Central London recorded 2.7 million square feet of office take-up in the second quarter of 2026, a figure that on its own reads as a healthy, active market. Look at what actually moved into that space and a more complicated picture appears. Technology, media and telecoms occupiers accounted for 27% of activity, according to CBRE’s July figures, with banking and finance almost alongside at 26%. AI-related companies accounted for 450,816 sq ft of the quarter’s transactions, including Anthropic’s 158,500 sq ft letting at 1 Triton Square, and more than 705,000 sq ft across the first half. These are genuine commitments, but none of those figures tells an owner how much additional office space the companies will ultimately occupy, when they will occupy it or which buildings might be left behind. A market can look busy and still be heavily influenced by a small number of large transactions; equally, a surge in one fast-growing sector need not mean established occupiers have stopped expanding. The real question for investors, developers and agents is what the composition of demand means for the income and future prospects of individual buildings, rather than which sector won the quarter.
Gross take-up is not the same question as net demand
Before assigning credit to any sector, the more fundamental distinction has to be established, because it changes what every subsequent figure actually means. Gross take-up records space transacted, including pre-lets that may not be occupied for years; it does not say what an occupier is doing with space elsewhere. A company agreeing to lease 100,000 sq ft in a new building while planning to vacate a similar amount in an older one produces a genuine transaction and a success for the new landlord, but not necessarily a net increase in occupied stock. Net absorption measures the change in occupied floorspace within a consistently defined market over a period, allowing for occupation and vacation. It cannot be calculated by taking gross lettings and subtracting a guessed share of relocations, nor can it be inferred directly from falling availability, which can also reflect refurbishment, withdrawal or demolition. Without a comparable occupancy series, the sectoral figures in this article establish who has committed to space, not how much the total occupied stock increased in 2026.
There is nevertheless evidence of occupiers pursuing genuine expansion, and it should not be dismissed simply because relocations feature prominently. Savills’ H1 2026 survey of active Central London requirements above 10,000 sq ft found that 43% of occupiers were seeking to expand their footprint, against 12% seeking to reduce it. Those are proportions of occupiers with requirements, not shares of floorspace already let, and the requirements may change or fail to transact. Another 15% reflected new entrants, occupiers moving out of serviced offices or businesses seeking additional space. Even within that group, the impact on occupied stock depends on what happens next: a fast-growing company leaving flexible space may create an opportunity for the serviced-office operator to refill it, while a business retaining its existing offices and adding another site represents a more direct expansion in its own footprint.
A further complication is the activity that conventional new-letting statistics barely capture. CBRE’s midyear office outlook identifies unusually high renewals and regears where suitable alternatives are scarce. An established occupier might commit to years more in the same building without generating an equivalent new-letting transaction, while another business could sign a major pre-let in 2026 that will not lead to physical occupation until 2029. For a developer, these decisions can underpin financing or affect competition for new schemes; for an investor judging actual current occupancy, they belong to a different period. Signed demand, occupied demand and space returning to the market are connected, but they are not interchangeable.
Finance and professional services remain the structural base
Financial and professional services occupiers have not ceased to underpin the market because AI leasing has attracted so much attention. CBRE’s 26% Q2 share for banking and finance was only one percentage point behind the technology, media and telecoms grouping, while Savills found that active financial and banking requirements rose by 30% from the start of 2026 to 5.8 million sq ft at the end of June. Professional-services requirements reached 3.58 million sq ft, an increase of 23%. These figures concern space being sought, not completed deals, but they undermine any reading of the market as depending on AI alone. They also show why treating an active requirement as though it were already a tenant signed to a particular building can distort a development appraisal.
Completed transactions provide the other half of the picture. Savills recorded 549,000 sq ft of professional-services lettings across 24 Central London deals in H1, with activity heavily influenced by a major law-firm pre-let. Its September 2026 market report showed that one 161,000 sq ft commitment by Mishcon de Reya represented almost 30% of all Central London take-up in August. A monthly sector chart can therefore be dominated by the property decision of one large occupier even where the broader market contains many smaller requirements. Established businesses may provide important credit and occupational depth, but their expansion should not be assumed from the size of one transaction any more readily than that of an AI company.
A still more striking example emerged on 24 September, when J.P. Morgan Asset Management announced a 465,000 sq ft pre-let to quantitative trading firm Jane Street at One Spitalfields. The existing occupier is due to vacate in 2027, redevelopment is scheduled for completion in 2029, and the agreement was described as the City’s largest off-plan pre-let in more than a decade. For the owner, a major pre-construction commitment can alter the risk of delivering the project; for the wider occupational market, it is not 465,000 sq ft of extra space occupied in September 2026. The scale of the transaction also reinforces how misleading it would be to treat financial services as the uneventful background against which technology alone is changing demand. An investment decision based on such deals must account for their delivery dates, occupier expansion plans and the future use of the premises vacated.
