BrainChip unveils AKD1500 SPI AI chip for microcontroller-based devices

BrainChip announced on 22 October that it had added a new co-processor to its AKD1500 family, using an SPI interface to work alongside microcontrollers. The company said the product is intended to bring Akida neuromorphic AI to wearables, smart-home equipment and sensor-based edge devices.

The AKD1500 SPI is described as a dedicated AI processor paired with an existing host microcontroller. In the stated workflow, the microcontroller passes sensor data to the chip, the neural network runs locally and the result returns to the host. The design is intended to let manufacturers add one chip without replacing their main processor.

BrainChip said the product’s distinguishing feature is on-chip learning of new words, sounds, users or patterns. The company said this can happen without sending data to the cloud or retraining the model. The release did not provide independent comparative measurements for power consumption, latency or accuracy.

Running the chip on-device could reduce reliance on cloud services when connectivity drops and allow voice or sensor data to remain on the product. Those are potential consequences of the architecture; the announcement did not report measured privacy or latency outcomes for a named customer device.

BrainChip said the AKD1500 SPI is positioned to run the same models and tools as its PCIe AKD1500 variant. Neuromorphyx executives said in the release that the product had been evaluated within real embedded-system constraints, but test conditions, performance results and third-party certifications were not disclosed.

The company said the chip was available through the BrainChip Shop, DigiKey and direct sales. The announcement also cited an ABI Research forecast that TinyML chipset shipments would exceed 4.1 billion units by 2031; this is a market-research projection, not a BrainChip sales forecast or reported shipment total.

The announcement places the SPI device beside the existing AKD1500 PCIe product rather than presenting it as a replacement for that design. BrainChip says both can use the same models and tools, while the new interface targets products whose host is a microcontroller. That positioning could make integration easier for manufacturers, but the release does not quantify design time, bill-of-material savings or the number of customers adopting the chip.

BrainChip named use cases ranging from wearables and home devices to robotics, industrial sensors and healthcare products. One cited partner described using the processor in EEG glasses intended for everyday wear by people with epilepsy. The company announcement did not identify a commercial deployment volume, provide clinical evidence or establish that the device is available to patients; the example describes intended product work.