Specialist for brain-inspired supercomputers

With the SpiNNcloud systems, artificial intelligence (AI) can complete even the most demanding tasks with drastically less energy consumption than with conventional graphics processors (GPUs) - in some cases reduced by a factor of 79.

At the heart of the SpiNNcloud systems is the specially developed SpiNNaker2 chip (Spiking Neural Network Architecture), which was brought to market maturity as part of the "Human Brain Project". Its revolutionary architecture, inspired by the human brain, uses event-driven communication. In contrast to conventional graphics processing units (GPUs), the system only calculates when data changes. This drastically reduces the power consumption of such computers.

The SpiNNaker2 architecture delivers up to 18 times higher energy efficiency than current GPUs for AI analysis. The advanced chip is manufactured at Globalfoundries in Dresden and underlines Saxony's innovative strength.

SpiNNcloud is the only European provider of such specialized AI supercomputers. As a result, the company has also acquired major orders in Germany and abroad. Users include the AI competence center ScaDS.AI and the University of Leipzig, which uses the system with around 650,000 cores for drug development. Internationally, the company equipped the renowned THOR consortium at the University of Texas at San Antonio (UTSA) in the USA with a system that simulates 393 million neurons. A milestone was achieved with the supercomputer "SpiNNcloud", which uses 35,000 chips and over five million processor cores in the first expansion stage. "We have built the world's largest brain-inspired supercomputer," the company estimates.

With these and other innovations, SpiNNcloud Systems is helping to strengthen European AI sovereignty. Through partnerships, such as with the French cloud provider Clever Cloud, the start-up is working to establish a cost-effective and sustainable AI infrastructure in Europe. At the same time, the company is pushing ahead with its internationalization strategy. A US spin-off, SpiNNcloud North America Inc. has now been established.

And research is also continuing: the Dresden engineers are already working on the next generation of neuromorphic architectures, which will be based on the new SpinNext chips. This should enable up to 78 times better energy efficiency compared to conventional GPUs.