What AI companies want from buildings: lessons from Oxford North

There is a useful paradox at the heart of designing for the AI sector. As the hosts of The Urban Forecast podcast put it, reflecting on conversations with those building for this market: "The greatest thing about an AI lab is it's just somebody with a laptop. But you do need the power. You do need the data." That tension between simplicity at the human scale and enormous demand at the infrastructure scale is reshaping what developers and architects have to deliver.

Oxford North is one of the most direct attempts in the UK to address it. A one-million-square-foot mixed-use innovation district on the ring road at the edge of Oxford, it is backed by St John's College, Ontario Teachers' Pension Plan and Stanhope. The project has hybrid consent for offices, labs, a hotel and a two-acre public park, alongside around 480 homes. Its ambition is not to replicate the science park model but to overturn it.

Simon Ruck, Managing Director at Oxford North Ventures, is direct about what that model gets wrong. Some science parks, he says, "wrap cars around individual buildings and people arrive at work, head up into their office or into their labs and go again at the end of the day and they've not really connected with anyone. This turns that whole philosophy on its head." Oxford North is designed as an urban district. Collision, not commute, is the point.

Power, data and cooling

The first thing AI-focused occupiers ask about is infrastructure. The current wave of AI development is driving significant new requirements around power, cooling and physical security. These are not afterthoughts that can be retrofitted into standard commercial space; they need to be resolved at the master-planning stage.

Ruck is confident Oxford North is in a good position. "We're in a good place in terms of our power solution right now for the million square feet," he told the podcast. He also notes that the technology is shifting: "There are solutions coming where some of the data center requirements are starting to become less power hungry." The trajectory matters to developers making long bets on infrastructure.

Campus flexibility and the hyperscaler problem

The second challenge is scale uncertainty. The fastest-growing AI companies, including hyperscalers, cannot reliably forecast their own footprint. They want optionality baked into the building offer. As the podcast hosts relayed from conversations across the market: "What a lot of businesses want now is campus-style propositions because they don't know, particularly with hyperscalers, how much they're going to need in the end. They don't want to just take a building; they want to know they've got a building where they can take another floor, a lab, and bring an ecosystem around them."

Oxford North's structure, with multiple buildings across a phased district rather than a single monolithic block, is well suited to this. Occupiers can start small and expand within the same ecosystem, without moving off-site and disrupting the relationships they have built.

Designing for uncertainty

Jonathan Kendall, Partner at Fletcher Priest Architects and one of the designers behind Oxford North, frames the core design challenge as one of managed flexibility. "Small can be next to large; small might grow into large," he says. The practice has thought carefully about what he calls "lab-enabled offices or office-enabled labs": buildings whose structure and services can accommodate different uses without requiring fundamental reconstruction.

The first-phase buildings at Oxford North were originally designed pre-Covid for general employment use. They were then adapted as life science wet labs during that sector's boom. They may now serve AI occupiers. That kind of serial adaptation is only possible if the original design decisions left enough room to manoeuvre.

But Kendall is clear-eyed about the limits of flexibility. "There's this whole thing about designing for uncertainty. The danger is that you pay a massive premium for an overly flexible scheme and you kill the scheme. It's trying to work out that sweet spot on how much flexibility is the right amount to design in." Over-engineering for every future scenario is its own form of failure.

Amenity as strategy

The fourth requirement is less technical but equally important: the physical environment has to attract and retain talent. Proximity to Oxford University, which Ruck describes as "number one in the world for computer science", gives the district a structural advantage. But the site itself has to do its share of the work.

Oxford North's approach goes beyond standard amenity provision. The Redhall building includes large-format collaboration spaces and a cafe. The two-acre park gives people a reason to move between buildings and linger outside. Olafur Eliasson has created his first permanent UK artwork for the district. Ruck is unambiguous about the intent: "It can't just be a case of pizza Thursdays... if you create more of a 24/7 innovation district, that's the transition point."

Early signals

Demand so far suggests the approach is landing. An AI hackathon on site drew around 50 people. An event with a Google AI team is also in the pipeline (note to editor: the exact event name was unclear in the audio; please verify before publish). Ruck's summary of the mood is straightforward: "The demand and interest in AI has never been greater."

What Oxford North represents, taken together, is a working hypothesis about what the UK's innovation districts need to become. Not car-park science parks, not anonymous business campuses, but genuinely urban places with the infrastructure grunt to support serious AI work and the density of people and amenity to make that work worthwhile. The hypothesis is being tested in real time.