Nor does a recognisable sector automatically make a covenant predictable. Financial institutions and professional partnerships also consolidate, merge, change workplace strategies and renegotiate accommodation, while the commitments of established organisations may involve substantial landlord incentives or bespoke fit-outs. What matters to the property’s value is the enforceable lease, effective rental income, credit support, ability to re-let and realistic alternative occupiers, not the reassuring or fashionable label attached to the tenant’s industry.
AI’s leasing activity is real, but treat the interpretation carefully
The acceleration of AI-related office leasing in Central London is documented rather than speculative. CBRE recorded 450,816 sq ft in Q2 and 705,371 sq ft across H1 2026, against 317,979 sq ft for the whole of 2025. Anthropic and OpenAI were responsible for two of the quarter’s four largest deals. These figures belong within, rather than in addition to, the wider technology sector total. AI’s Q2 activity was roughly one-sixth of CBRE’s overall quarterly take-up, not a third; confusing the half-year AI total with a quarterly share exaggerates the degree of immediate sector concentration. The growth is nevertheless rapid enough to matter for particular submarkets and for buildings capable of accommodating large technology occupiers.
The picture also depends on what an agency is measuring. JLL’s Q2 analysis put overall Central London take-up at 2.6 million sq ft, with technology, media and telecoms representing 33%, compared with CBRE’s 2.7 million sq ft and 27%; Savills reported 2.4 million sq ft under its own market coverage. These are separate research series, and differences in geographic coverage, classification and transaction recording need to be understood before their percentages are compared. The narrower conclusion survives: well-funded AI occupiers have become important bidders for London offices, but a single agency’s quarterly sector split is not proof that UK office demand has changed permanently or uniformly.
The commercial exposure also varies materially between AI occupiers. Some are part of large, well-capitalised international groups; others have short trading histories and forecasts of rapid expansion. A landlord may be asked to provide expansion rights, sizeable works or fit-out expenditure to secure a prominent tenant, and an impressive reported headline rent may not reflect the effective cash return over the committed term. The correct underwriting question is not whether AI demand is real, but whether the particular tenant’s obligations and the building’s alternative letting potential justify the capital being committed. A specialised layout that works exceptionally well for the first occupier may be costly to adapt if its future space needs prove different from the growth trajectory assumed when the lease was signed.
Where government and public-sector demand fits, carefully stated
The regional story is not simply a smaller version of London. In Savills’ six major regional office markets, public services, education and health accounted for 362,023 sq ft, or 22%, of H1 2026 take-up, while professional services contributed 336,066 sq ft, or 21%. The first grouping includes health and education and should not be described as government-only demand. The Government Property Agency’s 114,967 sq ft letting at Havelock in Manchester was the largest deal of the regional half-year, showing how a single public-sector commitment can make a substantial difference to a city’s leasing statistics without demonstrating equivalent expansion across all its office locations.
The longer-term Manchester Digital Campus programme is even more revealing. It is planned to bring approximately 8,800 civil servants together in around 900,000 sq ft of purpose-built space, with the government targeting completion for occupation in 2032. That is not an H1 2026 lease take-up statistic, and consolidation from existing government buildings must not be confused with 8,800 entirely new jobs or 900,000 sq ft of net additional occupancy. For a developer or investor, the campus can still have significant implications for nearby demand, competing accommodation, future releases of older office space and the ability of existing landlords to retain public-sector occupiers. The financial value of an announced estate programme depends as much on what it replaces as on what it creates.
Public-sector and regional professional-services requirements can also favour the best-quality buildings while leaving a difficult market for secondary stock. Savills recorded prime and Grade A offices taking 61% of H1 regional lettings, against a regional availability pool in which around half of the space was secondary. A major public-sector letting can validate a new scheme without rescuing the wider stock of ageing offices in the same city, just as strong City of London requirements do not guarantee occupiers for every older building outside the prime core.
Reading building preference by occupier type
Different occupiers have genuine operational preferences, but those preferences carry more information than a simple list of which industries favour which districts. Financial institutions may prize established client-facing locations and resilient building services; professional firms may need large, flexible floors that can accommodate changing teams; technology businesses may value connectivity, recruitment reach and room for growth; and public-sector bodies may have particular procurement, security and estate-management requirements. None of these categories confines the occupier to one building type. What matters is whether the property’s layout, performance and overall cost meet the business need, and whether the building could appeal to another sector if the current requirement disappeared.
