Arm has announced that its most advanced edge AI platform, Armv9, is now available to startups through its Flexible Access program, a move aimed at accelerating innovation in artificial intelligence and Internet of Things (IoT) hardware.
Arm Ltd.
The initiative allows early-stage companies to experiment with Arm’s latest technology at little or no upfront cost. Under the Flexible Access model, chip designers can explore and prototype using Arm’s intellectual property, tools, and resources before paying any licensing fees — costs are only incurred once a company decides to use specific Arm components in a final, commercial design.
Lowering Barriers to Entry for AI Hardware Innovation
Arm describes Flexible Access as a “try before you buy” model that has already proven to be a catalyst for innovation. Over the past five years, the program has enabled roughly 400 successful chip tape-outs, helping bring new products from concept to reality.
Notable participants in the program include Raspberry Pi, Hailo, and SiMa.ai, companies known for pushing the boundaries of efficient computing and machine learning at the edge.
With the inclusion of Armv9 — featuring the Arm Cortex-A320 CPU and Arm Ethos-U85 neural processing unit (NPU) — startups now gain access to a platform capable of running AI models with over one billion parameters directly on-device, without needing a cloud connection.
This makes it ideal for applications such as smart cameras, home automation systems, and interactive robots that rely on local data processing for faster, more private, and secure performance.
Bringing Intelligence Closer to Where Data Lives
According to Paul Williamson, Senior Vice President and General Manager of Arm’s IoT business, the next wave of AI breakthroughs will happen “at the edge – in the devices, interfaces, and systems that bring intelligence closer to where data is created.”
Processing AI workloads locally offers significant privacy and security benefits, as data never needs to leave the device. The Armv9 platform includes built-in protections like Pointer Authentication Code (PAC) and Memory Tagging Extension (MTE) to defend against exploits and ensure user data remains secure on-device.
In effect, devices using Armv9 can “perceive and respond like humans” — analyzing information in real time while maintaining strict privacy boundaries.
Edge AI Expected to Dominate by 2028
Industry analysts expect this local-first approach to become mainstream within just a few years. Research from VDC forecasts that AI will be the dominant technology across IoT projects by 2028, with Arm technology already playing a central role in this evolution.
Ryan Daws
For developers eager to begin building with Arm’s next-generation architecture, the Cortex-A320 will be available through Flexible Access starting November 2025, followed by the Ethos-U85 NPU in early 2026.
