The Netherlands is establishing a premier hub for neuromorphic computing.

      Our newest and most sophisticated technologies — including AI, Industrial IoT, advanced robotics, and self-driving vehicles — face significant challenges: high energy consumption, limited edge capabilities, system hallucinations, and notable accuracy discrepancies.

      A potential solution is emerging from the Netherlands, where the country is cultivating a promising neuromorphic computing ecosystem, leveraging neuroscience to enhance IT efficiency and performance. Billions of euros are being invested globally in this novel computing approach. The Netherlands aspires to lead the market by uniting startups, established companies, government entities, and academia within a neuromorphic computing ecosystem.

      A Dutch mission to the UK

      In March, a Dutch delegation arrived in the UK to conduct an “Innovation Mission” in partnership with local tech and government representatives. The mission, led by Top Sector ICT, a government-supported organization from the Netherlands, aimed to strengthen and explore the future of neuromorphic computing within Europe and the Netherlands.

      We reached out to Top Sector ICT, which connected us with Dr. Johan H. Mentink, a computational physics expert at Radboud University. Dr. Mentink discussed how neuromorphic computing could address current energy, accuracy, and efficiency issues faced by today’s computing architectures.

      “Modern digital computers utilize energy-intensive processes for data management,” Dr. Mentink stated. “Consequently, some contemporary data centers consume so much power they require their own power plants.”

      Current computing systems store data in one location (memory) while processing it elsewhere (processors), leading to significant energy expenditure on data transport, Dr. Mentink noted.

      Conversely, neuromorphic computing architectures differ at both hardware and software levels. Instead of relying on separate processors and memory, neuromorphic systems utilize new components like memristors, which function as both memory and processors.

      By enabling data processing and storage in a single hardware component, neuromorphic computing eliminates the energy-draining and error-prone process of data transfer. This immediate on-component data handling allows for quicker processing, resulting in faster decision-making, fewer hallucinations, improved accuracy, and enhanced performance. This approach is being utilized in edge computing, Industrial IoT, and robotics to enhance real-time decision-making.

      “Similar to how our brains handle and store information in the same location, we can create computers that combine data storage and processing within one unit,” Dr. Mentink elaborated.

      Early applications of neuromorphic computing

      Neuromorphic computing is beyond just an experimental phase, as numerous emerging and established tech firms are heavily investing in the development of new hardware, edge devices, software, and applications in this field.

      Major tech entities such as IBM, NVIDIA, and Intel with its Loihi chips are engaged in neuromorphic computing, while Dutch companies, aligned with a 2024 national white paper, are taking a prominent regional stance.

      For instance, Dutch firm Innatera — a frontrunner in ultra-low power neuromorphic processors — recently acquired €15 million in Series-A funding from Invest-NL Deep Tech Fund, the EIC Fund, MIG Capital, Matterwave Ventures, and Delft Enterprises.

      Innatera is merely the beginning, as the Netherlands continues to foster this burgeoning industry through funds, grants, and other incentives.

      Immediate applications for neuromorphic computing include event-based sensing technologies embedded in smart devices such as cameras or audio sensors. These neuromorphic devices process only changes, which can significantly reduce energy consumption and data load, according to Sylvester Kaczmarek, CEO of OrbiSky Systems, a company that integrates AI for space technology.

      Neuromorphic hardware and software could be pivotal in enabling AI on the edge, especially for low-power devices like mobiles, wearables, or IoT.

      Leading use cases include pattern recognition, keyword spotting, and simple diagnostics—such as real-time processing of complex sensor data streams for biomedical applications, robotics, or industrial monitoring, Dr. Kaczmarek explained.

      In pattern recognition and classification or anomaly detection, neuromorphic computing can make decisions very quickly and effectively.

      Professor Dr. Hans Hilgenkamp, Scientific Director of the MESA+ Institute at the University of Twente, concurred that pattern recognition is a strong area for neuromorphic computing.

      “One might also consider failure prediction in industrial or automotive contexts,” he added.

      The challenges creating neuromorphic opportunities

      Despite notable advancements, the establishment of robust neuromorphic computing ecosystems in the Netherlands faces hurdles. Globalized tech supply chains and the standardization of emerging technologies offer little leeway for hardware-level innovation.

      For instance, while optical networks and chips have exhibited superior performance compared to existing systems, this technology has not yet achieved widespread deployment. Introducing new hardware necessitates coordinated efforts between the public and private sectors. The global rollout of 5G technology exemplifies these challenges, requiring cooperation among telecommunications and governments to implement not only new antennas but also smartphones, laptops, and a range of compatible hardware.

      On the software front, 5G systems mandated global standards for integration