That leaves room for legitimate scrutiny—but also for assumptions that the available evidence does not support. Microsoft occupies the campus, for example, but it is not identified as the developer. Nor is there evidence yet that the facility would necessarily host AI systems or create a particular data-security risk. The most useful debate is therefore not whether all data centres are good or bad, but whether this project can answer measurable questions before planning permission is considered.
What is actually being proposed
The emerging proposal would replace three Microsoft-occupied office buildings at the Thames Valley Park campus. Consultation reporting identifies the landowner, FH Trustees, as pursuing the redevelopment and says Microsoft is a tenant rather than the developer.
That distinction matters. The campus name may invite assumptions that this is a Microsoft or Azure expansion, but neither the future operator nor any prospective tenants have been identified. The intended mix of workloads—AI training, AI inference, conventional cloud services, colocation or something else—also remains unknown.
The project team has indicated that it intends to submit a planning application in autumn 2026, with construction expected during 2027 to 2031 and operation from 2031. These are promoter targets, not approved dates. At the stage reflected in the consultation material, no planning permission had been established.
The proposal therefore should not be treated as a finished design. Project-specific assessments of noise, air quality, traffic, heat, landscape and ecology are expected to become important if a full application is submitted.
Natural gas is central to the initial power strategy
The most consequential confirmed detail is the proposed use of on-site fuel cells running on natural gas while additional electricity-grid capacity is brought forward.
The project team says the fuel cells would generate electricity through a chemical reaction rather than conventional combustion. That technical distinction may affect efficiency and pollution performance, but it does not establish that the system is low-carbon. Natural gas remains a fossil fuel, and the published information is not sufficient to verify lifecycle-emissions claims.
Several variables are still missing:
- The type, capacity and efficiency of the fuel cells.
- Expected annual and peak gas consumption.
- Direct greenhouse-gas emissions and relevant air pollutants.
- The emissions baseline against which any claimed reduction would be measured.
- The size and delivery date of the eventual grid connection.
- How much normal operation would continue to depend on gas after that connection becomes available.
Biomethane, hydrogen and offsets have been raised as possible future options, but possibilities are not the same as contracted fuel supplies or enforceable planning commitments. Hydrogen availability, production method and associated emissions would also matter. Merely stating that equipment could use a different fuel later does not demonstrate when—or whether—that transition will occur.
For residents and planning authorities, the practical question is not whether fuel cells sound cleaner than conventional generators. It is how many tonnes of emissions the proposed system would produce under realistic operating conditions, including any continued gas use after more grid power becomes available.
The water question has not been quantified
The cooling system has been described as a closed water loop that would carry server heat to rooftop air-cooled chillers, where heat would be rejected to the outside air. Such a description suggests that water would be recirculated rather than continuously consumed in an evaporative cooling process.
However, the available project information does not quantify what claims of “minimal” replacement water would mean. There is no published litres-per-day figure or complete water balance covering initial filling, routine losses, maintenance and other operational uses. Indirect water consumption associated with producing electricity or fuel has not been established either.
Consequently, two confident conclusions should be avoided. It is premature to claim that this site would necessarily place an extreme burden on local water supplies, but it is equally premature to dismiss water concerns on the strength of the phrase “closed loop.” A planning application should provide peak and annual demand figures, identify potable and non-potable sources and explain performance during hot weather or equipment failures.
One figure used in opposition to the development also requires care. A cited estimate of £25 billion concerned housing development halted by general water scarcity; it was not a calculation of health and water losses caused directly by data centres. Data-centre demand may still aggravate wider supply constraints, but the figure cannot accurately be presented as a project-specific damage estimate.
Heat concerns are plausible but not yet site-specific
A 72MW IT facility would generate substantial heat that must ultimately be moved outside the buildings. That makes questions about rooftop heat rejection, noise and local temperature effects reasonable subjects for environmental assessment.
It does not, however, establish the temperature effect at Thames Valley Park. A research preprint cited in the wider debate estimated an average land-surface-temperature increase of about 2°C after AI data centres began operating. It did not model this proposed facility, and the scheme has not been established as an AI data centre.
Site geometry, equipment, operating load, weather, landscaping and heat-rejection design would all affect local outcomes. An average found across other sites should not be converted into a prediction for Reading without a project-specific model.
The appropriate response is to require evidence: heat-plume modelling, seasonal assumptions, rooftop-equipment specifications and an assessment of effects on nearby properties and habitats. Generic research can justify asking those questions, but it cannot supply the answers.
