Leo CXL® Smart Memory Controllers

Purpose-built memory expansion, sharing, and pooling for AI and cloud platforms

Memory Solutions for the AI Era

  • Accelerates AI and cloud infrastructure with memory expansion, sharing, and pooling for enhanced performance
  • Eliminates bandwidth/capacity bottlenecks, reduces total cost of ownership, and optimizes memory utilization
  • Ensures end-to-end data integrity and protection with best-in-class industry standard security features
  • Delivers server-grade customizable RAS, advanced diagnostics, and fleet management via COSMOS suite

Leo Highlights

Accelerating AI with CXL Memory

Increase Memory
Capacity

With multiple DDR5 channels,
up to 5600 MT/s

Seamless
Interoperability Across

Stress tested with all major xPUs and Memory vendors

Enhanced Diagnostics
& Telemetry

Advanced capabilities through in-band and out-of-band management

Leo Delivers Proven Application Value with CXL

Astera Labs’ Leo CXL Smart Memory Controller is the industry’s first purpose-built solution that supports both memory expansion and memory pooling to solve performance bottlenecks and capacity constraints in cloud servers.

AI Inferencing: Recommendation System

73% more recommendations per second

AI Inferencing: Chatbot Services

40% faster time to insights with LLM

HPC: Computer Aided Engineering

50% more iterations per second

In-Memory Databases: Transaction Processing

150% more transactions per second

E-Series

  • Memory Expansion

P-Series

  • Memory Expansion
  • Memory Pooling
  • Memory Sharing

Leo A-Series Hardware Solutions

Leo A-Series CXL Smart Memory Hardware solutions offer all the benefits of Leo Controllers and enable quick plug-and-play deployment with faster time-to-market for system OEMs and data centers.

  • PCIe x16 CEM add-in card form factor
  • Up to 4x DDR5 RDIMMs supporting up to 2TB
  • On-board debug connectors for fleet management on Cloud Servers
  • Temperature and health monitoring of Leo controller and memory
  • RDIMM fault isolation and error correction
  • High volume production-qualified solutions with robust supply chain

Use Cases

Read more

Memory Disaggregation

Expand, pool and share memory between multiple servers to increase memory bandwidth and capacity while providing the option to reclaim stranded or under-utilized…

Resources

Ordering Information

Orderable Part NumberImageDocumentsDescriptionCXL SpecCXL LinkMemoryCapacityExpansionPooling / SharingOrdering
A1000-1254ABA1000-1254ABhttps://www.asteralabs.com/wp-content/uploads/2025/05/Leo_CXL_Smart_Memory_Controllers_Portfolio_Brief.pdfLeo A1000 CXL™ Smart Memory Add-in Card CXL 1.1/2.016x32G4x DDR5-5600 RDIMM slots, 2TB2TBYesYeshttps://www.asteralabs.com/product-details/a1000-1254ab/
CM5082E-*CM5082E-*https://www.asteralabs.com/wp-content/uploads/2025/05/Leo_CXL_Smart_Memory_Controllers_Portfolio_Brief.pdfLeo E-Series CXL 2.0 x8 Smart Memory ControllerCXL 1.1/2.08x32G2ch DDR5 Up To 5600MT/s2TBYesNohttps://www.asteralabs.com/product-details/cm5082e/
CM5162E-*CM5162E-*https://www.asteralabs.com/wp-content/uploads/2025/05/Leo_CXL_Smart_Memory_Controllers_Portfolio_Brief.pdfLeo E-Series CXL 2.0 x16 Smart Memory Controller CXL 1.1/2.016x32G2ch DDR5 Up To 5600MT/s2TBYesNohttps://www.asteralabs.com/product-details/cm5162e/
CM5162P-*CM5162P-*https://www.asteralabs.com/wp-content/uploads/2025/05/Leo_CXL_Smart_Memory_Controllers_Portfolio_Brief.pdfLeo P-Series CXL 2.0 x16 Smart Memory Controller CXL 1.1/2.016x32G2ch DDR5 Up To 5600MT/s2TBYesYeshttps://www.asteralabs.com/product-details/cm5162p/

Why Your Mixture-of-Experts Model Is Only as Good as Your Fabric

Introducing Hypercast™ for Improved Intelligence Benchmarks and Tokens-Per-Watt PerformanceThe more capable you make a frontier AI model, the harder it becomes to run. More parameters, more experts, more sophisticated routing: every architectural improvement that lifts benchmark scores also lifts the communication demands on the hardware connecting compute accelerators such as GPUs…. Read more

