Where AI convergence becomes real- the UK’s opportunity starts with the chip
We talk a lot about AI converging with quantum, photonics, robotics and other frontier technologies. But convergence doesn't happen in the abstract. It happens on a chip, in a package, inside a data centre rack. A breakthrough stays in the lab until it can be built in silicon (or light), powered, cooled and manufactured at scale.
That makes semiconductors the place where AI convergence becomes real. As AI meets photonics, quantum, robotics and new materials, the biggest opportunities for the UK will be where those technologies come together on the chip.
AI is reshaping what chips look like
For decades, progress in computing followed a simple formula: shrink the transistor, and performance improves. That formula is running out of road. At the same time, AI has created an appetite for compute, and for energy, that general-purpose chips were never designed to satisfy.
The result is a shift from "faster" to "smarter". At Microelectronics UK last month, I chaired a panel on balancing performance and efficiency in next-generation chip design, and the prevailing theme was that performance alone is no longer the sole marker of a "good" chip. Power, area and cost now matter just as much, and what counts as "performance" looks very different in a data centre, a car or a device at the edge. That is driving a move towards specialised hardware designed around specific AI workloads, from inference accelerators to brain-inspired neuromorphic chips.
AI is also changing how chips are made. AI tools are increasingly used across the design flow, helping engineers manage complexity that would be impossible to handle by hand. In other words, AI is now helping to build the hardware that AI runs on. That is convergence in its most literal form.
Where AI convergence is already happening
Look closely at the frontier of chip design and you'll find AI meeting several technologies at once.
AI and chiplets. AI workloads are driving the move from one giant chip to systems built from smaller, specialised "chiplets", often from different suppliers and technologies, in a single package. This shift has been characterised by some in industry as moving from system-on-chip to system-on-package. That opens the door to mixing silicon, photonics and specialist accelerators in one system, but it also raises big questions about standards, verification and security.
AI and advanced materials. AI is both driving demand for more efficient materials and accelerating the search for them, with AI-driven platforms and automated labs testing candidates far faster than humans could. UK companies are already moving beyond silicon: Paragraf, which produces graphene wafers at scale, has begun manufacturing molybdenum disulphide (MoS₂), which could help make data centre chips more energy efficient. New materials matter for resilience too. At techUK's recent critical materials webinar, panellists warned that the UK is highly exposed on inputs like gallium, germanium and indium. The challenge is turning modelled materials into proven, manufacturable ones, the gap Innovate UK's new National Materials Innovation Programme aims to close.
AI and quantum. AI is already helping quantum technologies develop, from reducing error rates to analysing the data quantum systems produce. At a 2026 techUK discussion on AI and Quantum, participants saw the clearest near-term opportunities in quantum sensing, such as combining quantum sensor data with AI to analyse cancer tissue. But all of this depends on semiconductors. Quantum systems rely on chips at every level, from silicon-based qubits to control electronics and photonics, and participants saw this dependence as a key factor in how fast quantum can scale.
The UK's hand, and the window to play it
The UK is well placed for this moment. We have world-leading strengths in chip design and IP, compound semiconductors, photonics and quantum, a world-class research base, and a growing cluster of companies building novel AI hardware. Government has also signalled real ambition through the £1.1 billion AI Hardware Plan and its commitment to act as an early customer for UK chips.
But there is a catch. These strengths have historically sat in separate silos, with different funding streams, policy teams and communities. These technologies are not innovating in silos, so neither should we.
There is also a sense of urgency. On frontier compute, techUK members have suggested the UK has a window of perhaps three to five years before dominant players set the standards and lock in the technology stack. Once that happens, catching up becomes structurally very difficult.
For government, that means coordinating the AI Hardware Plan, public compute through the AI Research Resource and UKRI’s research funding as one convergence strategy, not running them as parallel programmes.
Nobody builds a chip alone
If there is one lesson from AI convergence, it's that nobody builds the future of computing alone. The next generation of AI hardware will combine designs, materials and expertise from many companies, disciplines and countries. The UK's opportunity is to be the place where they come together and scale.
At techUK, that is exactly the work we are focused on, bringing together our semiconductor, quantum, photonics, AI and frontier compute communities to make the case for the UK's place in the next era of computing. If you'd like to be part of that conversation, I'd love to hear from you.