Nor should the flight to quality be mistaken for a market in which every transaction involves a newly constructed building. CBRE reported that secondhand accommodation represented 64% of Q2 2026 Central London take-up, against 29% for newly completed space and 7% for pre-lets. Secondhand means previously occupied; it does not necessarily mean poor-quality, obsolete or unrefurbished. A well-repositioned building can compete successfully for demanding tenants without being new, while a newly delivered project can still struggle if its rent, floorplate or location does not suit the requirements available. For owners considering redevelopment, the question is not merely whether new offices are popular but whether refurbishment can deliver a competitive product at a lower capital cost and with a wider future tenant pool.
Why the same building type can serve very different tenants
The buildings sought by AI companies, law firms, financial businesses and public bodies are not necessarily separate pools. A large, adaptable building with modern services, good transport connections and floors capable of different fit-out configurations might attract proposals from all of them. That can improve the owner’s prospects, but comparing offers requires more than comparing headline rents. One tenant may provide strong financial backing but demand expensive adaptation, another may offer attractive rent growth but need breaks or expansion options, while a pre-let may support a development long before the occupation starts and rent becomes payable. An owner selecting between them is making a judgement about the value, timing and security of income alongside the ability to recover and re-let the property if the original plan changes.
For investors, the most resilient building may therefore be the one capable of serving several plausible occupier types without extensive reconstruction, not the one tailored most completely to the sector currently dominating leasing news. Apparent diversification across multiple addresses is of limited protection where the tenants are exposed to the same financing cycle or corporate estate strategy. Equally, a technology occupier supported by a substantial balance sheet may be a more dependable covenant than a familiar professional brand with little independently recoverable credit support. Industry labels help describe the market; they cannot substitute for the lease and property analysis that determines the investment return.
What this means for each part of the market
For landlords and developers, the distinction between signed commitments, genuine expansion and redistributed occupation determines the risk of bringing another scheme to market. A pre-let secured in 2026 may support construction expenditure without creating occupied floorspace until several years later; a tenant moving out of a secondary property may support a prime development while leaving its former landlord with a costly refurbishment or vacancy. The apparent success of one owner can therefore create a capital problem for another. The economically useful measure is not simply the quantity of requirements circulating among agents, but how many are credible, when they will turn into occupation, what existing space will be released and what rent can be achieved after incentives and building expenditure.
Agents face a similarly demanding interpretation problem. A company’s interest in several buildings may appear in market conversations long before one firm commitment emerges, while requirements can be delayed, reduced, withdrawn or satisfied by a renewal. CBRE’s August 2026 office figures recorded take-up 41% below its long-term monthly average, while Savills found that one law-firm pre-let dominated its measure of August activity. These are not inconsistent with substantial requirements being active or with significant transactions being agreed in September. They show why the timing and concentration of deals matter more to forecasting than a simple run of monthly sector rankings.
For property investors and lenders, the danger is to capitalise short bursts of leasing momentum into assumptions of sustained rental growth across an entire portfolio. Recent transaction evidence is most relevant where the property shares the location, specification, lease economics and alternative tenant market of the buildings being let. Elsewhere, headline Grade A rents can move upward while a secondary asset requires considerable expenditure merely to remain competitive. A scheme underwritten on a pipeline of occupier requirements should be tested against the possibility that some are relocations, some are years from occupation and some are competing expressions of interest rather than independent sources of new demand. For occupiers, meanwhile, competition for genuinely suitable premises can be intense even when an apparently large quantity of offices remains available elsewhere in the market.
The real measure of a healthy office market
Total leasing volume, taken alone, answers the wrong question. The more useful assessment asks who is adding to the amount of office space they occupy, who is moving, which buildings are gaining and losing occupation, and whether the resulting income can sustain the capital invested in those properties. The evidence through September 2026 does not describe a UK market newly dependent on AI, nor one in which established finance and professional services have become marginal. Technology has produced a concentrated run of major London lettings; financial and professional firms remain important sources of demand, including commitments to future buildings; and public services have contributed meaningfully to particular regional markets. These facts show real occupational activity, but they are not a substitute for a consistent measure of net occupied stock. The central risk for a landlord or developer is mistaking the volume and visibility of leasing deals for durable demand available to the next building brought to market — and overlooking the premises and liabilities left behind when an occupier chooses somewhere else.
This article provides general UK commercial-property and office-market analysis, not investment, valuation, financial or legal advice. Market figures, occupier requirements and lease arrangements depend on the location, period and research methodology. Landlords, developers, investors, agents and occupiers should obtain appropriate specialist advice before committing capital or entering into property transactions.
