On the opening day of Computex 2024 in Taipei, Arm CEO Rene Haas outlined the company’s ambitious goal to enable 100+ billion Arm devices equipped for AI, spanning from the cloud to the edge, by 2025.
Do We Have Enough Energy for AI?
A critical question addressed by Haas was whether the world has sufficient energy to meet the growing compute demands of AI. He highlighted the paradox where advancements in AI processors increase energy requirements. Given Arm’s history of developing battery-efficient devices, Haas argued that this efficiency is a key reason for the widespread adoption of Arm technology by leading cloud service providers like AWS, Google, and Microsoft. According to Haas, Arm’s position as the world’s most power-efficient CPU is already driving significant AI innovations.
It’s All About the Software
Haas emphasized that while Arm is renowned for its power efficiency over its 30+ year history, what truly sets it apart is its unique software ecosystem. He noted that Arm is now the preferred platform for every major operating system, with AI PCs announced by Microsoft running apps natively on Windows on Arm. With an ecosystem of 18 million software developers, more developers design on Arm CPUs than any other processor today.
This developer support is crucial for realizing AI’s potential societal impacts, from edge devices to cloud-based training and inference. Haas introduced Arm Kleidi AI, a suite of AI compute libraries designed to simplify running AI applications on Arm hardware, ensuring performance optimization and predictability. He stated, “If there’s anything we’ve learned over Arm’s 30-plus years, as great as the hardware is, if you don’t have something that developers can get access to, the hardware is not going to do you much good.”
Kleidi AI allows developers working with frameworks like TensorFlow, PyTorch, Llama 3, and MediaPipe to leverage hardware features for optimal AI workload performance. Haas showcased videos from executives at Samsung Mobile, Meta, and Google, highlighting how Arm and Kleidi AI will accelerate AI innovation across various markets.
Arm Compute Subsystems for Client
Chris Bergey, SVP and GM for Arm’s Client Line of Business, joined Haas to discuss the newly announced Arm Compute Subsystems (CSS) for client devices. This marks the first time Arm has introduced physical implementations of its CPUs and GPUs using advanced 3 nanometer nodes. Bergey explained that this allows the Arm ecosystem to bring proven cores to market more rapidly and with greater confidence, enabling chipmakers to focus on differentiation using the Armv9 architecture.
Combined with Kleidi AI, developers now have the tools to fully exploit next-generation hardware technologies. With innovations like Arm CSS and Kleidi AI, Haas expects over 100 billion Arm devices to be AI-ready by the end of 2025, a milestone he believes will transform both the industry and the world.
