Our most advanced technologies — AI, Industrial IoT and advanced robotics to self-driving vehicles — all share serious problems. Massive energy consumption, limited capabilities on the edge, system hallucinations and serious accuracy gaps. In the Netherlands, a possible solution is emerging. The Netherlands is developing a promising eco-system for neuromorphic computer, which uses neuroscience to boost IT efficiency and performance. This new form of computing is attracting billions of euros in investment worldwide. The Netherlands wants to be a leader on the market, bringing together startups and established companies, government organizations, and academics, in a neuromorphic computer ecosystem.
Dutch mission to the UK.
A Dutch delegation arrived in the UK in March to host a ““Innovation Mission “with local tech and governments. Top Sector ICT, a Dutch government supported organisation, led the mission. The mission aimed to discuss and strengthen the future of neuromorphic computer in Europe and the Netherlands.
Top Sector ICT connected us with Dr Johan H. Mentink a computational physics expert at Radboud University, Netherlands. Dr Mentink discussed how neuromorphic computer architectures can be improved to solve energy, accuracy and efficiency problems. Dr Mentink stated that current digital computers use energy-hungry processes in order to handle data.
Some modern data centres consume so much power that they need their own electricity plant.
Computers today store data in memory and process it in processors. Dr Mentink explained that this means a lot energy is used to transport data.
Neuromorphic computing architectures, on the other hand, are different both at the hardware level and at software level. Neuromorphic systems, for example, rely on new hardware components, such as memristors, instead of processors and memory. These are both processors and memory.
Neuromorphic computing eliminates the energy-intensive, error-prone task that is data transport by processing and saving the data on the same component. Data can be processed faster because it is stored on these components. This results in better performance, reduced hallucinations and improved accuracy. This concept is being used in edge computing, Industrial IoT and robotics, to make faster real-time decisions.
Dr Mentink explained that “just like our brains store and process information in the same location, we can build computers that combine data storage and computing in one place.”
Early applications of neuromorphic computing.
The use of neuromorphic computing is not just an experiment. Many established and new technology companies are heavily investing in the development of new hardware, edge devices and software, as well as neuromorphic computing applications.
The biggest tech brands, such as IBM, NVIDIA, Intel, are all involved in Neuromorphic Computing, while companies from the Netherlands are playing a leading role in the region, aligned to a 2024
The Dutch company Innatera, a leader in ultra-low-power neuromorphic processors, recently secured EUR15million in Series-A financing from Invest-NL Deep Tech Fund and the EIC Fund. MIG Capital, Matterwave Ventures and Delft Enterprises also contributed. Innatera represents just the tip of iceberg as the Netherlands continues its support for the new industry with funds, grants and other incentives.
The immediate use cases of neuromorphic computing are event-based sensing technology integrated into smart sensors, such as cameras or sound. Sylvester Kaczmarek is the CEO of OrbiSky Systems. The company provides AI integration for space technology.
Neuromorphic software and hardware have the potential to transfer AI on the edge. This is especially true for low-power devices like mobiles, wearables or IoT. Dr Kaczmarek explains that pattern recognition, keyword spotting and simple diagnostics – such as real time signal processing of complex sensor streams for biomedical applications, robotics or industrial monitoring – are some of the most popular use cases.
Neuromorphic computing, when applied to pattern detection and classification, or anomaly identification, can make decisions quickly and efficiently.
Prof. Dr Hans Hilgenkamp of the MESA+ Institute of the University of Twente agreed that pattern recognition was one of the areas where neuromorphic computer excels. “One can also think of [for example] fail prediction in industrial and automotive applications,” he added.
The neuromorphic opportunities created by the gaps
Despite recent progress in the Netherlands, the road towards establishing robust neuromorphic computer ecosystems is challenging. Globalised tech supply chain and standardisation of new technology leave little room for innovation at the hardware level.
For instance, andoptic networks and optical chip outperform the traditional systems currently in use, but this tech hasn’t been deployed globally. The public and private sectors must work together to develop new hardware. The global rollout 5G technology is a good example. It required telcos around the world to deploy new antennas as well as smartphones, laptops and a large amount of hardware to support the new standard.
On software, 5G systems needed global standards for integration, interoperability and smooth deployment. In addition, established telcos were forced to shift from pure competition to a strategic collaboration — a new concept for an industry that has been built on siloed operation. Neuromorphic computing eco-systems face similar challenges. The Netherlands recognizesthat the success of the entire industry depends on innovation in materials and devices, circuit designs, algorithms, hardware architecture, and applications. These challenges and gaps create new opportunities for tech companies and startups, vendors and partners.
Kaczmarek said that neuromorphic computing required full-stack integrated. This requires expertise that can link novel materials and devices, circuit design and architectures, algorithms and applications. “Bringing these layers to together is critical but challenging,” he stated.
On both the software and algorithmic side, it is important to develop new paradigms for programming, learn rules (beyond deep learning backpropagation) and software tools that are native to neuromorphic equipment. “It’s crucial to make hardware usable and efficient – co-designing algorithms and hardware because they are intimately linked in neuromorphic system,” said Dr Kaczmarek. Other industries that have developed or are looking at research on neuromorphic computers include healthcare (brain computer interfaces and prosthetics), sustainable energy, agri-food and sustainable energy. Neuromorphic computing components or modules can be integrated with conventional CMOS technologies, photonics, AI and even quantum technology. Long-term opportunities for the Netherlands
Dr Hilgenkamp was asked what expertise or innovations were most needed, and offered the greatest opportunity for contribution and growth in this emerging ecosystem.
Dr Hilgenkamp explained that the long-term development involves new materials and research, both of which are already happening on an academic scale.
Dr Hilgenkamp added that the concept of “materials which can learn” opens up new concepts in materials sciences that are exciting to researchers. Dr Mentink, on the other hand, emphasized the potential to transform our economies that rely heavily on the processing of massive amounts data. He said that even replacing a small portion of this with neuromorphic computing would lead to massive energy savings.
Moreover, with neuromorphic computation, much more processing is possible close to the place where the data was produced. This is great news for situations where data contains sensitive information.
According to Dr Mentink concrete examples include fraud detection in credit card transactions, image processing by robots and drones as well as anomaly detection of heartbeats and processing of telecom data. Dr Mentink said that the most promising use cases involve large data flows, high demands for fast response times and low energy budgets.
Dr Mentink anticipates that the development of software toolschains to enable rapid adoption of neuromorphic platforms will grow as the use cases of neuromorphic computing expand. This new sector will include services that streamline deployment.
Dr Mentink stated that “longer-term sustainable development requires a concerted, interdisciplinary effort across all computing layers to enable seamless integration between foundational discoveries and applications in new neuromorphic computer systems.”
The bottom-line
The promise of neuromorphic computer has led to billions of dollars of investment in the Netherlands, Europe, Asia and other parts of the world. Businesses that can innovate and integrate neuromorphic technologies at the hardware and software level will benefit most.
The potential for neuromorphic computing to improve energy efficiency and performance can be felt across industries. The integration of the technology will benefit energy, healthcare, robotics and AI, industrial IoT and quantum tech. If the Dutch ecosystem is successful, the Netherlands could be in a leading position.
The TNW Conferencewill be taking place in Amsterdam on June 19-20. Its mission is to support Dutch tech. Tickets are now available and– use the code TNWXMEDIA2025 to save 30%.
