The Age of Light II
A Five-Chapter Exploration of Physical Intelligence
Executive Summary — Scanable Insights
- AI is hitting a physical limit: moving data consumes more energy than computing it. The bottleneck is no longer software, nor transistors, but infrastructure.
- Photonics is the solution: using light to move and process information enables computation that is faster, parallel and energy-efficient.
- Physical Intelligence emerges: computation embedded in hardware, memory and interconnects simultaneously — intelligence becomes a property of the system itself.
- Human-centered AI requires edge deployment: responsive, local intelligence can only exist near humans, within energy and latency constraints.
- Neuromorphic photonics enables brain-like systems: optical hardware performs learning and adaptation in parallel, mimicking biological efficiency.
- Practical applications: autonomous vehicles, care robotics, industrial smart grids — systems that sense, compute and act at light speed.
- Strategic and geopolitical implications: control over photonic infrastructure defines industrial and national intelligence sovereignty.
- Energy and materials are now strategic assets: the physical substrate of AI is as critical as software, influencing global power and competitive advantage.
- The Age of Light is here: intelligence is embodied, local and physically constrained — human-centered AI is now enforceable by physics, not aspiration.
Chapter I — Introduction: The Limits of Digital AI
Artificial intelligence has long been measured by algorithms, model size and software sophistication. Yet beneath this digital narrative, a quieter crisis is emerging: the energy required to move data now exceeds the energy required to compute it. The bottleneck is no longer software, nor even transistors — it is the infrastructure connecting everything together.
“The energy required to move data between the memory and the processor is orders of magnitude higher than the energy required to actually perform the computation. We are hitting a physical wall where electricity alone cannot scale AI further.”
Prof. Yurii Vlasov, Grainger Distinguished Chair in Engineering, Professor of Physics, University of Illinois Urbana-Champaign
Photonics — the manipulation of light to move and process information — is no longer a futuristic promise. It is a structural necessity. Moving data with light is the first step; thinking with light is the next.
Chapter II — From Electronic Intelligence to Physical Intelligence
Why human-centered AI is ultimately a physical question
Human cognition is embodied, constrained by energy, time and structure. Intelligence evolved within a medium that allowed it to act locally and efficiently. Similarly, AI cannot truly be human-centered unless it exists within the physical limits that humans inhabit.
The invisible bottleneck
Electronic AI has scaled to its physical limits: computation is cheap, but moving information is expensive.
“The energy required to move data between the memory and the processor is orders of magnitude higher than the energy required to actually perform the computation. We are hitting a physical wall where electricity alone cannot scale AI further.”
Prof. Yurii Vlasov, Grainger Distinguished Chair in Engineering, Professor of Physics, University of Illinois Urbana-Champaign
From symbols to fields
Photonics allows computation to occur in the physics itself. Matrix multiplications — the heart of AI — can be performed directly by manipulating light waves.
“We are moving away from traditional Von Neumann architectures. By using light to perform calculations, we can execute matrix multiplications—the heart of AI—at the speed of light, with almost zero energy consumption for the calculation itself.”
Prof. Harish Bhaskaran, Professor of Applied Nanomaterials, University of Oxford
Computation becomes structural. Memory, processing and communication merge. Latency collapses. Energy efficiency emerges as a physical property, not an optimization.
Toward physical intelligence
Embodied computation enables local, adaptive intelligence. Human-centered AI is no longer aspirational — it is physically constrained. Systems can exist alongside humans, within their energy budgets and temporal rhythms, because physics now defines what intelligence can do.
“Photonics allows us to build ‘brain-like’ hardware that processes information in a way that is inherently parallel. This isn’t just a faster chip; it’s a hardware architecture that mimics the biological efficiency of the human brain.”
Prof. Peter Bienstman, Professor in Photonics and Neuro-inspired Computing, Ghent University / IMEC
Chapter III — When Intelligence is Embodied
Redefining intelligence
Chapter III examines what happens when computation is no longer abstract but embedded in physical, adaptive systems.
Intelligence as a physical phenomenon
Embodied intelligence cannot exist independently of its medium. Physical substrates define capabilities, speed and learning.
“Photonics allows us to build ‘brain-like’ hardware that processes information in a way that is inherently parallel. This isn’t just a faster chip; it’s a hardware architecture that mimics the biological efficiency of the human brain.”
Prof. Peter Bienstman, Professor in Photonics and Neuro-inspired Computing, Ghent University / IMEC
Human-centered AI in a photonic world
Local, edge-deployed intelligence ensures responsiveness and autonomy.
