you expect and reserve headroom in the regions you are launching into, while there is still time to change either. A launch plan written the week of release is a guess.

  2. 02Rehearsal

    Load tested on the fleet you'll launch on

    Testing happens on the same hardware and the same routing that will carry release day, so what you measure is what you get — not a staging environment that flatters the real one.

  3. 03Launch

    Named engineers on the window

    An agreed escalation channel staffed for the window itself, so a problem at peak concurrency reaches someone who can act on it rather than entering a queue behind everyone else's tickets.

  4. 04After

    Scaled back as the curve settles

    Launch concurrency is never the steady state. Bringing capacity back down is part of the plan agreed up front, not a renegotiation you have to win a month later.

Community program

Your name, someone else's server.

Authorising a host is a risk decision, not a feature purchase. These are the five places it usually goes wrong — and what stops each of them.

The riskThey fall behind on your patches

Community servers on the build you shipped

Rented servers follow the same update path as everything else we host, so the fleet moves onto a patch without each owner being chased individually — and can be held on a known-good build when a release has to be rolled back.

SteamCMD auto-updateVersion locking
The riskSomeone tampers with the server binaries

Renters can't reach what you protect

You decide which files and directories a server owner can see. Because the binaries run as you shipped them, anti-cheat and server-side validation behave exactly as they do on your own infrastructure.

Restricted pathsHidden server files
The riskModded servers fragment after every update

Workshop support operators will actually use

Mods, collections and version pinning are handled in the panel rather than over SFTP — the difference between a modded community that grows and one that scatters across broken servers every patch day.

Mod managerCollectionsFile sync
The riskA server breaks your rules and it becomes your problem

Your rules, enforced as our terms

Your EULA and server rules become part of the terms every owner in the program accepts, and we act on reports from your community team — suspensions and takedowns included — through a named contact rather than the public abuse queue.

Takedowns and suspensionsNamed studio contact
The riskHosting complaints land in your support inbox

Hosting tickets stop at us

Outages, port questions, config mistakes and "my server won't start" are answered by our team around the clock. A community program should take load off your support channels, not redirect it into them.

24/7 human support

Mitigation

Questions worth asking every host.

Most providers publish a capacity number and stop. The rows that decide whether an attack takes your servers down are further along — whether filtering is always on, whether anyone will write a filter for your protocol, and whether the people fixing a leak work there or at an upstream vendor.

Terabit
OVHcloud
Nitrado
Hetzner
Edge filtering capacity
Total volumetric headroom available across the network.
500+ Tbps / 13+ Tbps50+ Tbps——
Mitigation latency
Added latency once mitigation engages.
0ms0-5msVariable0ms
Layer protection
Which protocol layers are inspected.
L3 / L4 / L7L3 / L4 / L7L3 / L4L3 / L4
Always-on mitigation
Continuous filtering that stays engaged instead of waiting for an attack to be detected.
Game-aware filters
Stateful filters tuned to specific game protocols.
Custom filters
Bespoke filters built on request for custom games, apps, or protocols.
Owned Infrastructure
Whether the provider operates its own mitigation infrastructure or relies on an upstream provider.
Attack-leak response
How quickly engineers tune filters when a novel attack slips through the edge.
Live engineer responseStandard—Standard

Competitor figures come from each provider's public marketing pages at the time of writing. More detail on the pipeline is on our DDoS protection page and the network map.

Commercials

The shape of the deal.

We do not publish rates, and it is worth saying why rather than leaving a gap: the split moves with scale and region mix, so a headline number would fit almost nobody. What does not move is the structure, which is below.

Revenue share
On community rentalsRentals under your program return a share to the studio. The split and the payout cadence are set per partnership.
Exclusivity
Not requiredWe do not ask to be your only host, and we do not ask for a category lock in return for a partnership.
Reporting
On your cadenceServer counts, regions, uptime and attack activity, delivered as often as your team actually reads it.
Data handling
NDA and DPASigned before builds change hands, with the region mix for player data agreed rather than assumed.
Escalation
Named contactsPeople on both sides who can make a decision, plus a direct channel for launch windows and incidents.
Co-marketing
OptionalLaunch coverage and creator support when it suits the release — never a condition of the deal.

Partner FAQ

The questions studios actually open with, answered including the ones where the answer is no.

One officially: SCP: Secret Laboratory, since 2024. The partner program is deliberately young and we would rather say so than imply a client list we do not have. What is not young is the operation underneath it — 15,000+ customers across 29 edge locations, and a network that has absorbed over 4.2 million attacks including a 30 Tbps flood. You would be an early partner, with the access to our engineering team that implies.

Tell us what you're shipping.

The useful first conversation is a technical one — your title, your expected concurrency, the regions that matter and the date you are working back from. Bring the constraints, not a brief.

What happens next: you send the details, an engineer replies with real questions about your architecture rather than a discovery call, and we get servers running under the arrangement before anyone signs anything long.

Join our Discord!

Get instant support and connect with other server owners

Join Now