Bridge the PCIe 6 Transition Without Requalifying Your NIC Fleet

How Aries 6 PCIe Smart Gearbox enables hyperscalers to scale I/O for AI workloads while preserving qualified PCIe 5 frontend NIC infrastructureAI Infrastructure Is Outgrowing Yesterday’s I/O ArchitectureAI training and inference workloads don’t just demand faster GPUs—they demand faster systems. As model sizes and datasets explode, every critical path matters: moving data into… Read more

What NVIDIA GTC 2026 Said About the Future of AI Connectivity

NVIDIA GTC has always been a window into where the AI Infrastructure industry is heading. This year, what that window revealed was a compute layer fragmenting, deliberately and by design. Jensen Huang’s keynote introduced not one new architecture but several: Vera Rubin for high-throughput GPU compute, the Groq LPU integrated as a decode accelerator for latency-sensitive inference, and a… Read more

Astera Labs Extends Leadership in Open, AI Scale-Up Networking with New 320 Lane Scorpio X-Series Smart Fabric Switch

Now Shipping to Leading Hyperscalers, Scorpio Smart Fabric Switch Family Delivers Breakthrough Accelerator Utilization Through Memory-Semantic Based Open and Platform-Specific ProtocolsNews Highlights:Largest open, memory-semantic fabric switch: The Scorpio™ X-Series 320 Lane AI fabric switch, shipping today, supports increased scale-up cluster sizes with low latencyIntelligent… Read more

Astera Labs Reports First Quarter 2026 Financial Results

Record quarterly revenue of $308.4 million, up 14% QoQ and up 93% year-over-yearMarket-leading PCIe 6 AI fabric and signal conditioning portfolio delivered strong growth during Q1Now shipping newly announced Scorpio™ X-Series 320-lane AI Fabric switch and expanded Scorpio P-Series PCIe 6 switch family supporting 32 to 320 lanesSAN JOSE, CA, U.S. – May 5, 2026 – Astera Labs,… Read more

Astera Labs Announces Conference Call to Review First Quarter 2026 Financial Results

SAN JOSE, CA, U.S. – April 2, 2026 – Astera Labs, Inc. (Nasdaq: ALAB), a leader in semiconductor-based connectivity solutions for rack-scale AI infrastructure, today announced that it will release its financial results for the first quarter 2026 after the close of market on Tuesday, May 5, 2026. Astera Labs will host a corresponding conference call at 1:30 p.m. Pacific Time, 4:30… Read more

Astera Labs Reports Fourth Quarter and Full Year 2025 Financial Results

Record quarterly revenue of $270.6 million, up 17% QoQ, and record full-year revenue of $852.5 million, up 115% year-over-yearBroadening Scorpio X-Series smart fabric roadmap to address expanding scale-up market opportunities supporting multiple customers, starting production ramp for lead platformAppointed Desmond Lynch as Chief Financial Officer with Mike Tate transitioning to the… Read more

Scorpio X-Series 320 Lane: Largest Open, Memory-Semantic Fabric Switch

Get your first look at the new Scorpio™ X-Series 320-Lane AI Fabric Switch! The frontier models driving today’s most demanding AI applications require connectivity infrastructure that keeps pace with the accelerators powering them.High-radix AI fabric switch replaces multiple legacy switches to enable larger scale-up cluster sizes in a single hop and reduce overall latencyHardware-accelerated… Read more

Scorpio P-Series 32 to 320 Lane: Broadest Family of PCIe 6 Fabric Switches

As GPU clusters scale for frontier AI, the fabric matters. Performance, scalability, and resilience all depend on how CPUs, GPUs, NICs, storage, and memory connect—and that connection is PCIe. This is exactly what Scorpio™ P-Series Fabric Switches are built for. Introducing the expanded Scorpio P-Series PCIe fabric switch family—now spanning 32 to 320 lane configurations—designed… Read more

Extending Reach of Scale-out Networks for AI Clusters

See our latest demo where our Scorpio P-Series Fabric Switches connect NVIDIA H200 GPUs and ConnectX-8 NICs – running real NCCL-based training workloads at max data rates. Our Taurus Ethernet-enabled AECs then extend reach up to 7m with full interoperability with NVIDIA Spectrum-X for switch-to-switch and ConnectX-8 for switch-to-NIC connectivity at 800G per linkLearn how this demo delivers:✅… Read more