Generative AI and Large Language Models (LLMs) are getting a lot of attention, but what’s often unknown is how much AI is already deployed in embedded devices and how it’s impacting applications across homes, cities, and industries—what we call the Intelligent Internet of Things (AIoT)—and it’s built on Arm.

Arm Holdings today announced the Arm Cortex-M52, designed for AIoT applications that demand increased performance without the cost of dedicated digital signal processing (DSP) and machine learning (ML) accelerators. The Cortex-M52 is said to offer the potential to deliver ML in embedded computing solutions at a lower cost than is currently available.

“As AIoT technologies evolve, on-device intelligence is being deployed at the lowest cost to smaller, cost-sensitive and often battery-powered devices, while also helping to improve privacy and trustworthiness by reducing device reliance on the cloud,” said Paul Williamson, senior vice president and general manager, IoT Business Line, Arm. “As shipments of AI-enabled IoT continue to grow, our partners will need access to more ML capabilities and a simpler development flow that will give them the agility to rapidly innovate and scale.”

■ Providing AI capabilities to a wide range of IoT devices

The Cortex-M52 includes Arm Helium technology, which delivers significant performance improvements in DSP and ML applications for small, low-power embedded devices. It enables compute-intensive ML inference algorithms to be deployed on endpoints without the need for a dedicated NPU. Helium technology is already implemented in products at the network edge, but the Cortex-M52 enables Arm’s partners to bring this capability to low-cost, power-constrained devices.

Arm is extending its Helium technology with a new Cortex-M class of microcontrollers, dramatically improving matrix and DSP computing in small embedded devices. The Cortex-M52 provides a simplified migration path from Cortex-M33 and Cortex-M4 to address a wide range of AIoT applications supporting richer UI, voice and vision experiences, such as automotive and industrial control, predictive maintenance, and wearable sensor fusion. The Cortex-M52 provides the flexibility needed to scale across a range of performance points and configurations, while saving silicon area and cost by providing DSP functionality without a separate processing unit.

AIoT runs on Arm, and Arm, together with our partners, has identified a need to bring DSP and ML compute performance to low-power embedded applications at a more affordable cost and access point for the market. The Cortex-M52 CPU enables the industry to further scale IoT device deployments by delivering higher levels of AI inference performance in ultra-small devices.

■ Providing optimal performance and cost selection along with security

The Cortex-M52 extends the Armv8.1-M Cortex-M lineup (including the Cortex-M55 and Cortex-M85) to a new efficiency point, a significant milestone in bringing ML capabilities to microcontrollers.

It offers the lowest area and power of any Helium-enabled Cortex-M family, providing a choice for the semiconductor industry struggling between performance and cost. The Cortex-M52 delivers up to 5.6x faster ML performance and up to 2.7x faster DSP performance over the previous generation, allowing developers to benefit from improved ML and DSP performance.

Security remains critical, especially in devices shipping in high volumes, and Cortex-M52 implements the latest security extensions for Armv8.1-M, including PACBTI and Arm TrustZone technology, which provide enhanced software threat mitigation capabilities.

Additionally, Cortex-M52 accelerates the process of implementing PSA Certified Level 2 silicon, enabling the next generation of PSA Certified devices. The latest Armv8.1-M cores (including Cortex-M55 and Cortex-M85) provide functional safety features for many automotive and industrial control applications. Cortex-M52 offers this expanded safety package and features, enabling partners to achieve FuSa certification more quickly and easily than previous generations of Cortex-M deployed in these applications.

■ Simplifying AI development for ultra-small endpoint devices

Traditional embedded developers struggle to acquire the math, data analytics, toolchain expertise, and programming skills needed for AI. As AIoT shipments grow, support for developers is essential, and the Arm Cortex-M52 delivers the key capabilities needed for today’s modern development workflows.

In the past, achieving the ML and DSP performance that the Cortex-M52 delivers required a combination of CPU, DSP, and NPU, which meant that once the hardware was built, developers had to use three toolchains, three compilers, and three debuggers to write, debug, and tune code for the chip.

Now, Arm can deliver AI on a single toolchain and a single, validated architecture, with the industry-standard, user-friendly Arm Cortex-M portfolio. This ensures a unified development flow for traditional DSP and ML workloads, a seamless developer experience with specific integrations and optimizations for leading machine learning frameworks, and the ability to get the best performance from every Cortex-M.

The Cortex-M52 is fully software compatible with the Cortex-M55 and Cortex-M85, enabling developers to leverage and benefit from the growing software and tools ecosystem around Helium, as well as the extensive knowledge base of free software libraries and the Arm partner ecosystem.

To simplify and accelerate the IoT and embedded development process, the Cortex-M52 is also available with Arm Virtual Hardware, a cloud-based offering that enables software development ahead of silicon.