A computing approach that requires up to 8,000 times less energy than conventional methods is emerging as a potential answer to one of technology’s most significant challenges: the unsustainable energy consumption of modern computing systems.
Neuromorphic technologies – which mimic the brain’s architecture by integrating storage and processing – are gaining momentum in the Netherlands, where a diverse ecosystem of researchers, startups and established tech companies is taking shape to advance the field.
“We’re essentially talking about a new way of computing,” said Frits Grotenhuis, director of Topsector ICT, a national organisation that coordinates and stimulates digital innovation in the Netherlands. “It particularly relates to storage capacity and data processing. With insights from how the human brain functions, we can compute faster, cheaper and with less energy.”
Unlike many emerging technologies that focus on a single aspect of computing, neuromorphic technologies span multiple dimensions. “It encompasses materials, devices, algorithms, architecture and applications,” he said. “That makes it fascinating but challenging – you need to consider advancing each dimension while maintaining their integration.”
At the hardware level, neuromorphic technology incorporates novel materials and components designed to mimic neural networks. The architecture fundamentally differs from conventional Von Neumann designs by eliminating the separation between memory and processing units. This co-location drastically reduces the energy required for data movement, enabling more adaptive and resilient computing.
“Data processing can be done more at the source itself, whether it’s sensors or otherwise,” said Grotenhuis. “This not only makes it many times more energy-efficient and sustainable, but also enhances privacy since sensitive data doesn’t need to be transmitted to external servers.”
Imagine a world where your computer runs on the energy of a light bulb instead of a power-hungry server farm. That’s the promise of neuromorphic computing. According to researchers behind a comprehensive whitepaper on neuromorphic computing in the Netherlands, analogue in-memory computing requires approximately 100 mW, compared with 100W for a typical central processing unit and 300-800W for a graphics processing unit. This represents a staggering efficiency gain of up to 8,000 times, drastically reducing power consumption while maintaining performance.
Practical applications
While the theoretical foundations of neuromorphic computing have existed for decades, recent advancements are accelerating its potential for practical applications. “The US and China are miles ahead on this topic,” said Grotenhuis. “But within Europe, we have a good position. The UK, Germany and Switzerland are quite advanced, and the Netherlands is somewhere in the middle or slightly above.”
Nevertheless, Dutch scientific output in this field is growing at a higher percentage rate than even China, according to a report on neuromorphic technologies commissioned by Topsector ICT. The country’s research ecosystem spans multiple institutions, including CogniGron at the University of Groningen, Mesa+/Brains at the University of Twente, and initiatives at TU Delft, TU Eindhoven and Radboud University.
Recent breakthroughs at these institutions align with work done elsewhere in the Netherlands. In 2022, researchers at the University of Twente announced they had developed molecular switches that mimic brain synapses – a critical component for neuromorphic systems. As professor Christian Nijhuis explained in previous coverage by Computer Weekly, these molecular switches could potentially make computing 10 to 100 times more energy efficient.
The Dutch ecosystem is also showing promising signs of market transfer. “The Netherlands now has several startups that are leading with neuromorphic technologies,” said Grotenhuis.
Companies like AxeleraAI and Innatera have already brought hardware to market, while others, such as Onward, GrAI Matter Labs and Ourobionics, are advancing various neuromorphic applications.
A spectrum of applications
One of the challenges in discussing neuromorphic technologies is the breadth of potential applications – from edge computing in internet of things devices to enhancing AI capabilities and enabling new healthcare services.
“It could involve precision farming in the agri-food sector, the energy sector becoming much more sustainable and smarter healthcare,” said Grotenhuis. “The potential is enormous, especially from the sustainability perspective and the energy efficiency of this new technology.”
According to Topsector ICT’s exploration, practical applications are expected to emerge at different timescales. Some, like “event-based” cameras – sensors that only consume energy when there is a change in the field of view – are already being deployed. These cameras represent an early example of neuromorphic sensing technology in action.
In the medium term, chemical detection, sound sensors and robotics applications are expected to gain traction. Longer-term applications (2030-2035) may include sophisticated systems for self-driving vehicles, EMG wearables that interface directly with neurons controlling muscles, and cooperative robot systems.
However, Grotenhuis is careful not to over-promise on immediate applications. “It’s difficult to predict exactly how this will develop,” he said. “Between now and five years, I think we’ll mainly see growth in companies interested in the technology, focused on innovation in the technology itself and initial applications. The truly disruptive effects will likely only become visible after five to 15 years.”
Building a national ecosystem
The Netherlands is working to create a more cohesive ecosystem around neuromorphic technologies to accelerate development in this field. Topsector ICT has supported initiatives, including the aforementioned white paper on the state of neuromorphic computing in the Netherlands and an upcoming trade mission to the UK to learn from and potentially collaborate with British researchers and companies.
This international connection fits within a broader European strategy. The European Commission has recently allocated hundreds of billions of euros for digitisation, including quantum computing and AI facilities. Similar investments in neuromorphic computing infrastructure could accelerate progress across the continent. “We really need to look at the European level together and seek collaboration where relevant and appropriate,” said Grotenhuis, noting that this is particularly important given the significant investments made by the US and China.
The efforts align with the Netherlands’ National Technology Strategy, which has identified neuromorphic technologies as one of 44 key technologies for the country’s future, and ten of these technologies have been prioritised for national action agendas. “We want to connect neuromorphic technologies to the agenda for AI and data, alongside crossovers with other technologies such as semiconductors and quantum computing,” he said.
An integrated approach is essential given the technology’s multidimensional nature and potential synergies with other advanced computing paradigms. The white paper on neuromorphic computing in the Netherlands emphasises that neuromorphic systems can complement existing technologies rather than replace them. For instance, neuromorphic computing can drastically improve AI’s energy efficiency by enabling local data processing. When combined with quantum computing, it could contribute to solving complex problems such as drug discovery. Integrated with photonics, optical neuromorphic systems could offer advantages where ultra-fast communication is crucial.
With projections of up to $19bn by 2030, neuromorphic computing is poised for exponential growth – provided key breakthroughs are achieved.
What’s certain is that the quest for more energy-efficient computing is not a luxury, but a necessity. In the Netherlands – a country with limited physical space and grid capacity for datacentres – neuromorphic computing offers not only economic opportunities, but also a concrete contribution to sustainability goals.
As the Dutch ecosystem for neuromorphic technologies continues to evolve, it represents a significant bet on a computing approach that draws inspiration from nature’s most sophisticated computing system – the human brain. While we may never match the brain’s incredible efficiency outside our bodies, even a fraction of its capabilities could transform computing as we know it.
#Netherlands #bets #braininspired #computing #greener #future