Adrian Allen
Part one of two
So, the picture so far: capital, attention and political energy are flowing towards AI at a scale with no real precedent; the built environment's own numbers haven't moved, and the regulatory floor just quietly changed in a way most of the market hasn't caught up with yet.
There is a particular kind of vertigo that comes from sitting in an asset management strategy meeting in 2026 and watching the agenda quietly reorder itself. Eighteen months ago, every board pack led with EPC trajectories, retrofit business cases and the green premium. Today, the opening slide is AI: co-pilots for underwriting, predictive maintenance, automated ESG reporting. These conversations are not wrong to have. The problem is that they are increasingly the only conversation being had, and the built environment's decarbonisation problem has not gone anywhere while everyone was looking at the shiny new thing.
This is a deliberately unfashionable argument: AI is not a substitute for Net Zero strategy in commercial real estate, and in some respects, it is making the underlying problem larger, not smaller. Owners and asset managers who let AI enthusiasm crowd out fabric-first retrofit planning are making a category error, treating a productivity tool as if it were a decarbonisation strategy.
The capital is moving, but not where the emissions are
The scale of capital reallocation towards AI infrastructure is genuinely without precedent. The International Energy Agency's 2026 analysis shows that capital expenditure by the largest technology companies exceeded USD 400 billion in 2025 and is expected to jump by a further 75% in 2026, a level of investment now larger than global spending on oil and natural gas production combined. Electricity consumption from AI-focused data centres surged 50% in 2025 alone, and "AI factories," data centres purpose-built for AI workloads, have more than tripled in capacity in eighteen months.
That is an extraordinary reallocation of capital, engineering talent and political attention, and it sits in direct tension with decarbonisation. The domestic picture matters most for a UK ESG and asset management audience. UK data centres now consume around 5.8% of national electricity generation, close to the 5% threshold that industry analysts identify as the point at which local political and community pushback typically intensifies. Grid connection demand has exploded alongside it: Ofgem's queue of contracted connection offers rose from roughly 41GW in November 2024 to around 125GW by June 2025, and the 140 proposed data centre schemes currently in the planning system are seeking around 50GW of electricity, 5GW more than Great Britain's entire peak demand of 45GW recorded in February 2026. A December 2025 London Assembly report warned that data centres could increase their energy needs in the capital by between 200% and 600% in the medium term, placing further strain on a grid already under pressure from electrification elsewhere. Slough, Europe's largest data centre cluster with 30 to 35 facilities, and Greater London, with over 1,000MW of capacity, sit at the sharp end of this. Six UK environmental NGOs, including Friends of the Earth, wrote to the technology secretary in March 2026 warning explicitly that AI data centre demand could drive up the UK's carbon emissions.
None of this makes AI inherently a climate villain. Hyperscalers (the largest scale providers – Microsoft, Amazon, Google & Meta) are signing genuine long-term renewable power purchase agreements, or PPA’s (long-term contracts to buy electricity directly from a renewable generator), and nuclear offtake agreements (advanced commitments to buy power from a nuclear plant, often signed before construction, to help finance the project). This suggests at least some of this demand growth will be met with low-carbon supply. But the honest summary, in the UK as much as globally, is that AI is not, on the current evidence, a straightforward decarbonisation dividend. It is a large new source of demand competing for the same constrained grid capacity, skilled labour and capital that the built environment retrofit challenge also needs. Every gigawatt of UK transmission capacity absorbed by a data centre campus is a gigawatt not available to electrify a building's heating system: the two are drawing on the same finite pipe.
Meanwhile, the built environment's numbers haven't improved
The built environment remains one of the largest single contributors to UK territorial greenhouse gas emissions, commonly estimated at somewhere between a quarter and over 40% of the total, depending on how embodied and operational carbon are scoped. Unlike AI's emissions trajectory, which is genuinely new and fast-moving, the built environment's problem is old, well understood, and stubbornly unresolved. That is arguably what makes it less exciting to talk about, and precisely why it deserves more airtime, not less.
Three data points illustrate the scale of what remains outstanding in UK offices specifically.
Savills estimates 58% of Central London office stock, by floor area, currently sits below an EPC B rating, and other industry analysis puts the proportion of English and Welsh office stock at risk from a future EPC C requirement as high as 73%.
Central London office rents are diverging sharply along sustainability lines. Savills' Q4 2024 data found BREEAM Excellent or Outstanding buildings achieving rents around 15% higher than lower-rated or unrated stock, consistent with JLL's global finding of a 7 to 12% green premium and a 42-study meta-analysis putting the average certified-building premium closer to 6%.
Q1 2026 Central London market data shows sustainability has moved "beyond preference and into leasing reality." Buildings without a credible improvement pathway now face longer voids and weaker investor appetite, regardless of location or specification otherwise.
This is not a hypothetical future risk. It is a present-tense repricing of the market, and it is happening at the same time as the industry's attention has visibly shifted towards AI tooling.
The regulatory goalposts just moved, and most owners haven't noticed
Here is where currency matters, and where I'd push back on the framing in a number of recent ESG briefings circulating the market. This includes Hollen+'s January 2026 presentation, which anticipated the government's long-delayed MEES review confirming an EPC B minimum by 2030.
That review has subsequently landed, and it isn't quite what was expected. In June 2026, the Government published its interim response to the 2019 and 2021 MEES consultations. The headline change is that the EPC B deadline has moved to 2031, and it now only applies to commercial buildings above 1,000 square metres, approximately 10,760 sq ft. Buildings below that threshold face a materially lower bar, EPC E rather than B. The previously proposed interim EPC C milestone for 2027 has been dropped entirely.
There are two ways to read this. The generous reading is that the government has finally provided certainty after years of delay, and the timeline extension gives owners genuine breathing room to plan retrofit capital expenditure sensibly rather than reactively. The less generous reading, and the one I'd weight more heavily, starts with the scale of what the 1,000 sqm carve-out actually excludes: approximately 85% of UK commercial property sits below that threshold. On the face of it, this looks like a policy that has just softened compliance for the overwhelming majority of the stock, at exactly the moment the market, per the Savills and CBRE data above, is already pricing sustainability credentials into rents and liquidity regardless of what the regulatory minimum requires.
The more useful way to look at it, though, is not simply that most buildings got let off the hook. Smaller assets carry smaller retrofit bills in absolute terms: less roof area, fewer plant rooms, a shorter run of glazing to replace.
The capital outlay required to move a sub-1,000 sqm building from EPC E towards EPC B or better is typically far more modest than the equivalent works on a large multi-let building. That changes the economics of acting early. For owners of this smaller-format stock, the relaxed statutory deadline is not really a reason to defer, because the cost of getting ahead of the market voluntarily is proportionately lower, and the leasing and valuation benefits documented earlier in this article apply just as much to a 5,000 sq ft building as a 500,000 sq ft one.
Occupiers and capital markets are moving faster than the regulatory floor across the size spectrum. Owners who treat the softened MEES timeline as licence to defer capex, whether they sit above or below the 1,000 sqm line, are, in my view, mistaking regulatory minimums for market reality. The two have already decoupled.
This is precisely the kind of granular, fast-moving regulatory development that a generic AI dashboard summarising "ESG risk" will not reliably catch, and that requires an expert reading the primary source, and not just the headline.
The question that leads to is, “What should owners and asset managers actually do about it, and where does AI genuinely fit into that answer rather than distracting from it?”
That's the subject of part two.
Contact Adrian Allen