Solutions
Player data private. Servers you place near the player.
CTOs and heads of platform at studios with live games and millions of accounts, who want AI for matchmaking, moderation and support without sending player data to a frontier API.
Why now
A breach now carries a nine-figure price tag.
Korea’s privacy regulator set a record
On 12 June 2026 the Personal Information Protection Commission fined Coupang 624.7 billion won, about $409 million, over a breach of 33.2 million accounts. For a studio holding player data, the cost of data leaving your control is now a published number. [1]
SourcePersonal Information Protection Commission of Korea, decision reported June 12, 2026. The full citation and link are in the footer notes.
What it means for you
Player data is personal data wherever your players sit, and the fine is now a published number. Your board will read it before the next title ships.
What waiting costs
Player accounts, payment tokens and chat logs on a shared cloud are a liability priced in headlines. One breach costs more than every server you will ever rent.
Why Triple X
Triple X puts game servers near the player and player data on a private network you control, with no exit fee when a title winds down. Launch regions by the hour; keep the data where the regulator expects it.
In plain English
Studios used to worry about latency and the bill. Now they worry about a fine bigger than the game’s revenue. Triple X gives you servers near the players and a private network for their data, priced by the hour.
What you get
Four things, each a product you can open.
Machines where the match runs
Region and zone chosen at launch; place servers near the players. Six sizes from $0.024 an hour.
OpenStandard Kubernetes for the game services
Up in 2 to 5 minutes, ingress and certificates included.
OpenYour own model on a private endpoint
Moderation, matchmaking and support on an open model you load. Prompts and player data do not leave your private cloud.
OpenSealed when it must be provable
No operator can read player data in use.
Open
Which way in
Start where it fits. Move later without a new vendor.
Triple X Cloud nodes
No hardware of your own. Join the waitlist and start when your seat opens.
Your own devices
Triple X connects your racks or devices to a private cloud of your own.
Both
Sensitive work on your devices, the rest on ours, one private cloud.
Objections you will hear
What someone in the room will say.
We are not an infrastructure company.
You are not asked to be. The platform runs the nodes; you load the model and call an endpoint. Your team sees an API, not a rack.
Open models are not as good as the frontier.
For moderation, matchmaking and support they are good enough, and you own the weights, the prompts and the data. Keep the frontier for the ten percent that needs it.
Latency kills games.
You choose the region and zone at launch and place the model where the match runs. No published region list yet; the API returns the live one.
Who can see the player data?
Nobody outside your private cloud. Sealed adds a hardware boundary so even the operator cannot read it in use.
What yes looks like
The list a buyer brings to the table.
- Their own open-weight model on a private endpoint
- A per-token cost they can compare to the API bill, measured in a pilot
- Player data that stays in their private cloud
- Servers placed near the players
- A ten-node pilot with a pass mark they set
- A path from one region to all regions without a new vendor
Get started
Point one feature at your own model.
Ten nodes, one region, 60 days.
- 01
Join the waitlist.
Tell us what you want to run and where. No credit card.
TakesTwo minutes
- 02
Run the pilot.
Ten nodes, 60 days, a pass mark you set in writing.
Takes60 days
- 03
Decide on day 60.
Extend, or leave at the transfer rate and nothing else.
WhoYou
We never ask for your root credentials, API keys or bank details by email.
The first 60 days
A pilot with a pass mark you set.
One feature, one region, ten nodes, and a latency budget the live-ops lead writes before anyone touches production traffic.
Week 1
Cluster, endpoint, one region
Create your private cloud, launch a cluster in the region nearest a player base, and load the open model on a private endpoint. The free tier covers the first machine.
Weeks 2 to 4
One feature on live traffic
Send a share of moderation or support traffic to your own model. Stop a node mid-session and time the recovery. The pass mark is the live-ops lead's: p95 latency, per-token cost beside the API bill, no player data outside your private cloud.
Day 30
Read the two bills side by side
Compare cost per million tokens on your traffic with the API invoice for the same month, with transfer included. Review the placement record.
Day 60
Decision
Next region, next feature, or stop. Leaving costs the transfer rate; the model and the data were always yours.
Questions
What buyers ask before they sign.
In the order they come up. Every answer ends with where to check it.
10 answers
All questionsCan we run match servers and game services here without a rewrite?
Yes. Six machine sizes from $0.024 an hour, and standard Kubernetes with 428 of 428 conformance tests passed. Helm charts and manifests work unchanged. A cluster reaches Running in 2 to 5 minutes; new nodes join in 1 to 2 minutes.
Proofthe Kubernetes overview in the documentation, August 1, 2026.[2]
How close to the players can we place the servers?
You choose region and zone at launch. There is no published region list; the API returns the live one before you launch. For a shard near a city, your own devices in a colocation you rent can join the same private cloud, on Tier C at index 80.
Proofthe region and zone fields on the create call in the API reference, August 1, 2026.[3]
What does a moderation model cost against the API bill?
GPU capacity is quoted in writing, hourly, naming the card class, the count and the price per GPU-hour. The pilot measures cost per million tokens on your own traffic. Triple X quotes no percentage; the number comes from your invoice.
Proofthe GPU answer on the Pricing page and a written quote for your private cloud.
What does outbound to millions of clients cost?
The first terabyte out is $0.01 per GB, 1 to 10 TB is $0.008, above 10 TB is $0.006. Inbound and traffic inside the same region are free. Set a budget with alert thresholds before launch weekend.
Proofthe transfer bands in the published pricing, August 1, 2026.[4]
Where does the player data go when the model answers a support ticket?
Nowhere. The model is an open one you load on a private endpoint in your private cloud. Prompts, player data and answers stay inside it. When a regulator wants it provable, Sealed keeps the data unreadable in use, including to the operator.
Proofthe access table on the Security page and the four steps on the Sealed page.
What happens on launch weekend when a node dies?
Work moves inside your private cloud first, then across 1,800+ nodes if you allow it. Tier B commits to 99.95%, about 22 minutes a month. Test it before launch by stopping a node. Raise account ceilings in advance; increases take 1 to 2 business days.
Proofthe fleet count, company-stated, October 2026, and the uptime table on the Reliability page.[5]
What do we show a privacy regulator who asks where 33 million accounts live?
The placement record, pinned before anything ran. The access table showing who can touch what. The SOC 2 Type II report under NDA. For Sealed runs, a receipt per run showing no one could read the data in use.
Proofthe Security page and the SOC 2 Type II report under NDA.[6]
What if we want to leave after the season ends?
No exit fee and no minimum term. Leaving costs the outbound transfer rate. Your kubeconfig and your S3 tools work on the last day as on the first, and the model weights were always yours.
Proofthe no-minimum-term rule in the published pricing and the no-exit-fee clause in the contract.[4]
How does the ten-node pilot start?
Create an account on the free tier, no credit card. Point one feature, moderation or support, at your own model: ten nodes, one region, 60 days. Write the pass mark first: a latency budget and a per-token cost you will compare with the API bill.
Proofthe free tier terms in the published pricing, August 1, 2026.[4]
What is not included?
No anti-cheat, no matchmaking service, no model weights: you bring the open model. No published region list. GPU availability is on quote. Keep the frontier API for the ten percent of calls that need it.
Proofthe AI and GPU page and a written GPU quote for your private cloud.
No questions under this term yet. Ask Triple X.