“To make AI truly autonomous and responsive in the real world, we need to move the intelligence out of the cloud and into the device. Photonics is the only medium that can provide the necessary bandwidth and low latency within the strict energy budgets of edge devices.”
Dr. Chang Gao, Assistant Professor in Neuromorphic Computing Hardware, TU Delft
Adaptive, embodied learning
Learning occurs directly in the hardware, leveraging interference patterns and optical dynamics. Intelligence becomes local, parallel and adaptive — inseparable from its environment.
Strategic and societal implications
Infrastructure, energy and proximity now define capability. Human-centered AI emerges from the physics of light.
“Silicon photonics is the bridge that allows the semiconductor industry to survive the post-Moore era. It is not an evolution; it is a fundamental shift in how we define a ‘computer’ in the 21st century.”
Prof. Roel Baets, Full Professor at the Photonics Research Group, Ghent University / IMEC
Chapter IV — Photonics Brains in Practice
Neuromorphic photonics: the first brains of light
- Signals propagate as light; computation occurs intrinsically.
- Memory, processing, and routing collapse into a single substrate.
“We are creating systems that do not merely simulate brain-like behavior; the hardware itself computes, learns and adapts in parallel.”
Prof. Peter Bienstman, Professor in Photonics and Neuro-inspired Computing, Ghent University / IMEC
Edge intelligence becomes reality
- Critical for autonomous vehicles, robotics and industrial systems.
- Low latency and energy-efficient local AI are now technically feasible.
“To make AI truly autonomous and responsive in the real world, we need to move the intelligence out of the cloud and into the device. Photonics is the only medium that can meet the strict latency and energy requirements.”
Dr. Chang Gao, Assistant Professor in Neuromorphic Computing Hardware, TU Delft
Case studies
- Autonomous mobility: real-time sensing and planning at light-speed.
- Healthcare and care robotics: continuous monitoring and adaptive assistance at human scale.
- Industrial infrastructure: instant decision-making in smart grids and factories.
“We are witnessing the first generation of brains built from light — systems that sense, compute and learn simultaneously. The distinction between hardware and intelligence is disappearing.”
Prof. Peter Bienstman, Professor in Photonics and Neuro-inspired Computing, Ghent University / IMEC
Chapter V — Geopolitics, Energy and Strategic Sovereignty
Intelligence as a strategic resource
Photonics transforms AI from software into physical intelligence, making control over materials, energy and infrastructure decisive.
“The energy required to move data between the memory and the processor is orders of magnitude higher than the energy required to actually perform the computation. We are hitting a physical wall where electricity alone cannot scale AI further.”
Prof. Yurii Vlasov, Grainger Distinguished Chair in Engineering, Professor of Physics, University of Illinois Urbana-Champaign
Sovereignty in the Age of Light
- Edge intelligence cannot be exported.
- Photonic supply chains are strategic chokepoints.
- National deployment requires control over energy, materials and fabrication.
“Silicon photonics is the bridge that allows the semiconductor industry to survive the post-Moore era. It is not an evolution; it is a fundamental shift in how we define a ‘computer’ in the 21st century.”
Prof. Roel Baets, Full Professor at the Photonics Research Group, Ghent University / IMEC
Industry and governance
- Corporations with photonic expertise gain leverage.
- Governments securing domestic infrastructure define national intelligence sovereignty.
- Policy decisions on energy and infrastructure directly affect AI deployment.
“To make AI truly autonomous and responsive in the real world, we need to move the intelligence out of the cloud and into the device. Photonics is the only medium that can provide the necessary bandwidth and low latency within the strict energy budgets of edge devices.”
Dr. Chang Gao, Assistant Professor in Neuromorphic Computing Hardware, TU Delft
Strategic implications
- Investment in photonic edge systems is now a strategic imperative.
- Intelligence is infrastructure. Control over it defines competitive and national advantage.
- Human-centered AI is enforced by physical law, not policy alone.
Closing: the Age of Light
The Age of Light is here:
- Intelligence is physical, embodied and local.
- Human-centered AI emerges naturally from the physics of light.
- Strategy, industry and governance converge around control of energy, infrastructure and photonic technology.
“We are witnessing the first generation of brains built from light — systems that sense, compute and learn simultaneously. The distinction between hardware and intelligence is disappearing.”
Prof. Peter Bienstman, Professor in Photonics and Neuro-inspired Computing, Ghent University / IMEC
The world now faces a new landscape of intelligence, where matter, energy, and cognition are inseparable — and where foresight, infrastructure, and physics define who leads in the next era.
Photo credit: IMEC – Physical intelligence is no longer a concept — it’s emerging in labs today. This is where AI meets photonics and where the first brains built from light are taking shape.
