AutoMQ — The Kafka Engine Behind
the World's Biggest Game Studios
Building a Multi-Cloud Real-Time Streaming Analytics Architecture on AutoMQ
Hungry Studio, based in Hong Kong, created Block Blast! — a drag-and-drop puzzle game that became a global phenomenon. #1 in 40+ markets, 200+ countries, 70M DAU— the world's largest ad-monetized mobile game. Winner of Sensor Tower APAC 2025 Best Puzzle Game and App Ape Award 2025.
Kafka Pain Points
Multi-cloud data silos block unified analytics
Business runs on both OCI and AWS. Ad spend, user behavior, and operational events are scattered across clouds with no unified analytics layer.
Kafka multi-cloud data latency impacts ad ROI
Advertising is the primary revenue driver. Operations teams need real-time visibility into campaign performance. Cross-cloud data fragmentation creates long analysis pipelines and high latency, directly impacting ad revenue.
Telemetry archival costs are prohibitive at scale
70M DAU generates massive daily telemetry needed for long-term retention. Traditional Kafka's local-disk storage makes this expensive with ongoing capacity planning overhead.
Why AutoMQ
Cloud-neutral, cloud-native multi-cloud solution
AutoMQ provides a cloud-neutral, cloud-native architecture supporting AWS, Azure, GCP, OCI and all major cloud regions globally. Managed Connector and Kafka Linking enable seamless multi-cloud data migration and synchronization — no vendor lock-in.
Fully managed Connector, out of the box
Provides fully managed Connector support for MySQL, PostgreSQL, Databricks and other mainstream data sources. Hungry Studio uses S3 Connector for low-cost telemetry archival.
Cloud-native architecture delivers additional savings
S3 replaces multi-replica EBS, zero cross-AZ fees, on-demand broker scaling. Significant savings at 70M DAU data volumes.
Real Production Numbers
us-east-2Partial production cluster data as of Q1 2026
Eliminating Kafka Over-Provisioning with True On-Demand Scaling
miHoYo (HoYoverse), founded in 2011, is one of the world's most successful game studios. Genshin Impact has 300M+ registered players, alongside Honkai: Star Rail and Zenless Zone Zero. 5,000+ employees across Shanghai, Singapore, Montréal, LA, Tokyo, and Seoul. ~$4.7B revenue in 2024.
Kafka Pain Points
Traffic spikes, but clusters can't scale back
New game versions cause multi-fold traffic surges. Traditional Kafka's local-disk dependency makes scale-down nearly impossible — data migration is slow, risky, and bandwidth-intensive.
Chronic over-provisioning wastes resources
Clusters sized for peak stay oversized during normal periods. Across multiple games and regions, idle compute and storage costs compound rapidly.
High migration risk blocks Kafka architecture upgrades
miHoYo runs massive Flink workloads consuming Kafka data in real time. If consumer offsets are lost or disrupted during migration, Flink jobs must reprocess from scratch — extremely costly. A reliable, out-of-the-box migration tool is needed to safely upgrade to a new Kafka architecture.
Why AutoMQ
Scale up for launches, scale back after
AutoMQ persists data to S3. Scaling involves only metadata changes, completing in seconds with zero data migration. Over-provisioning eliminated at the architecture level, dramatically reducing operational complexity.
Unified S3 storage, significantly lower costs
All data persists directly to S3, reducing storage costs by an order of magnitude. No more disk capacity planning or expansion headaches.
Zero-downtime migration to AutoMQ, seamless Flink handover
Kafka Linking migrates from existing Kafka to AutoMQ with zero downtime. Consumer offsets are preserved — Flink jobs switch over and continue consuming without reset.
Real Production Numbers
Partial production cluster data as of Q1 2026
Diskless Kafka Eliminates AWS Cross-AZ Traffic Costs for Real Savings
FunPlus, founded in 2010 and headquartered in Switzerland, has 2,000+ employees across China, Singapore, Portugal, and Spain. State of Survival has 150M+ downloads. Other titles include King of Avalon, Guns of Glory, Sea of Conquest, and DC: Dark Legion. A leading global SLG publisher celebrating 15 years in 2025.
