Breadth, depth and value: Arm empowers developers for the agentic AI era 

Great hardware matters, but software is what helps unlock its full value. Arm’s DNA is hardware, and our sustained software investments help make that hardware easier to use, optimize, and scale. Since our inception, Arm has invested across the software stack, from open source and developer tooling to partner enablement deep inside production codebases, forming the foundation for increasingly complex AI workloads. Together, these investments have helped build a global ecosystem of more than 22 million developers on Arm. 

As compute scales across cloud, edge, and the physical world, software demands are becoming more complex. The new landscape for AI development is creating systems that can reason, act, and adapt across a growing range of requirements and deployment targets.  

Arm’s vision is to enable AI for everyone, everywhere. We are obsessed in making it as simple as possible for our developers when moving from prototype to production and portability regardless of the complexity of their overall system.  

In this landscape, developer success depends on a platform that provides flexible options. This means combining breadth across the entire compute spectrum, depth through the software stack, and the ability to translate both into real value for developers and their applications.  

This is the advantage Arm provides through our platform and software ecosystem: hardware designed for efficient compute, software investment that unlocks that hardware, and partner enablement that helps developers build, optimize, and scale wherever compute happens. This provides a consistent foundation for AI-driven solutions, wherever they run, from cloud to edge to emerging physical AI environments.  

Breadth to run everywhere

The ubiquity of the Arm compute platform helps developers build once and scale across vastly different environments. This reach matters because today’s most demanding workloads increasingly span the full spectrum of compute, from smart sensors and ultra-low power edge devices to hyperscale cloud environments, including AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure.  

Across the landscape, the challenge remains the same: build applications that solve real-world problems while meeting business requirements for performance, cost and power efficiency. Arm’s software ecosystem brings together hyperscalers, commercial software companies, industry leaders, AI pioneers, open-source communities and edge innovators to help developers optimize wherever compute happens.  

Depth for full stack optimization 

Arm’s full-stack optimization is the result of many years of sustained software investment, including a focus on open, extensible software that developers can use, modify, and deploy consistently across environments. This includes contributions to more than 1,300 open-source projects and collaborations with more than 50,000 companies.  

Given the scale of software that Arm supports — from cloud to edge — our approach is guided by three core priorities that shape how we invest across the stack and where we focus optimizations for the greatest impact: 

• Maximizing performance per watt;  

• Making it as easy as possible to build on Arm; and  

• Enabling developers to deliver innovation to market as fast as possible. 

Maximizing performance per watt: AI workloads demand sustained, efficient performance, not just peak capability. Arm takes a strategic approach to optimization by focusing on durable layers of the stack — compilers, operating systems, runtimes, core libraries, and AI frameworks — where improvements persist and scale across the ecosystem. This ensures that performance gains are not isolated to individual applications, but realized broadly, enabling workloads to run more efficiently without requiring constant rework at the application layer. 

Making it as easy as possible to build on Arm: Arm’s long-standing investment in open, extensible software makes it straightforward for developers to build, deploy, and scale across environments. Our work spans many of the industry’s most widely used foundations, including Linux, Android tooling, AI frameworks like PyTorch, and runtimes such as LiteRT and ExecuTorch. Arm Kleidi directly integrates with frameworks and runtimes developers already use, working behind the scenes so developers can benefit from optimized performance using the tools they already know. 

Enabling developers to deliver innovation faster: By optimizing the layers that matter most, Arm reduces the need for developers to rewrite or extensively modify their applications. In many cases, workloads simply run better on Arm — with improvements delivered through tuned libraries, kernel enhancements, and framework-level optimizations. This allows developers to focus on innovation rather than infrastructure, accelerating time to market while maintaining performance and efficiency at scale. 

That is the Arm advantage: consistent architecture, durable software investment, and optimized performance without unnecessary engineering effort. 

Value through real-world gains that scale

Arm’s ongoing partnership with Meta is a clear example of how software collaboration and silicon innovation reinforce each other. That collaboration now extends directly to the recent launch of the Arm AGI CPU, where Meta is the lead partner and customer for Arm-designed production silicon built for AI data centers and agentic AI infrastructure. Arm AGI CPU launch partners, such as Cerebras, Cloudflare, OpenAI, Positron, Rebellions, SAP, and SK Telecom, show how this work extends across cloud, networking, enterprise, and agentic AI infrastructure. Together, these collaborations show how Arm-based infrastructure and software optimization can help partners scale agentic workloads in production.  

We’ve worked closely with Meta for years across both hardware and software, helping them get more value from their unified, fast-moving development model.  

Beyond the data center, where Meta trains models at a massive scale, the company also needs to run AI on the devices people use every day. To do this, it is moving more AI inference on-device across its family of apps with ExecuTorch.  

In Instagram Cutouts, for example, ExecuTorch can leverage Scalable Matrix Extension 2 (SME2), an Armv9 architecture feature, to accelerate SqueezeSAM segmentation, enabling faster and more efficient mobile AI experiences. Arm is also bringing AI acceleration closer to the GPU pipeline with Arm Neural Technology, so developers can create richer, more responsive visual experiences within mobile power and performance constraints.  

Arm worked with Meta to optimize AI runtimes and deployment tools, including ExecuTorch, improving AI model deployment on Arm-based platforms. This helps deliver richer, more responsive experiences in the apps that billions of people rely on every day, while also extending those gains across the broader open-source ecosystem.  

By combining breadth and depth, Arm enables significant leverage for developers: the ability to optimize once and realize benefits across a wide range of use cases and environments. This is what value looks like in practice: optimizations at the right layer carry across tools, devices, and AI-native data center infrastructure, so developers spend less time on repetitive deployment work and more time building the applications that matter. 

Empowering developers in a new era of AI 

Arm has spent decades creating hardware and software platforms that help developers build across generations of compute. Through continued investment in deep and broad technology ecosystems, Arm lowers the barriers to advanced AI and empowers developers to build business critical solutions.  

As compute demands grow from cloud to edge, and agentic AI accelerates, Arm is well positioned to help partners, customers, and developers capture the full value of this next wave of innovation.  

Arm continues to invest in the software, tools, and ecosystem partnerships that give developers a faster path from prototype to production across cloud, edge, gaming, mobile, and robotics. Arm Create extends this commitment by giving developers access to hands-on projects, live events, and practical challenges that help turn ideas into working applications. Built around the maker mindset, the program supports developers as they imagine, prototype, optimize, and deploy AI applications across Arm-based platforms, from cloud to edge.