At COMPUTEX 2025, Arm made it clear: the company is fully committed to delivering a top-tier hardware and software stack for AI, whether it’s deployed in the cloud or at the edge.
A major message from Senior VP Chris Bergey’s keynote was that AI is no longer confined to the cloud. It’s rapidly moving closer to users – into smartphones, laptops, and vehicles. This rise of “physical AI” marks a shift toward systems that can sense, interact with, and respond to the real world.
That’s where Arm’s long-standing focus on power-efficient performance stands out, drawing on over 40 years of expertise in efficiency-driven design.
He also emphasized Taiwan’s critical role in the global AI ecosystem. Known for its strength in design, engineering, and manufacturing, Taiwan is a cornerstone of the AI supply chain. But Arm believes the next step is to scale that leadership – and doing so will require deeper innovation, greater collaboration, and bold vision.
At the core of Arm’s AI strategy are three pillars: a ubiquitous computing platform from cloud to edge, robust support for the largest developer ecosystem, and industry-leading performance per watt. Arm isn’t starting from scratch – more than 310 billion Arm-based chips have already been shipped.
One notable advancement is KleidiAI, an AI-optimized library suite launched last year. It’s already integrated with major frameworks such as ONNX Runtime, PyTorch, MediaPipe, and ExecuTorch, and supports leading Chinese AI models like Qwen and Hunyuan. The objective is clear – to make AI run better on Arm CPUs across a wide range of use cases and devices.
While training AI models often grabs the spotlight, Bergey highlighted that the real value to users lies in inference. Whether on a smartphone, in a car, or at the edge, inference is where efficiency becomes critical – and where Arm’s technology shines.
Cloud computing is also being reimagined. Today, Arm-based CPUs power over half of AWS’s new compute capacity, driven by the efficiency of Graviton chips. Other cloud providers are following suit, reporting energy savings of at least 40%. With datacenter power demand in Taiwan projected to rise eightfold by 2028, energy-efficient computing is quickly becoming a necessity.
Arm’s Neoverse CPUs are central to NVIDIA’s Grace Blackwell platform, which unifies CPUs, GPUs, and DPUs for AI-centric workloads. This level of system-wide optimization is essential, Bergey noted, because legacy architectures weren’t designed with AI in mind.
Most smartphones today already run on Arm, but they are not having it as they want as much as possible with expansion and venture into new form factors like PCs, Chromebooks, and ultra-light laptops. For example, MediaTek’s Kompanio Ultra chip, built for thin-and-light laptops, is one such example aimed at bringing AI to students, creatives, and everyday users.
Looking ahead, Arm is already building the next generation. Its upcoming flagship CPU, codenamed Travis, will feature Scalable Matrix Extensions (SME) for significant AI performance improvements and double-digit gains in IPC. On the GPU side, Drage – the next-gen Mali – is expected to deliver console-quality graphics and sustained performance for rich media and mobile gaming.
Like our Facebook Page here at NasiLemakTech.com for more news and in-depth reviews! Also, join our Facebook Group for insightful information and memes!
Subscribe to our YouTube channel too!