Kafka Pain Points
AWS MSK costs remain stubbornly high
Growing data volumes drove MSK bills higher. Migrating to MSK Express didn't help — both rely on traditional Kafka's multi-replica replication, the fundamental cost driver.
Cross-AZ traffic is the biggest hidden cost
Traditional Kafka's multi-replica mechanism means every message crosses AZ boundaries multiple times. For data-intensive gaming workloads, cross-AZ fees can dominate total Kafka cost.
Traditional Kafka cost architecture can't support business growth
Traditional Kafka architecture introduces cross-AZ traffic fees, local storage costs, and management complexity that cannot support large-scale business growth.
Why AutoMQ
Cross-AZ traffic fees eliminated architecturally
AutoMQ persists data directly to S3, eliminating multi-replica replication entirely. No inter-broker cross-AZ copying. The key to real cost reduction on AWS.
Storage costs drop by 10x
Traditional Kafka uses 3x EBS replicas plus reserved headroom. AutoMQ stores single-copy data on S3, pay-as-you-go.
Kafka cost decouples from growth
Zero cross-AZ fees + S3 storage + stateless on-demand brokers. Not incremental optimization — architectural transformation.
Real Production Numbers
Partial production cluster data as of Q1 2026
Replacing AWS MSK to Power 450M+ Monthly Real-Money Tournaments
Avia Games, based in San Mateo, CA (founded 2015), runs Pocket7Games — a platform of 15+ skill-based games including Bingo Clash (App Store Top 5) and Solitaire Clash. Players compete for real cash prizes. 60M+ downloads, 450M+ monthly tournaments, $714M+ in cash prizes paid out.
Kafka Pain Points
MSK's unpredictable maintenance disruptions
AWS MSK forces OS patch updates that restart brokers at unpredictable times. Each restart triggers partition rebalancing, causing latency spikes during business peak hours.
Scaling is too risky for production
Every broker holds its own data replicas. Any change requires massive data migration that's slow, I/O-intensive, and risky. Engineering teams avoid touching production clusters.
Fragmented game telemetry management
15+ games generate massive telemetry. Schemas are scattered across consumers and Flink jobs with no single source of truth. Each new topic requires hand-written Flink DDL.
Why AutoMQ
Maintenance no longer impacts live games
AutoMQ offloads data to S3, making brokers stateless. Restarts, upgrades, and scaling involve only metadata changes — no data migration, no rebalancing storms.
Scaling becomes routine, not high-risk
Storage-compute separation enables second-level scaling with zero data copying. Teams can now freely adjust cluster size and pursue optimizations previously shelved due to risk.
Table Topic: unified streaming data platform
AutoMQ's Table Topic auto-materializes Kafka topics as queryable Iceberg tables — no ETL pipelines needed. End-to-end automation from telemetry collection to analytics.
Real Production Numbers
Partial production cluster data as of Q1 2026
Why AutoMQ for Gaming
Instant Elasticity for Gaming Traffic Patterns
Game launches and events create massive traffic spikes that quickly subside. AutoMQ scales in seconds, on demand — no over-provisioning for peak capacity.
Technology Innovation Driving Cost Reduction at Scale
Data persisted to S3 cuts storage costs by 10x. Cross-AZ replication traffic fees eliminated. Stateless brokers scale on demand — zero resource reservation waste.
Multi-Cloud + Enterprise-Grade, Out of the Box
Zero-downtime Kafka Linking migration, Iceberg real-time data lake (Table Topic), Managed Kafka Connector, consistent multi-cloud deployment, AutoMQ BYOC fully managed.
Available on Cloud Marketplaces
Subscribe to AutoMQ directly from your preferred cloud platform
