Largely, public attention to AI has centred around models, applications and software – but the AI economy depends on physical infrastructure: chips, interconnects, data centres, energy systems, materials, manufacturing capacity, a skilled workforce and architectures that move and process data as efficiently as possible. 

The UK Government’s recently announced AI Hardware Plan recognises this critical fact, and for the photonics sector, this is a particularly significant moment. 

The plan places photonics firmly within the national AI hardware conversation, not as a peripheral enabling technology, but as part of the core architecture required for the next generation of AI systems.  

It acknowledges that as AI workloads diversify, the hardware that supports them must diversify too – training, inference, edge AI, embodied AI and secure AI systems will not all be served by a single model of compute on conventional systems.  

Becoming a global AI powerhouse will not be achieved by consuming technologies developed elsewhere. It will require capability across the infrastructure layers that make AI scalable, secure and economically viable. 

Infrastructure layers that photonics is perfectly poised to provide. 

AI leadership is an infrastructure challenge

The Government’s plan is built around a clear objective: to strengthen the UK’s capability in the chips and semiconductor technologies that underpin AI. 

This doesn’t just address industrial, economic or policy aspects, but sovereignty too. 

AI infrastructure is becoming one of the defining sources of economic and strategic power. Countries that can develop, deploy and scale critical AI hardware will have greater control over how AI is used, how their data is protected, where value is created, and the resilience of their technology ecosystems. 

The UK cannot and should not attempt to replicate the supply chains of say, the US or China, but we can build huge leverage in the areas that set us apart – areas like photonics. 

The UK has world-class expertise in photonics, optical systems, semiconductor design, advanced materials and secure computing from our universities to our companies. These strengths sit at the intersection of the challenges now facing AI infrastructure: performance, energy efficiency, data movement, security and scale.

The memory wall and the data movement bottleneck

Larger models, larger datasets and larger clusters have driven extraordinary advances in AI, but as these systems scale and demand grows, new limitations are emerging. 

Conventional systems are fast hitting data bottlenecks and memory walls, constrained by how efficiently data can move through and to systems. 

Modern AI infrastructure depends on vast flows of data between processors, memory, accelerators, servers and data centres. Every movement introduces latency, consumes energy and adds complexity. Transporting these swathes of data can become more costly than computing on it. 

This is where investing in and implementing photonics becomes imperative. 

Photonics uses light to transmit and manipulate information and already underpins the fibre optic networks that form the backbone of the internet. In AI infrastructure, photonics offers a route to higher bandwidth, lower latency and greater energy efficiency. 

But the opportunity extends beyond moving data faster – the next step is developing optical infrastructure that can become more computationally useful. 

Photonics as a double-pronged sovereignty technology

Photonics matters to AI sovereignty in two ways. 

The first is efficiency. 

For AI to scale sustainably, infrastructure must deliver more performance without proportional increases in power, cost and complexity. Photonic technologies can help address some of the key physical constraints facing AI systems: data movement, bandwidth, heat and energy consumption. 

This is essential for data centres, cloud infrastructure, frontier compute and distributed AI systems. Sovereign AI must be economically and energetically viable because infrastructure that can’t scale efficiently can’t support long-term AI leadership. 

The second is security. 

The most valuable AI use cases will involve sensitive data: financial information, healthcare records, defence workloads, national infrastructure, proprietary models and confidential enterprise data. To unlock those use cases safely, the UK will need infrastructure that supports secure and privacy-preserving computation. 

That means post-quantum cryptography, fully homomorphic encryption, and confidential AI. It means protecting data not only when it is stored or transmitted, but also when it is being processed. 

These workloads are mathematically demanding and require infrastructure capable of handling large, structured and data-intensive operations efficiently. 

Photonics is well suited to this class of challenge. Light-based systems can support high-throughput operations and structured mathematical transformations, offering a route to making secure computation more practical at scale. 

If the UK wants to position itself as a global safe harbour for privacy-preserving AI and confidential data processing, it needs the hardware infrastructure to support that ambition, underpinned by photonics to boost performance as well as trust. 

Where Compute-in-Transit fits 

At Optalysys, we develop photonic chip technology that enables Compute-in-Transit for data centres: processing data directly within optical transport at communication line speeds.  

Today, compute and transport are typically treated as separate layers. Data is moved to a processor, computation is performed, the result is moved again. 

That separation creates latency, energy cost and system complexity. 

Compute-in-Transit changes the relationship between compute and data movement. By bringing processing into the data transfer past, we can help reduce pressure on memory, interconnects and electronic processing systems. 

For AI, this unlocks new levels of efficiency and scaling. 

For secure computing, it creates a route to accelerating structured workloads such as those found in post-quantum cryptography and fully homomorphic encryption. 

For sovereignty, it gives the UK an opportunity to lead in a differentiated layer of the AI hardware stack: not by copying the largest global players, but by defining the future of compute architecture. 

Making the UK AI Hardware Plan a reality

The UK has the research base, technical expertise and intellectual property to become a world leader in photonics — a domain that will be central to the future of AI hardware, but leadership will not happen by default.  

To seize this opportunity, the UK must turn academic strength into commercial scale, connect regional innovation into a coherent national ecosystem, and unlock the long-term capital needed for deep-tech companies to grow from here. 

The AI Hardware Plan is an important signal of intent. Now the task is to convert that intent into action: backing the technologies, supply chains and architectures that will make AI more efficient, more secure and more sovereign. 

Photonics gives the UK a chance to shape the next era of compute rather than simply consume it.

---

At Optalysys we’re pioneering the architectural revolution enabling photonic compute-in-transit. Get in touch with us to find out how we can bring efficiency gains to your use case →