Synaptics and Google Team Up on Edge AI

Generative artificial intelligence (AI) applications — such as large language models and image and video generators — have really stolen the spotlight in the past few years. So much so, that many people now see these tools as being synonymous with AI itself. But that, of course, is not the case. All sorts of predictive algorithms were deployed in healthcare, retail, and industry long before ChatGPT ever became a household name.

It is in this world where substance reigns over hype that Synaptics is focusing its attention. The company has just announced the Astra SL2600 Series — the next generation of its multimodal Edge AI processors designed to bring high-performance, low-power intelligence to the Internet of Things. The first to launch in this new line, the Astra SL2610 product family, integrates advanced AI acceleration, security, and connectivity technologies to power everything from smart appliances and industrial automation to robotics, healthcare devices, and even casual gaming systems.

At the core of the SL2610 series is the Synaptics Torq platform, which combines neural processing unit (NPU) architectures with open source compilers for flexible development. The processors feature Arm Cortex-A55 and Cortex-M52 cores, a Mali GPU, and built-in AI security mechanisms. Together, these components deliver efficient multimodal AI performance — handling tasks like speech, vision, and sensor fusion — while maintaining strong privacy protections by keeping data at the edge.

The SL2610 line consists of five pin-to-pin compatible families, ranging from energy-efficient versions for battery-powered devices to high-performance configurations for vision-based systems. Each supports integration with Synaptics’ Veros Connectivity technologies, offering Wi-Fi 6/6E/7, Bluetooth, Thread, and UWB. This unified design approach simplifies development and reduces time-to-market for manufacturers looking to embed AI into their products.

The Astra platform was developed in partnership with Google to deliver the first production implementation of Google’s new open source, RISC-V-based Coral NPU. This collaboration brings AI acceleration to low-power devices while giving developers access to an open compiler stack. The Coral NPU, with its scalar, vector, and matrix execution units, is built to eliminate the software fragmentation that has long plagued entry-level AI accelerators, making them difficult to program.

Sampling for the SL2610 family has begun, with general availability planned for the second quarter of 2026. For developers eager to get a head start, Synaptics is offering early access to its Astra Machina SL2610 Development Kit.