The survey responses implied a staged transition in practices over several years.
Less than two years later, evidence emerging from Cambridge suggests this transition has proceeded at a materially faster pace than anticipated.
Our latest Bidwells Cambridge Offices & Labs Market Databook provides empirical confirmation. AI-related activity is visible across the full spectrum of occupier requirements, from application-layer tools and automation to robotics, clinical discovery laboratories and infrastructure-heavy computing environments. In Cambridge, AI is increasingly embedded in physical research processes, specialist infrastructure and IP-heavy workflows.
Depth and Dispersion of Adoption
Business space requirements from AI specific companies now span the full stack of operations, with 56% categorised as application-layer activity, 27% hardware-related, and 17% infrastructure-focused, reflecting both breadth and depth within the AI sector.
However, while the reach extends beyond purely technology-focused companies, the depth and nature of adoption within individual organisations varies considerably. While the lines are increasingly blurred between technology and biotech, and our analysis of current office and lab requirements finds that in some ‘non-tech’ organisations, AI is embedded as core research infrastructure, integrated into experimental design, data generation, and workflow execution, often in ways that cannot easily be replicated by generic, off-the-shelf agents. In others, the use of Ai remains more limited or exploratory. Distinguishing between substantive integration and more superficial positioning is increasingly necessary when interpreting both occupier strategy and investor narratives.
Nonetheless, the acceleration in adoption is reflected in market composition. Science and technology companies accounted for 89% of all office and laboratory take-up in Cambridge in 2025, up from 68% in 2020. This represents not incremental change but a structural reallocation of demand. For example, H2 saw ARM committed to take up over 95,000 sq ft of space, the largest single transaction in 3 years, which represent neither incremental change nor displacement of scientific activity. At the same time, AstraZeneca committed to extend its presence in the cluster with a new 100,000 sq ft building.
The growth of AI-enabled life sciences, digital biology, and platform-driven research has materially reshaped Cambridge’s leasing market. While there are concerns about AI-driven disintermediation, demand in Cambridge is increasingly concentrated among businesses where AI complements specialised science, proprietary data, and infrastructure, rather than serving as a direct substitute. These occupiers build experimental capabilities and data assets that are difficult for horizontal AI platforms or agents to commoditise, reinforcing the durability of their space requirements.
Capital Context and Substitution Risk
The broader capital context adds an important dimension. Equity markets in recent weeks have reflected concerns over the durability of segments of the enterprise software model, given the capacity of advanced AI systems to automate coding-intensive tasks and compress traditional software value chains. In that environment, differentiation between AI-substitutable and AI-complementary business models becomes critical.
Cambridge’s cluster structure is relevant here. Its concentration in deep technology, healthcare R&D, and AI-enabling hardware suggests greater exposure to AI as a complement to domain-intensive, IP-rich science rather than as a substitute for core outputs. Value in these segments is closely tied to proprietary data, regulatory pathways, experimental capability, and specialist infrastructure, while more conventional software activities may face greater substitution risk.
Academic Foundations and Innovation Dynamics
More fundamentally, the cluster’s academic foundations are significant. Frontier research environments generate paradigm-challenging inquiry that extends beyond the training data of AI systems. While current AI systems are powerful tools for pattern recognition and acceleration, they do not independently reframe underlying research questions. The presence of university-led science and spin-out activity therefore reinforces a model in which AI augments discovery rather than replaces it.
The impact of this academic depth is clear in the Cambridge ecosystem. 47% of current laboratory floorspace demand come from university spin-outs, with most, but not all, originating from Cambridge itself. Proximity to frontier research and early-stage capital appears to accelerate convergence dynamics, producing a concentration of smaller, rapidly scaling, data-intensive ventures.
Software, Hardware and Spatial Implications
The interaction between software and hardware further strengthens this dynamic. AI-enabled design, simulation, and control systems are increasingly narrowing the boundary between computational and physical research processes. AI-enabled design, simulation, and control systems are increasingly narrowing the boundary between computational and physical research processes. At the same time, AI is accelerating the development of hardware itself, from initial design through to prototyping and testing.
Robotics provides one illustration, but similar patterns are emerging across laboratory automation, advanced instrumentation, and healthcare R&D equipment, and other sectors. As AI both drives and becomes embedded within new physical systems, demand is anticipated to extend beyond office space toward laboratories and other infrastructure-intensive facilities, with evidence already seen in some current mid-tech space requirements.
Next-Stage Convergence
In 2023, convergence was framed as a forthcoming transition in research practice. In Cambridge, it is now measurable in occupier composition, capital allocation, and leasing activity. Convergence has therefore moved from an anticipated shift to an observable market structuring driver.
Its evolution will continue rapidly, with opportunities emerging across the real estate spectrum that cannot yet be fully anticipated. As AI agents evolve into a new interaction and orchestration layer for enterprise systems, the assets likely to prove most resilient are those anchored in irreplaceable data, experimental platforms and regulated workflows; precisely the areas where Cambridge’s demand is deepest. Crucially, the cluster’s strong academic foundations will remain central to enabling transformative advances.
