How Safaricom Scaled M-Pesa to 4,000 Transactions per Second and Is Now Eyeing 8,000
How does Safaricom’s M-Pesa handle the relentless demands of millions of daily transactions? Recent reports reveal a bold architectural transformation driven by cloud-native design, active-active redundancy, and AI-driven efficiency to ensure M-Pesa remains seamless and scalable.

The Evolution of M-Pesa’s Architecture
M-Pesa’s journey is one of constant reinvention. Felix recalls the early days of monolithic systems, designed to support only hundreds of transactions per second. Today, the platform has been re-engineered into a cloud-native, active-active system that is scalable, resilient, and capable of meeting Africa’s growing fintech demands.
What It Takes to Run a Fintech Platform at 4,000 TPS
1. An Ultra-Scalable, Cloud-Native, Active-Active Architecture
-
Cloud-native & Active-Active: M-Pesa transitioned to a 100% cloud-native, active-active system that allows simultaneous operations across multiple environments for true resilience and scalability.
-
Multiple Environments: It runs across three environments—one production and two disaster recovery. If one fails, traffic can switch over in under four minutes, and database failover happens in under a minute.
2. Massive Capacity Scaling Over Time
-
Historic Growth: In 2015, M-Pesa peaked at 616 TPS. Today, it's comfortably handling 4,000 TPS.
-
Future Targets: Safaricom plans incremental steps: 6,000 TPS by April 2026 and 8,000 TPS by year-end.
-
Daily Volume: At 4,000 TPS, that translates to roughly 100 million transactions per day.
3. Database Architecture and Infrastructure
-
High-Performance Databases: Sharded databases are now running on 64-core systems (up from 16-core), enhancing both performance and reliability.
-
Robust Infrastructure: The platform spans 700+ servers deployed across production and DR environments to ensure smooth load handling and failover.
4. Zero-Downtime Deployments
-
Frequent Seamless Upgrades: M-Pesa performs nearly two system upgrades every other week or month—without impacting service availability. Given the volume, even one minute of downtime could disrupt 240,000 customer requests.
5. AI-Driven Fraud Detection and Monitoring
-
AI-Native Core: The architecture incorporates AI at its core, enabling intelligent monitoring, fraud detection, and predictive analysis.
-
Proactive Security: AI models monitor transaction patterns and forecast system traffic, making threats easier to detect and system blowouts easier to prevent.
6. Massive Ecosystem & Partner Integrations
-
Expansive Integration Capabilities: The platform supports 45,000+ integrations, over 100,000 developers, and interfaces with more than 10 internal systems, including Fuliza and Zidisha.
-
Daraja Platform Ecosystem: As noted elsewhere, the Daraja API platform supports over 55,000 integrations and more than 100,000 developers globally.
7. Strategic Investment and Future Proofing
-
Major Capital Outlay: Safaricom is investing over $300 million annually to overhaul M-Pesa into “M-Pesa 2.0,” a platform designed for capacity, functionality, stability, and resilience.
-
Regulatory Alignment: The upgrade involves moving to a cloud-native, API-first architecture that aligns with regulatory pushes for interoperability (e.g., Kenya’s Fast Payment System)
Summary Overview
Component | Key Feature |
---|---|
Architecture | Cloud-native, active-active with multi-site failover |
Scalability & Uptime | From 616 TPS (2015) to 4,000 TPS now, aiming for 8,000 TPS in 2026 |
Infrastructure | 700+ servers, 64-core sharded DBs, seamless upgrades |
Security & Intelligence | AI-native for fraud detection, monitoring & predictive analytics |
Ecosystem & Integration | Massive partner API ecosystem with 45k+ integrations and 100k+ developers |
Investment & Strategy | $300M+ annual spend; built for resilience, growth, and regulatory alignment |
What's Your Reaction?






