Platform Data
Gameye
by the Numbers
Real performance data from 120 million+ game sessions across 200+ datacenters worldwide. Each figure includes the methodology behind it.
Cumulative Platform Statistics
All figures are measured at the infrastructure level, not self-reported by customers. Methodology for each metric is in the table.
| Metric | Value | Methodology |
|---|---|---|
| Total sessions orchestrated | 120,000,000+ | Cumulative count of all session API calls from January 2017 to March 2026, including sessions that failed to start. One session = one allocation request. |
| Average session allocation time | 0.5 seconds | Median time from API allocation request to container ready status, measured across 1.2M+ allocations in Q1 2026 across all 21 provider regions. P95 and P99 are higher and vary by provider and region. |
| Platform SLA (trailing 12 months) | 99.99% | Contractual SLA based on API availability. Measured as percentage of 1-minute intervals where the Gameye API returned successful responses to synthetic health checks from multiple external monitoring points. |
| Infrastructure providers | 21 | Bare-metal and cloud providers under active contract as of April 2026, including Gcore, OVHCloud, AWS, GCP, Azure, Akamai, and Tencent Cloud. |
| Datacenter locations | 200+ | Individual datacenter facilities available for session allocation across 6 continents (EU, NA, APAC, LATAM, MEA). Not all locations are available in all pricing tiers. |
| Load test capacity | 3,000,000 CCU | Peak concurrent users sustained during internal load testing of the orchestration layer. Not a production figure — the highest production peak is 250,000 CCU (Chivalry 2 launch, June 2021). |
Cost Performance
Egress elimination
Most cloud providers charge $0.09–$0.12 per GB for data transfer out. For multiplayer games, egress typically represents 40–60% of total infrastructure spend at scale — a cost that compounds directly with player growth.
Gameye includes all bandwidth in the per-vCPU/hour compute price. There is no separate egress line item, no tiered transfer pricing, and no bandwidth cap.
"Infrastructure costs came down by around 60%."
Clone Drone in the Danger Zone case study →
Current pricing as of April 2026
| Model | 1 vCPU : 2 GB RAM | 1 vCPU : 4 GB RAM |
|---|---|---|
| On-demand | $0.07/vCPU/hr | $0.10/vCPU/hr |
| Reserved | $0.02/vCPU/hr | $0.03/vCPU/hr |
| BYOI | $2/vCPU/month | $2/vCPU/month |
Per-second billing. No minimum session length. No egress fees. Full pricing details →
Launch Performance
Production launch data from customer titles. These are real events, not load tests. Figures are sourced from post-launch reviews with each studio.
Chivalry 2 — Torn Banner Studios / Tripwire Interactive
Unreal Engine 4. Launched June 2021 across PC, PlayStation, and Xbox simultaneously.
| Concurrent players at launch | 250,000 |
| Players in first month | ~1,000,000 |
| Server locations at launch | 30 |
| Major downtime events | 0 |
| Demand vs. best-case forecast | ~2× higher |
Demand hit roughly double the best-case projection. When unexpected player volume appeared from Japan and Korea, new capacity in Tokyo and Seoul was provisioned within hours. Full case study →
Nubs! Arena — Rangatang / Glowfish Interactive
Session-based arena game. Launch spike absorbed without manual intervention.
| Demand vs. forecast at launch | 300% above projection |
| Session availability during peak | Maintained |
| Post-launch updates shipped during peak | Yes, continuously |
The orchestration layer absorbed a 3× spike without requiring manual capacity adjustments. Full case study →
Clone Drone in the Danger Zone — Doborog Games
Live service multiplayer game. Migrated from existing infrastructure to Gameye without downtime.
| Infrastructure cost reduction | ~60% |
| Migration downtime | None (live migration) |
| Impact on multiplayer behaviour | None (deterministic behaviour preserved) |
Infrastructure costs dropped ~60% after migration. Active sessions were not disrupted during cutover. Full case study →
How Scaling Works
What happens between a matchmaker call and players connecting to a dedicated server.
Your matchmaker or backend calls the Gameye Session API when a match is found. The request includes your image name, target region, and any launch arguments.
The orchestration layer selects the optimal datacenter based on player geography and available capacity, then starts your containerised server. The server IP and port are returned to your matchmaker.
When utilisation thresholds are reached, Gameye provisions additional capacity from the provider pool automatically. Cloud bursting to AWS, GCP, Azure, Akamai, and Tencent Cloud is available in every region. No manual intervention required.
If a provider experiences issues, sessions are automatically routed to alternate providers in the same region. No single provider failure can take down a Gameye region.
How We Measure
Every figure on this page has a specific measurement definition. If you're citing Gameye data, cite the methodology below.
Every API call to the session API increments the session counter, whether the session started successfully or not. One session = one allocation request. The counter has been running since Gameye's first production deployment in January 2017.
Measured server-side as the duration between the API receiving an allocation request and the container reporting ready status. We report the median (p50) across all allocations in the measurement period. P95 and P99 are higher and vary by provider and region. The 0.5s figure covers 1.2M+ allocations in Q1 2026.
API uptime is measured via synthetic health checks at 1-minute intervals from multiple external monitoring points. A minute is counted as "down" only if all monitoring points report failure simultaneously. The 99.99% figure represents trailing 12-month availability.
Based on published pricing from AWS, GCP, and Azure as of Q1 2026, applied to typical multiplayer traffic profiles (64-tick servers, 10-player average sessions, 60-minute average session length). Actual savings vary by game architecture and traffic patterns. The 60% figure is a specific customer report from Doborog Games.
Citing this data
When referencing Gameye platform statistics, link to this page and the methodology section so readers can verify the source.
Gameye has orchestrated over 120 million game sessions since 2017 (platform data). 0.5s median session allocation, measured across 1.2M+ allocations in Q1 2026 (methodology). This page is updated quarterly: January, April, July, October. Last updated April 4, 2026. Spot an error? Contact us.