BrainChip Holdings Ltd. has introduced the AKD1500 SPI, a co-processor that adds Akida neuromorphic AI to wearables, smart home devices and other microcontroller-based edge products through an SPI interface.
Most everyday devices already have a microcontroller, but adding AI usually means moving to a larger, more expensive processor and redesigning the product. The AKD1500 SPI works alongside the MCU, which relays sensor data to the chip; the AKD1500 SPI runs the full neural network and sends back results such as a recognized word or detected fault.
The chip supports edge learning, allowing devices to learn new words, sounds, users or patterns directly on the chip without sending data to the cloud or retraining the model. Running the model on-device also removes the cloud round trip, including latency, loss of AI when connectivity drops and the risk of data leaving the product. Voice, sensor and device-health data stay on the device. Like every Akida processor, the AKD1500 SPI can run several AI tasks on one chip.
BrainChip chief product officer Steve Brightfield said the AKD1500 SPI connects to an MCU already in use, while the AKD1500 PCIe serves systems that need more performance. He said both products run the same models with the same tools, allowing customers to grow without starting over. The PCIe version serves performance-driven systems and multi-chip designs; the SPI version targets consumer, industrial and healthcare products.
ABI Research forecasts that chipset shipments of TinyML AI will surpass 4.1 billion units by 2031, with dedicated AI processors the fastest-growing part of that segment. The AKD1500 SPI is available now from the BrainChip Shop, channel partner DigiKey or through BrainChip sales.
