BrainChip (ASX: BRN) has begun shipping an Akida AKD1500 PCIe card, a plug-in accelerator that lets any desktop, workstation, industrial PC or single-board computer run the company's ultra-low-power neuromorphic engine without dedicated hardware. The card, announced on September 17, 2026 and available now from BrainChip's online store, is the first time the second-generation Akida silicon has been packaged for mainstream evaluation.
Neuromorphic Compute In A Standard Slot
The AKD1500 is BrainChip's event-based neuromorphic processor. It runs convolutional, transformer and temporal neural networks locally on-device, learns new classes on-chip and sips power in the sub-milliwatt range for typical edge workloads. Historically that capability required either a custom design win or one of BrainChip's Raspberry-Pi-sized dev kits. The PCIe form factor changes the entry point: any team with a spare Gen3 slot can now drop the card in and start benchmarking.
From IP To Off-The-Shelf Silicon
"Our customers no longer have to choose between proving out neuromorphic AI and productizing it," said BrainChip CEO Sean Hehir. "Akida is available in every form they need, from card to IP." Chief product officer Steve Brightfield framed the play in plainer terms: "If you have a PC with a spare slot, you can be running your own models on Akida this afternoon."
BrainChip already licenses Akida as IP to partners such as MegaChips, Renesas and Prophesee, and sells USB dongles, mini-PCIe modules and full evaluation kits. The AKD1500 card sits alongside those channels rather than replacing them. It targets research groups, defense integrators, automotive suppliers and industrial-inspection vendors that need to profile edge inference against real workloads before committing to silicon.
Edge AI Widens Its Front
The launch lands in a market where every AI-accelerator vendor is trying to move inference off the cloud. Traditional GPU makers are pitching smaller data-center parts for edge boxes, while Qualcomm, Ambarella, Hailo and NXP are chasing dedicated neural silicon. BrainChip's angle is that event-driven, sparsity-aware computation lets it match those competitors on accuracy at a fraction of the power — a claim its customers can now stress-test on their own hardware, not BrainChip's.
Related coverage on The Robotics Media: Seyond's Hummingbird D1R LiDAR moves from cars to robots, Nexstrom's 12-inch 2D semiconductors and Vantora's physical-AI venture studio track the same shift toward on-device intelligence.
Reporting based on coverage from BrainChip's official press release, HPCwire/AIwire and Edge AI + Vision Alliance.