Arm Holdings plc Unclaimed

Transforming AI Infrastructure With Scalable, High-Performance Computing.

Enhances AI performance for smartphones with unmatched speed and personalization.

Accelerates AI workloads with modular, standards-based compute subsystem solutions.

Accelerate innovation in edge AI with low-cost access to Arm's cutting-edge technology.

High-density, power- and cost-efficient CPUs and production silicon optimized for large-scale AI inference, orchestration and rack-scale deployments in cloud and data centers.

System-level SoC platforms combining SME2-enabled Arm CPUs, Mali GPU and integrated AI libraries to deliver real-time on-device AI for assistants, translation, speech and audio with low latency and efficient power.

Licensing and developer programs that provide low-cost access to Arm IP, CPUs and NPUs for startups and OEMs to prototype, tape-out and deploy edge, embedded and IoT devices running on-device AI models.

Standards-based compute subsystem IP that defines chiplet boundaries, coherent interconnects and industry-standard interfaces (AMBA CHI C2C, UCIe, PCIe, SerDes) to enable modular, secure and scalable multi-die systems.

Arm AGI CPU

Arm AGI CPU is Arm-designed data center silicon built to run agentic AI workloads at scale. It expands Arm’s compute platform into production silicon, delivering high-density, power-efficient compute for AI inference and orchestration. The CPU can feature up to 136 Arm Neoverse V3 cores per CPU, with 6 GB/s memory bandwidth per core and a 300 W thermal design power. This enables dense rack configurations—from up to 8,160 cores per rack to more than 45,000 cores per rack in liquid-cooled systems. Arm claims more than 2x performance per rack versus x86 platforms and up to $10B in CAPEX savings per gigawatt of AI data center capacity. Meta serves as the lead partner and co-developer, aligning roadmaps for production-scale deployment. Arm provides production silicon and pre-integrated Compute Subsystems (CSS) to accelerate development, and supports a broad software ecosystem. The Arm AGI CPU marks a major expansion of Arm’s compute platform into silicon, enabling scalable AI infrastructure from cloud to data centers.

• Delivers More Than 2x Performance
  Arm AGI CPU significantly enhances AI workload performance, providing over twice the performance per rack compared to traditional x86 processors, thus ensuring efficient processing for agentic AI tasks.
• Enables Production-Scale Deployment
  With strong backing from Meta and numerous partners, the Arm AGI CPU is designed for quick integration and deployment across various AI data center environments.
• Accelerates Development With CSS
  The Arm AGI CPU includes integrated Arm Compute Subsystems that streamline system design, helping accelerate the development of AI infrastructure.
• Supports Up to 136 Cores
  Offering a maximum of 136 powerful Neoverse V3 cores per CPU, the Arm AGI CPU provides substantial memory bandwidth, crucial for handling complex AI tasks efficiently.
• Enables High-Density Rack Configurations
  The Arm AGI CPU can scale dramatically, accommodating more than 8,160 cores in standard air-cooled systems, and further extending to over 45,000 cores in optimized liquid-cooled deployments.
• Maximizes Energy Efficiency
  Designed for efficient energy use, the Arm AGI CPU facilitates high-density deployments without compromising on power efficiency, allowing for sustainable AI computing.

Arm Lumex CSS Platform

Lumex CSS is Arm’s most advanced compute subsystem platform, purpose-built to accelerate AI experiences on flagship smartphones and future PCs. It unites SME2-enabled Arm CPUs with a Mali G1-Ultra GPU, DSU, and system IP, enabling real-time on-device AI from assistants to translation and personalization. Lumex includes tiered CPU configurations (C1-Ultra, C1-Premium, C1-Pro, C1-Nano) and software integration via KleidiAI libraries. SME2-enabled CPUs deliver up to 5x AI performance uplift, 4.7x lower latency for speech workloads, and 2.8x faster audio generation. The platform supports 3nm-optimized implementations and on-device AI workloads across a range of devices while enabling flexible deployment with efficient power use.

• Achieves 5x AI Speed
  Leveraging SME2, Lumex CPUs provide significant performance gains in AI tasks.
• Ensures Power Efficiency
  Offers cutting-edge technology that strikes a balance between performance and power consumption.
• Integrates AI Flexibly
  Allows developers to swiftly implement AI capabilities without extensive rewrites.
• Elevates Graphics
  Ignites the visual experience with enhanced graphics capabilities for gaming and media.
• Delivers Speedy Audio Processing
  Accomplishes near-instantaneous audio output using advanced processing techniques.
• Reduces Response Time
  Ensures faster and smoother interaction during voice-based applications.
• Offers Flexible Configurations
  Provides a range of CPUs tailored to different performance needs and market segments.

Arm Compute Subsystems (CSS)

Compute Subsystems (CSS) are Arm’s platform of pre-integrated, standards-based system blocks that combine Arm IP and interconnects to accelerate AI workloads across data centers and edge devices. CSS provides a cohesive path from IP licensing to production silicon and software, enabling partners to deploy Arm-based AI solutions with coherent interfaces. CSS emphasizes modularity and interoperability with industry standards, including chiplet expansion via AMBA CHI C2C, UCIe, PCIe, and SerDes. It defines chiplet boundaries and responsibilities for system partitioning, and addresses security, synchronization, interrupt and debug access, and timing to ensure data integrity across chiplets.

• Enhances Chiplet Integration
  Facilitates a unified approach to designing multi-vendor chiplet systems, promoting efficiency and compatibility across various architectures.
• Ensures Data Consistency
  Supports seamless data flow between components for optimal performance in AI applications.
• Enhances System Reliability
  Provides necessary tools for monitoring and debugging, ensuring dependable system operations.
• Guarantees Data Accuracy
  Ensures that data integrity is maintained through synchronized operation across diverse chiplet components.
• Facilitates Integrative Solutions
  Empowers designers with the flexibility to integrate various industry-standard interconnects, enhancing system performance.

Arm Flexible Access

Arm Flexible Access is Arm’s program to give up-front, low-cost or no-cost access to Arm technology, tools, and resources, enabling startups and OEMs to experiment and iterate designs. Qualifying startups pay license fees only for the technology used in final designs. The program extends to Armv9 edge AI platform access, featuring a Cortex-A320 CPU and Ethos-U85 NPU to enable on-device AI models with more than one billion parameters. Arm Flexible Access has supported rapid ecosystem innovation with more than 400 tape-outs and 300 active members across embedded and IoT.

• Enable On-Device AI
  Use advanced Armv9 architecture to run complex AI models directly on edge devices, enhancing responsiveness and efficiency.
• Foster Collaboration
  Join a vibrant community of innovators, sharing knowledge, resources, and best practices to drive rapid advancements in edge AI technologies.
• Lower Barriers to Entry
  Empower startups to innovate without heavy upfront costs, paying only for the technology used in their final products.
• Enhance IoT Solutions
  Utilize Arm Flexible Access to develop and refine IoT devices that require high performance and efficiency for real-time data processing.