Arm announced it will provide its most powerful edge AI platform, Armv9, to startups through its Flexible Access program.
The Flexible Access approach acts like a "try before you buy" option for chip designers. It offers companies upfront, low-cost or no-cost access (for qualifying startups) to a broad portfolio of Arm technology, development tools, and support materials. Participants can test hardware and software combinations, iterate designs, and pay licensing fees only for the IP that ends up in finished products.
Arm says the program has been a catalyst for innovation, supporting about 400 successful chip designs, or "tape-outs," over the past five years. Names that have taken advantage of the path include Raspberry Pi, Hailo, and SiMa.ai, showing how smaller teams can move from concept to production with reduced financial risk.
The Armv9 edge AI platform combines the energy-efficient Arm Cortex-A320 processor with the Arm Ethos-U85 NPU, the latter responsible for heavy neural processing workloads. That pairing can run models with more than one billion parameters directly on-device, removing the need for constant cloud connectivity and cutting inference latency for time-sensitive tasks.
Target applications include next-generation edge AI use cases such as smart cameras that analyze scenes rather than only record them, smart home devices that adapt to occupants’ routines, and interactive robots that respond to vision, voice, and gesture inputs. The architecture is designed to let systems carry out sophisticated perception and decision tasks close to where data is generated.
Paul Williamson, who runs the IoT business at Arm, said the next wave of AI innovation will happen “at the edge – in the devices, interfaces, and systems that bring intelligence closer to where data is created.”
Privacy and security are central selling points for on-device AI. Processing information locally lets machines perceive and respond like humans and keeps inference and data processing securely on-device. That reduces the need to send raw personal data to remote servers for routine recognition or control tasks, which lowers exposure to interception or misuse.
Arm has integrated hardware security measures such as Pointer Authentication Code (PAC) and Memory Tagging Extension (MTE) to help protect on-device execution and memory, giving designers a stronger baseline for building secure products.
Research firm VDC predicts that by 2028 AI will be the "dominant technology used across IoT projects." Arm says its technology is already at the center of that shift, and making Armv9 available through Flexible Access strengthens its role as device designers shift toward local intelligence.
Developers who want early access can obtain the Arm Cortex-A320 through the program beginning in November 2025. The Ethos-U85 AI processor is scheduled to follow in early 2026, giving startups a clear path to test, validate, and optimize edge AI systems before moving into full production.