Data security depends on the workload, not the campus name
Some objections have suggested that a data centre operated by, shared with or owned by an international company would place its activities beyond UK data-protection enforcement. That is too broad. UK GDPR can apply to organisations outside the UK when they offer goods or services to people in the UK or monitor their behaviour.
Foreign ownership also does not, by itself, establish that UK personal data would be insecure. The relevant issues would include who acts as controller and processor, what information is stored, where it is transferred, how it is encrypted, who can obtain access and what contractual and legal safeguards apply.
None of those arrangements can be assessed yet because the operator, tenants and workloads remain unknown. It is therefore unsupported to assert either that the proposed facility would create a particular public data-security danger or that it would be inherently secure.
For Windows users and enterprise administrators, the practical lesson is that the physical location of servers is only one part of data governance. A building on a campus associated with Microsoft should not automatically be assumed to host Azure, Microsoft 365 or Windows workloads. Security and residency claims must be tested against the actual service architecture and contracts.
Biodiversity claims need proper context
The project team promises a minimum 10% Biodiversity Net Gain. That may produce a better ecological outcome than leaving losses uncompensated, but 10% is broadly the statutory minimum for qualifying major developments in England. It is not, on its own, an exceptional voluntary benefit.
The meaningful evidence would be the habitat baseline, the statutory metric calculation and the proposed delivery plan. Those details would show whether the gain is expected through habitat improvements on the site, off-site biodiversity units, credits or some combination of these measures.
Until that information is published, neither supporters nor opponents can reliably judge the quality, location or durability of the promised gain. The percentage alone says little about which habitats would be affected or how long replacements would take to mature.
Jobs and investment are forecasts, not guarantees
The promoter forecasts up to 245 construction jobs, at least 115 full-time-equivalent operational roles and more than £200 million in construction investment. These figures help indicate the proposed scale, but they remain promoter projections. Their methodology and assumptions have not been independently demonstrated in the available material.
There is, nevertheless, independent evidence of weakness in the Thames Valley Park office market. A Wokingham Borough Council employment review recorded Microsoft vacating two of five campus buildings, Shell vacating four buildings and three vacant buildings at the Oracle campus.
That provides a credible context for considering redevelopment. It does not prove that a data centre is the best alternative use, nor does it establish that projected jobs will be local, permanent or additional. A planning case should distinguish temporary construction work from long-term roles and clarify whether operational employment estimates include contractors or jobs transferred from elsewhere.
Similarly, the existence of AWS, Azure, Google and other computing facilities in England proves neither that another 72MW facility is needed nor that the region already has excess capacity. A convincing need argument would require evidence about demand, resilience, grid constraints and suitable regional capacity.
What a planning application needs to settle
The strongest objections are those framed as testable questions rather than predetermined conclusions. A full application would need to make several points materially clearer:
- The applicant’s exact legal identity and its relationship with FH Trustees.
- The future operator, where disclosure is possible, and the anticipated classes of workload.
- Fuel-cell specifications, gas use, emissions and air-quality effects.
- The grid-connection timetable and the expected long-term role of natural gas.
- A quantified operational water balance.
- Noise, heat-plume, traffic, landscape and ecological assessments.
- The biodiversity baseline and the mechanism for achieving at least 10% net gain.
- The assumptions behind employment and investment forecasts.
Some commercially sensitive details may not be available at an early stage. That does not remove the need for enforceable performance limits. Planning conditions could ultimately be more important than broad assurances, particularly for noise, emissions, operating equipment and environmental mitigation.
A debate that should become more precise
Thames Valley Park’s vacant offices make redevelopment understandable, and digital infrastructure can bring investment and long-term employment. But the proposed initial dependence on natural gas, the absence of quantified water demand and the lack of project-specific environmental assessments give residents substantive reasons to seek more information.
At the same time, several prominent arguments go beyond the evidence. Microsoft is not identified as the developer; the facility is not confirmed as an AI site; overseas corporate involvement does not make UK data law impossible to enforce; and generic water or heat studies cannot be treated as forecasts for this location.
The scheme’s targeted 2031 opening leaves time for a more disciplined assessment. The decision should turn on published technical specifications, measurable impacts and enforceable commitments—not the familiarity of the Microsoft name, optimistic promotional language or worst-case assumptions imported from unrelated data centres.