Groundbreaking chip patents as the key to neuromorphic computing

The TiF platform technology based on bismuth iron oxide brings reconfigurable memristors and transistors to industrial applications for the first time and addresses the very resource-intensive operation of matrix-matrix multiplication, which is essential for AI systems.

• 90 % energy savings, 400 times faster processing speed

Matrix multiplication is the central computing operation of modern AI systems and at the same time the largest energy consumer in data centers. Reconfigurable memristors and transistors make it possible to adapt the chip design directly to the mathematical structure of matrix multiplications - the foundation for specialized AI hardware that goes far beyond what classic GPUs, APUs or TPUs can do today. Test series on recognized simulation platforms show: TiF platform technology can achieve energy savings of up to 90% and a 400-fold increase in processing speed compared to the classic Von Neumann architecture.

"We are talking about a complete paradigm shift in chip architecture," says Prof. Dr. Heidemarie Krüger, co-founder of Techifab GmbH. "The reconfigurable memristors and transistors of our TiF platform technology are not small, incremental optimizations, but enable hardware to be designed for specific tasks and domains - be it for neural networks, quantum-safe encryption or industrial sensor technology."

• Techifab's path to industrial scaling

The start-up, which is funded by the Federal Agency for Leap Innovations (SPRIND), is pursuing a clear path to commercialization of its invention with this innovation. In addition to the search for further investors, the focus in the short term is on developing the first commercially viable applications.

"In the medium term, we are aiming to work closely with global integrated design manufacturers (IDMs) and design enablement providers as well as further R&D collaborations. Our goal is to achieve highly scaled integrations in AI data centers before the year 2030. In addition to the use in AI infrastructures, we also see enormous potential in edge computing (e.g. for autonomous driving)," comments Guy Vroemen, CEO of Techifab GmbH.

• About Techifab GmbH

Techifab GmbH was founded in 2021 by Prof. Dr. Heidemarie Krüger and Stephan Krüger and develops the platform technology for neuromorphic chips based on the TiF platform technology based on a novel semiconductor material. The start-up from Silicon Saxony addresses key weaknesses in traditional semiconductor architectures and offers a new, highly energy-efficient alternative for AI, robotics, cybersecurity and communication technologies. Techifab GmbH currently employs around 50 people at its Dresden site. With 24 registered patents and over EUR 20 million in investments to date, the company is one of the most promising representatives of the European DeepTech scene. The first memristor components, including administration software, have been available as a proof of concept in the Techifab online store since 2024.