Scorpio Smart Fabric Switches

Fabric switches for the AI Era

Purpose-built portfolio for deploying rack-scale AI platforms at unprecedented pace and scale

  • Improve energy efficiency with maximum performance per watt fabric solutions
  • Increase GPU utilization with predictable, secure and high-performance data flows
  • Maximize server uptime with COSMOS delivering unprecedented data center observability
  • Reduce time to market with flexible software-defined architecture and seamless interoperability

Why Use Scorpio Smart Fabric Switches?

Maximize Utilization

  • High Performance Data Ingest

    With PCIe 6 support and AI-optimized architecture, Scorpio P-Series delivers maximum predictable performance to increase GPU utilization

  • High Performance Scale-up

    With a software-defined architecture, Scorpio X-Series delivers protocol and performance optimization for increased data transfer efficiency

  • Performance Monitoring

    Real-time telemetry data provides valuable insights to optimize performance and improve utilization

Ease-of-Use

  • Purpose-built Board Design

    Small package size with integrated power supply decoupling to enable easier routing

  • Future-proof

    Mix current and next-generation devices on the same platform

  • Seamless Software Integration

    Common COSMOS APIs enable Scorpio to integrate seamlessly into existing infrastructure management solutions that use the widely-deployed Aries Retimers

Increase Uptime

  • Fleet Management

    COSMOS software with deep telemetry and diagnostics for link and device health monitoring enables unprecedented data center observability

  • Quick Debug

    Built-in protocol analyzer with link state history and timestamps, full non-destructive eye scan for RX Lane margining, self-test features to minimize link downtime and accelerate fault isolation

  • Rigorous Interop Testing

    System testing with diverse ecosystem of PCIe hosts and endpoints in our Cloud-Scale Interop Lab minimizes interoperation risk for improved reliability

P-Series and X-Series

Application-specific products built from the ground up for AI servers and AI clusters

Architected for mixed traffic AI head node connectivity (GPU-to-CPU/NIC/SSD)

  • Industry’s first PCIe 6 fabric switch
  • Predictable and high-performance GPU-to-CPU/NIC/SSD data flows
  • High bandwidth and low latency peer-to-peer for GPU data ingest
  • Interoperability with diverse ecosystem of PCIe hosts and endpoints

Architected to deliver highest back-end bandwidth for AI scale-up (GPU-to-GPU)

  • Software-defined architecture enables flexibility for diverse AI accelerators
  • Platform-specific customization for homogeneous GPU scale-up fabrics
  • Protocol and performance optimization for increased data transfer efficiency
  • Expanded reliability and telemetry features for maximum uptime

Resources

Ordering Information

SeriesPart #PCIeLanesPackage Size [mm]Status
P-SeriesPF60641L6.x6429×29Pre-Production
P-SeriesPF50641L5.x6429×29Pre-Production
P-SeriesX6064-1M04Xami MCIO-MCIO Evaluation KittPre-Production
X-SeriesCustomer SpecificContact Us

How Your Hyperscale Data Center Can Overcome the Challenges of Next-Gen AI Infrastructure with Scorpio Smart Fabric Switches

Request the White Paper

 

Astera Labs at Supercomputing 2025: Advancing Rack-Scale AI Connectivity with Open Ecosystems

As AI models surge in complexity and require unprecedented reasoning performance, the industry is undergoing a transformative shift toward what we call “AI Infrastructure 2.0”—where the rack, not the server, becomes the new unit of compute. The scale and tight coupling required by today’s GPU-dense workloads demand ultra-low-latency, high-bandwidth communication across hundreds of… Read more

Astera Labs Completes Acquisition of aiXscale Photonics

Today, Astera Labs completed its acquisition of aiXscale Photonics GmbH, following the definitive agreement announced in October 2025. This marks an important step in our journey to deliver comprehensive connectivity solutions for AI Infrastructure 2.0.Why Photonics Matter for Scale-Up AIAs AI systems scale to hundreds of tightly integrated accelerators per rack, optical connectivity… Read more

Building the Software Stack for AI Infrastructure 2.0: Why Standards-Based Connectivity Management Matters

How our collaboration with ASPEED and Insyde on OpenBMC support advances the vision of open, interoperable AI rack infrastructure We’re at an inflection point in AI infrastructure. As I watch hyperscalers architect their next-generation AI racks, I see a fundamental shift happening—one that goes far beyond just faster GPUs and higher bandwidth connections. We’re witnessing the… Read more

Astera Labs Announces Fourth Quarter 2025 Financial Conference Participation

SAN JOSE, CA, U.S. – Nov. 13, 2025 – Astera Labs (Nasdaq: ALAB), a leader in semiconductor-based connectivity solutions for rack-scale AI infrastructure, today announced that it will participate in the following upcoming financial conferences during the fourth quarter 2025:2025 RBC Capital Markets Global Technology, Internet, Media & Telecommunications Conference on Nov. 19th… Read more

Astera Labs Announces Financial Results for the Third Quarter of Fiscal Year 2025

Record quarterly revenue of $230.6 million, up 20% QoQ and 104% YoYStrong Q3 revenue growth driven by new AI platform ramps featuring multiple product familiesScorpio fabric switch design wins expand to several platforms at multiple hyperscaler customersSAN JOSE, CA, U.S. – November 4, 2025 – Astera Labs, Inc. (Nasdaq: ALAB), a leader in semiconductor-based connectivity solutions… Read more

Astera Labs to Acquire aiXscale Photonics

Deal is expected to help enable scale-up photonic chiplets in AI Infrastructure 2.0 and accelerate Astera Labs’ deployment of rack-scale solutionsSAN JOSE, Calif.—October 22, 2025—Astera Labs, Inc. (Nasdaq: ALAB), a leader in semiconductor-based connectivity solutions for rack-scale AI infrastructure, today announced that it has entered into a definitive agreement to acquire aiXscale… Read more

Astera Labs Joins Arm Total Design to Accelerate Custom AI Infrastructure Solutions 

Collaboration addresses growing rack-scale infrastructure demands with custom AI chiplet architecture connectivity solutions SAN JOSE, Calif.–October 14, 2025–Astera Labs, Inc. (Nasdaq: ALAB), a leader in semiconductor-based connectivity solutions for rack-scale AI infrastructure, today announced it has joined Arm® Total Design, a comprehensive ecosystem dedicated to accelerating… Read more

COSMOS Explorer: Link Management Tools Demo

Astera Labs’ COSMOS Explorer is a powerful tool for link monitoring, management, and diagnostics. It offers a rich set of system configuration options for you to optimize performance with critical signal quality telemetry such as bit error rate, eye metrics, and link event logs. COSMOS Explorer allows you to modify link settings and get customized diagnostics in one GUI for all of our products.In… Read more

COSMOS and OpenBMC Demo at OCP 2025

Through collaboration with ASPEED Technology and Insyde Software, Astera Labs has extended COSMOS features into production AI systems to standardize management, monitoring and diagnostics of retimers and scale-up switches with OpenBMC and DMTF Redfish APIs. COSMOS now offers:Standard APIs for configuration, monitoring and lifecycle controlActionable insights with granular visibility into… Read more

CXL Memory Innovation at OCP 2025

Rack-scale memory innovation is unlocked with CXL. At OCP Global Summit 2025, Astera Labs demonstrated how Leo CXL Smart Memory Controllers can eliminate AI infrastructure bottlenecks with regards to memory capacity and bandwidth.3x concurrent LLM instances at higher AI throughput3x increased throughput w/ CXL at higher user count3x lower latency with CXLHigher GPU utilization on average with… Read more