Traditional processors, such as computer processors (CPUs), graphics processing units (GPUs), and microcontrollers (MCUs), have formed the foundation of digital data processing for decades. These systems are predominantly based on von Neumann architectures or variations thereof. However, with the emergence of deep neural networks (DNNs), these traditional systems are reaching their limits in terms of energy consumption and processing speed. Neuromorphic processors (NPUs) are considered the solution to these challenges, as their biologically inspired architecture differs fundamentally from established processor types.

The parallel processing structure and the co-location of the processing unit and memory enable extremely efficient execution of AI tasks on these NPUs. In addition, the biologically inspired architecture enables the implementation of novel AI approaches, such as spiking neural networks (SNNs). SNNs are a special class of neural networks that mimic the functioning of the human brain even more closely than the widely used DNNs. Key differences include event-driven data processing via spikes and the ability to capture temporal information. This gives SNNs the potential to be more powerful and efficient in various applications, which explains the increased research interest in recent years.

When implementing SNNs and NPUs in practice, industrial companies face significant obstacles:

1. Difficult execution and training
2. Fragmented software and hardware market and lack of acceptance
3. Lack of unified methodology

The Neuromorphic Computing Solutions Group systematically addresses these challenges in the development and implementation of neuromorphic AI solutions and their industrial adoption.