Google and semiconductor manufacturer Synaptics Inc. announced a strategic partnership on Tuesday to bring artificial intelligence (AI) out of distant data centers and directly into everyday consumer and industrial devices.
The collaboration debuts Synaptics’ new Astra SR200 series of processors alongside the expanded release of the jointly developed Coralboard development platform. Built to power hardware like smart glasses, factory cameras, and security sensors, the technology enables devices to run sophisticated AI models locally without relying on constant cloud connections or heavy power supplies.
The initiative marks a crucial milestone in Google’s strategy to democratize AI hardware. In October 2025, Google Research open-sourced its Coral neural processing unit (NPU) architecture under an Apache 2.0 license, allowing silicon manufacturers to integrate the design without paying licensing fees or royalties. The open standard utilizes the RISC-V instruction set to streamline development across disparate hardware platforms.
“One of our goals was to try to minimize developer pain,” said Billy Rutledge, director at Google Research, told VentureBeat. Rutledge noted that creating a standardized NPU design helps developers bypass the friction of building for various proprietary chips, adding that “the most useful devices that will use AI at the Edge are still yet to be designed.”
Synaptics serves as Google’s inaugural commercial silicon partner for the Coral ecosystem. Its new Astra SR200 lineup features three distinct chips combining Arm Cortex-M cores with Google’s Coral NPU: SR210 (entry-level) delivers 200 billion operations per second (GOPS) for always-on audio and sensing tasks; SR230 (mid-range) reaches 300 GOPS, adding display and vision support; and SR250 (flagship) provides 400 GOPS for complex, multimodal workloads.
To address standard battery constraints, Synaptics embedded up to 8MB of SRAM and custom DRAM configurations into the hardware, claiming the setup cuts memory power consumption to roughly one-third of traditional system DRAM.
The architecture supports lightweight AI models such as Google’s Gemma 3 270M, Liquid AI’s LFM2.5, and Ultralytics’ YOLO vision tools. According to Synaptics, the SR200 hardware can process Gemma 3 270M at 12 to 16 tokens per second, enabling real-time local processing for enterprise clients like audio brand Sonos, security firm Verisure, and AR maker Jorjin Technologies.
The announcement comes as Synaptics enters an increasingly crowded edge-AI market, competing against established microcontrollers from STMicroelectronics and Renesas. Additionally, Synaptics is preparing for a proposed $5.7 billion acquisition by onsemi, scheduled to close by mid-2027.
Engineers can immediately begin testing edge-AI applications using the Coralboard SL development kit, available through DigiKey for $99.79. Sampling for the SR200 processor family is currently underway, with the companion Coralboard SR set to release in the future.
