Few forces are reshaping commercial real estate as rapidly or as fundamentally as artificial intelligence. The evidence is already plain to see: AI businesses are expanding aggressively into London's office market, data centre development is surging at a pace the infrastructure network is struggling to match, and the characteristics that make a building valuable are changing. For clients, investors, and their advisers, understanding what is driving these shifts and what they mean for individual assets and transactions is becoming an essential part of commercial decision-making.

As with any major technological transition, the benefits will not be distributed evenly or arrive quickly. The pace at which businesses integrate AI, and the degree to which productivity improvements translate into economic expansion, will be the key determinants of which parts of the market thrive and which come under pressure.

London's Office Market: Quality is Everything

A clear pattern has emerged in London's office market: demand is concentrating at the top of the quality spectrum. Grade A buildings, those that are newly constructed or substantially refurbished, offering strong environmental credentials, modern technical infrastructure, and high-quality communal and amenity space, are commanding strong interest from occupiers across all sectors. Lower-quality stock, by contrast, is finding it increasingly difficult to attract lettings.

The numbers bear this out. Overall office take-up in London reached 2.4 million sq ft in the first quarter of 2026, but that aggregate figure conceals a sharper story: 73% of all lettings involved Grade A space. Pre-letting activity reached 932,000 sq ft, and Grade A vacancy in the City core fell to 1.1%, a five-year low. In the West End, pre-lets accounted for 41% of total take-up.

AI Companies Are Transforming Leasing Activity

One of the most striking developments of the past 18 months has been the volume and pace of office leasing by AI businesses. In the first six months of 2026 alone, AI occupiers took more than 660,000 sq ft across London, exceeding their total for the whole of 2025. Since the beginning of last year, cumulative commitments have approached 1.2 million sq ft, with analysts projecting the full-year 2026 figure at around one million sq ft, approximately double 2025 and four times the 2024 level.

The scale of individual transactions reflects the ambition of these businesses. Anthropic secured 158,000 sq ft at One Triton Square; Databricks committed to 137,000 sq ft in Fitzrovia for its EMEA headquarters; and OpenAI took 88,500 sq ft at Regent Quarter in King's Cross. Together these three transactions accounted for nearly 60% of AI office take-up in the first half of the year.

Importantly, this is no longer a story about a handful of large model developers. Legal AI platforms, enterprise software companies, and cloud infrastructure providers are all actively seeking space, competing with banks, law firms, and hedge funds for the same limited supply of best-in-class buildings. With active demand across London sitting at around 11.4 million sq ft, well above the long-term average of 9.5 million sq ft, and prime City vacancy at just 0.3%, the gap between demand and available supply is significant.

AI occupiers are also changing the specification that buildings are expected to meet. High electrical capacity, specialist cooling, enhanced ventilation, and on-site data infrastructure are increasingly standard requirements. Landlords with technically capable buildings are winning lettings; those whose assets cannot accommodate these demands are being passed over, regardless of location.

The Data Centre Surge

The computing infrastructure required to develop and operate AI systems is generating extraordinary levels of investment in data centres. Global spending on AI-driven data centre capacity is forecast to reach £114.2 billion by 2028, at a growth rate exceeding 30% per year.

The pressure on UK infrastructure is particularly acute. Applications for grid connections rose by 460% in the first half of 2025, and in some locations the wait for a suitable connection now extends to between 12 and 15 years. Land with proximity to existing power infrastructure has been rapidly repriced as a result, and construction costs across the EMEA region have increased by close to 10% year-on-year. Across the region as a whole, contracted data centre capacity has grown 12% year-on-year, with occupancy rates at 91% and around a quarter of new capacity pre-let before opening.

A Balanced View: The Downside Risks

The opportunity created by AI is real, but so is the potential for disruption. Market analysis points consistently to one conclusion: AI is more likely to produce a wider range of outcomes across the real estate sector than to lift all assets uniformly.

In the central scenario, AI adoption proceeds gradually. Businesses direct early efficiency gains towards cost reduction rather than growth, which keeps a lid on aggregate office demand and widens the performance gap between prime and secondary assets. In a more pessimistic scenario, reminiscent of the dot-com era, a wave of overinvestment is followed by financial disruption, setting back the broader economic benefits of AI and placing stress across the market before a recovery takes hold.

The underlying message for clients is clear: AI is not a straightforward demand multiplier. It is reshaping which businesses succeed, where they choose to locate, and what they need from property. Those holding or acquiring assets need to think carefully about quality and adaptability, not just aggregate market trends.

What This Means in Practice

For clients active in the commercial real estate market, whether as occupiers, landlords, investors, or lenders, a number of practical considerations now require attention.

Leasing documentation must address technical requirements explicitly. Conventional office leases were not designed to accommodate the technical demands that AI occupiers bring. Where a letting involves substantially elevated electrical capacity, specialist cooling plant, or bespoke data infrastructure, the lease must deal with these matters expressly. In practice, this means agreeing at heads of terms stage on the landlord's obligation to provide and maintain adequate power supply; the tenant's entitlement to carry out technical fit-out works; how the costs of enhanced building services are treated within the service charge; and the scope of the tenant's reinstatement obligations at lease end. Left unaddressed, these become the source of significant disputes. Landlords should also carry out an honest assessment of whether their building's power infrastructure can realistically meet AI-grade requirements before marketing to this occupier class.

Technical due diligence is now as important as legal due diligence. The factors that determine a building's appeal to AI-related occupiers, principally its electrical capacity, grid connection status, cooling capability, and data connectivity, are also material to its investment value and long-term lettability. Purchasers and investors who confine their due diligence to title, planning, and environmental matters risk overlooking features of the asset that will directly affect void risk, rental growth, and realisable value on exit. Valuers and lenders should equally reflect technical adequacy in their assessments.

Data centre transactions call for specialist legal input. Grid access is the most pressing issue: given connection timelines of 12 to 15 years in some locations, the contractual framework governing the connection process needs to be established at the earliest opportunity. Early advice is also needed on planning matters, including whether proposed use requires a specific consent; construction and fit-out contracts, where cost escalation risk is significant; and the structure of colocation or occupational arrangements, which are governed by their own specialist legal framework.

Lease flexibility requires careful structuring on both sides. AI businesses can grow materially within a relatively short period, making break rights, expansion options, and alienation provisions governing assignment, subletting, and sharing of occupation more commercially significant than in a conventional letting. Tenants need the ability to contract or expand space as circumstances require; landlords need to preserve the income security that supports asset value. These competing interests need to be resolved in heads of terms rather than left to negotiation once solicitors are instructed.

The gap between prime and secondary is structural, not cyclical. The forces driving occupiers towards Grade A space are not short-term. Owners of secondary assets should form a clear view now on whether the capital investment required to reposition a building is commercially viable, rather than waiting for market conditions to improve. For lenders, technical adequacy should be incorporated into asset quality assessments alongside location, unexpired lease term, and tenant covenant.

How Edwin Coe Can Help

Edwin Coe LLP's commercial real estate team works with occupiers, landlords, developers, and investors across the full range of transactions, including office lettings, investment acquisitions, and development projects. If you would like to discuss any of the issues covered in this article in the context of your own portfolio or transaction, please get in touch with a member of the team.