BRN Discussion Ongoing

[Guzzi62]

12-top-emerging-technologies-to-catch​

10.Neuromorphic Computing
​

The energy inefficiency of traditional GPU architectures has become a bottleneck. To tackle that, neuromorphic computing introduces brain-inspired hardware designed to replicate how biological neurons process information.

What happens is that neuromorphic systems use spiking neural networks and parallel architectures that activate only when needed. They do not process data sequentially like conventional chips, so they reduce power consumption and enable real-time learning and pattern recognition.

Neuromorphic chips are gaining attention in 2026 for edge AI, robotics, autonomous systems, and IoT devices. They are useful where low latency and ultra-low power usage are critical.

BrainChip, a leader in the field, launched its AKD2500 silicon development project in February 2026. This $2.5 million project aims to integrate its next-generation Akida 2.0 neuromorphic architecture into silicon using TSMC’s 12-nanometer process. Prototype silicon is expected in Q3 of 2026.

Industry analysts predict commercial neuromorphic solutions will hit the market soon, as energy demands from AI continue to grow.

12 Top Emerging Technologies to Catch in 2026

Every decade has a moment where the future seems to arrive all at once.

iexchange.substack.com

 

[Fullmoonfever]

Another update a few hours ago to close the gap between Python model "training" and production C++ "deployment".

GitHub - Brainchip-Inc/AkidaCPPInferenceDeploymentPackage

Contribute to Brainchip-Inc/AkidaCPPInferenceDeploymentPackage development by creating an account on GitHub.

github.com

Brainchip-Inc/AkidaCPPInferenceDeploymentPackage

Akida C++ Inference Deployment Package​

Overview​

This package provides a complete, working example of deploying a trained Akida neural network model as a standalone C++ application on a Linux host. It bridges the gap between Python-based model training and production C++ deployment without requiring Python at runtime.

What's Included​

• Complete C++ inference pipeline (src/inference.cpp) with detailed comments explaining each API call
• Linux system implementation (src/system_linux.cpp) for required runtime hooks
• Model conversion script (scripts/convert_model.py) to transform .fbz models to C++ binaries
• Test input generator (scripts/generate_test_input.py) for validation against Python results
• Ready-to-build CMake configuration (src/CMakeLists.txt)

Target Use Case​

You have:

• A trained Keras model converted to Akida format (.fbz file)
• Validated the model works in Python using the Akida SDK
• An AKD1000 or AKD1500 device connected to a Linux host via PCIe

You want:

• A standalone C++ application that runs inference without Python
• Production-ready code you can integrate into your existing C++ codebase
• Full control over the inference pipeline (batching, latency measurement, pre/post-processing)

 

[7für7]

Is the buy side increasing maybe??

 

[HopalongPetrovski]

Is the buy side increasing maybe??

Don't think so.
Still those 10 million in prop bids down the bottom distorting the picture for anyone taking a quick glance at it.

 

[Pom down under]

Don't think so.
Still those 10 million in prop bids down the bottom distorting the picture for anyone taking a quick glance at it.

Isn’t it great when you see the SP increase by nearly 4% then realize the sell line has massively dropped from 2,855,332 to 2,855,329