With technological innovations such as Arm CSS and KleidiAI, Arm CEO Rene Haas predicts that by the end of 2025, there will be over 100 billion ARM based devices available for AI.

Prior to the COMPUTEX 2024 exhibition, Arm CEO Rene Haas shared how the company will achieve over 100 billion Arm based devices available for artificial intelligence (AI) from the cloud to the edge by the end of 2025.

 

Do we have enough energy to support AI?

One of the key questions that Haas has responded to is whether there is enough energy globally to meet the growing demand for AI computing. Many outstanding innovations have emerged in the field of AI, but the industry is facing a dilemma where the better processors can handle the performance requirements of AI, the greater their energy demand.

Arm’s DNA was born for devices that use batteries. Haas believes that it is precisely because of this that mainstream cloud service providers such as Amazon Web Services (AWS), Google, and Microsoft are increasingly adopting the Arm architecture. He also mentioned that Arm is the world’s most energy-efficient CPU, and the energy savings it saves drive amazing AI innovation.

Software is crucial

In the more than 30 year history of the company, Arm is best known for its leading energy efficiency and the foundation that makes it the most widely used computing platform globally. Not only that, Haas further pointed out that what truly sets Arm apart is its unparalleled software ecosystem.

As Haas said, Arm has become the choice for various major operating systems around the world. The AI PC recently released by Microsoft is a good example, and the most commonly used applications can now run natively on Windows on Arm (WoA). In fact, in an ecosystem with 18 million software developers, there are now more developers designing on Arm CPUs than on any other processor.

If we want to experience the social changes that AI can unleash, from the smallest devices on the edge to cloud based training and reasoning, developer support is crucial. Developers need to be able to leverage AI for innovation and enable it to run on hardware in high-performance, power optimized, highly executable, and predictable states.

To ensure this is achieved on the Arm computing platform, Haas introduced the newly launched Arm KleidiAI, which is included in a new set of AI computing libraries, making it easier for developers to run their AI applications on the Arm platform. Haas added, “In the development of Arm over the past 30 years, we have truly learned that providing developers with the resources they can access is equally important as hardware, otherwise hardware cannot play its role.”

 

For example, when developers process their models in frameworks such as TensorFlow, PyTorch, Llama 3, and MediaPipe, KleidiAI enables them to enter the abstraction layer and leverage the underlying hardware capabilities of AI workloads to fully unleash their performance potential.

From an ecosystem perspective, KleidiAI also has significant advantages. On site, Haas shared videos from Samsung, Meta, and Google executives, each discussing how Arm and KleidiAI will help accelerate AI innovation on the Arm platform across multiple markets.

Arm Terminal CSS

In addition to the excellent work done by Arm in empowering the AI developer ecosystem, Arm will continue to provide higher performance and more energy-efficient products. Chris Bergey, Senior Vice President and General Manager of the Terminal Business Unit at Arm, shared more information about the product on site and introduced the Arm Terminal Computing Subsystem (CSS) released last week.

 

This is the first time Arm has delivered physical implementations of Arm CPUs and GPUs based on cutting-edge three nanometer process nodes. Bergey emphasized its importance, as the Arm ecosystem can now utilize empirical evidence, CPU and GPU physical implementations based on leading three nanometer process nodes, and push products to the market faster and more confidently.

Arm terminal CSS enables chip manufacturers to focus on creating platform differentiation and effectively consider how the Armv9 architecture will empower their next-generation products. At the same time, by combining terminal CSS with KIeidiAI, developers will have the tools necessary to fully utilize these next-generation hardware technologies.

Finally, Haas once again emphasizes that Arm is committed to the development of AI technology by providing the world’s most complete computing platform. With technological innovations such as Arm CSS and KleidiAI, he expects to have over 100 billion devices based on Arm architecture available for AI by the end of 2025. Realizing this amazing quantity will not only reshape this industry, but also reshape the entire planet.