Can Data Center Market Intelligence Be Trusted?

Data center market intelligence increasingly sits between a site announcement and an investment decision. The map may look precise: a pin, a megawatt figure, a status and a delivery year. But precision of presentation is not the same as quality of evidence.

That is the question I want to address here. Can data center market intelligence be trusted when projects change faster than conventional datasets, capacity definitions vary by operator and much of the decisive power information remains conditional?

My answer is yes—but only when the record keeps the claim, the source and the uncertainty together. A map without provenance is a picture. A market-intelligence system should be an auditable chain.

Start with the decision, not the pin

A developer, investor, operator and end user can all open the same map and ask different questions. The developer may want to benchmark a site. The investor may be testing whether a market is becoming oversupplied. The operator may need capacity by a specific quarter. The end user may be looking for 20 MW of deployable IT load rather than a 200 MW campus vision.

That is why our market-intelligence workspace does not treat a project record as a single fact. Location, phase, power, status, timing and source need separate fields because they answer separate decisions. The public coverage view is useful for orientation; the work begins when a team asks what each record actually proves.

The power context makes this more urgent. The IEA’s April 2025 Energy and AI analysis projects global data-centre electricity use at roughly 945 TWh by 2030 in its base case. More demand means more announcements, more grid applications and more competition for the same substations. It does not mean every announced megawatt will be delivered.

The provenance test for data center market intelligence

I use a simple test before relying on a record: can I move backwards from the conclusion to the original evidence without losing the meaning of the claim?

A primary operator announcement may prove that a facility exists, but not that the full campus is live. A planning document may identify a site and maximum electrical envelope, but not a contracted connection. A utility map can show topology, but not spare capacity. A transaction release may prove a change of ownership while leaving phase-level delivery unchanged.

The PowerlandMap methodology separates reported facts from estimates and inferred readiness for precisely this reason. The distinction is not academic. It determines whether two sites can be compared at all.

The European Union is moving in the same direction on operational evidence. The Commission’s data-centre energy-performance framework requires reporting for significant facilities and creates a European evidence base for energy and water performance. In 2026, the Commission also described the minimum-performance standards work as building on mandatory reporting above 500 kW. That does not solve project-level diligence, but it makes a wider point: comparable markets require comparable definitions.

First check: what kind of megawatt?

A campus can be reported in IT load, utility connection capacity, gross electrical capacity, phase capacity or long-term masterplan capacity. These are not interchangeable. I covered the practical distinctions in Data Center Capacity Is Not One Number.

For data center market intelligence, the safe approach is to preserve the original capacity basis. If the source does not specify it, the record should say so. Converting every headline to “IT MW” may make a table easier to rank, but it quietly creates information that the source never provided.

Second check: what date does the status describe?

“Under construction” is meaningful only if we know when the evidence was observed. A groundbreaking in 2024, an operator update in 2025 and a target opening in 2027 are three different dates. Collapsing them into a single last-updated field makes it impossible to see whether the project is progressing or merely being repeated.

This is where market data often ages badly. The headline survives; the evidence date disappears. Our development-readiness framework therefore treats land, power, permitting, design, finance, construction and commissioning as distinct milestones rather than one status label.

Third check: does the grid layer prove capacity?

Grid geography matters, but proximity is not a reservation. ENTSO-E’s Transparency Platform and national datasets such as RTE’s éCO2mix data are valuable public infrastructure sources. They describe systems, generation and flows. They do not automatically disclose whether a specific data-center load can connect on a specific date.

My view is firm here: a nearby transmission line is a screening signal, not a bankable power right. The gap between network context and project entitlement is exactly what Data Center Grid Upgrade Costs: Who Pays? is designed to expose.

Read the gaps as data

A good dataset should not hide missing evidence. It should make the gap usable.

If a site has a verified developer and location but no public capacity basis, that is not a failed record. It is a record with a specific limit. If a project has a public grid application but no connection agreement, the application is still relevant—provided the status is not upgraded by inference. If a campus has a target date without a construction milestone, that uncertainty belongs beside the date.

This is also why I resist scoring systems that produce a single readiness number without showing their inputs. A 72/100 score can look authoritative while mixing verified land, estimated power and an old delivery target. The number may be useful for triage, but the evidence must remain available underneath it.

The powered-land due-diligence guide applies the same discipline at site level. “Powered” is not a property type. It is a stack of evidence covering rights, capacity, works, timing and conditions.

A live record needs an audit trail

The operational challenge is not only collecting data. It is updating it without destroying history.

When a project moves from announced to permitted, the earlier claim should remain traceable. When a capacity is revised, the previous basis should not vanish. When a source is corrected, the record should show why the change occurred. This is particularly important in fast-moving AI infrastructure, where campus visions, equipment orders and tenant discussions can be mistaken for delivered supply.

RTE’s February 2026 review of French electricity consumption illustrates the value of a dated baseline: it reported almost 1 TWh of 2025 consumption for dedicated data-centre sites connected to the transmission system, while also distinguishing distribution-network consumption. The figures are useful because the scope and period are explicit.

At PowerlandMap, the practical rule is similar. Every material claim should carry a source, an observation date, a capacity basis where relevant and a confidence level. The product overview explains how those layers connect across markets, sites, power and demand.

What I would ask before using any market map

I would not begin by asking how many sites it contains. I would ask five less comfortable questions.

Can I open the source behind the record? Can I tell when it was published and when the record was observed? Can I distinguish IT load from electrical or campus capacity? Can I see whether status is reported, estimated or inferred? Can the dataset preserve a correction without rewriting history?

If the answer is no, the map may still be useful for discovery. It should not be the only basis for capital allocation, site selection or a delivery commitment.

This is the practical standard behind data center market intelligence: more records are helpful only when provenance scales with coverage. My own bias is to retain a visible uncertainty rather than manufacture a clean answer. That makes the dataset slightly less comfortable to read and far more useful in a real decision.

Teams that want to inspect the evidence structure across markets, sites, power and project phases can request access to PowerlandMap. The objective is not to make every project look comparable. It is to show where comparison is valid—and where further diligence is still required.

*Matthieu Gallego*