Capital Is Not Capacity: The Test Data Centre Announcements Must Pass

The data-centre market has become fluent in large numbers. Investment announcements arrive in hundreds of millions of dollars. Campuses are described in acres or square feet. National strategies are measured in gigawatts. These numbers signal intent, but they do not answer the question that determines commercial value: what power-backed IT capacity can actually be delivered, and when?

Equinix's latest Chicago announcement is a useful test case. On 29 October, the company described an initial US$215 million investment in CH5 at Elk Grove Village, its sixth site in the Chicago region. It said the facility is expected to provide more than 190,000 square feet of colocation space at full buildout, with roughly 150 construction jobs and up to 50 permanent roles. It also highlighted air cooling designed to avoid routine water consumption.

Those are meaningful facts. They establish capital commitment, location, facility scope and parts of the operating design. But the announcement does not disclose a megawatt figure, commissioned area, customer commitments, the grid connection detail or the date at which the full-build footprint will be available. The correct conclusion is not that CH5 lacks capacity. It is that investment and floor area cannot be converted into deliverable IT megawatts without another layer of evidence.

That distinction is becoming more important as electricity moves from an operating input to a gating asset. The IEA's latest assessment expects global data-centre electricity consumption to rise from about 485 TWh in 2025 to 950 TWh in 2030. In that environment, the market will increasingly reward projects that can prove not only that a building can be financed, but that power can be contracted, connected, commissioned and allocated to a customer on a credible timetable.

A live evidence gap, not a theoretical one

A PowerlandMap snapshot taken on 1 November 2026 contains 7,574 published, non-demo supply records. It is a working evidence inventory, not a census of unique campuses and not a statement of available capacity.

Within that bounded population:

The relevant ratio is simple: 27 of the 77 records with capex, or 35.1%, have no direct IT or campus capacity field. That is not a market-wide disclosure rate. It reflects the records and fields in this snapshot. It nevertheless demonstrates why a capital headline cannot be treated as a capacity proxy even inside a structured database.

The same caution applies in reverse. A capacity value does not automatically establish available capacity. It may describe an ultimate campus envelope, a single phase, a utility request, a contracted block or an estimate derived from public evidence. The basis matters as much as the number.

PowerlandMap therefore keeps direct capacity, estimated capacity and the evidence basis in separate fields. The 1,070 estimated-only records are not discarded, but they should not be blended with direct disclosures as if the two populations carried the same confidence.

Why the conversion fails

There are four reasons why dollars, square feet and megawatts should remain separate until the evidence connects them.

First, capex scope varies. One announcement may include land, shell construction, electrical infrastructure and several future phases. Another may cover only the first fitted-out hall. A third may include network equipment or renewable-energy assets. Dividing capex by an assumed MW value can therefore produce a precise-looking ratio with an undefined denominator.

Second, floor area is not IT load. A published area may refer to gross building space, technical space, colocation area or a full-build total. Different cooling systems, density assumptions, redundancy designs and non-IT areas can materially change the relationship between the footprint and usable IT capacity.

Third, grid access is staged. A project can have a financed building while the final connection, transformer programme or upstream reinforcement remains conditional. It can also secure an electrical envelope that is larger than the first commissioned phase. Neither situation is unusual, and neither should be collapsed into one "capacity" number.

Fourth, time changes the commercial meaning. A 100 MW ultimate campus that can deliver 12 MW in 2028 is a different product from a 20 MW operating facility with 4 MW currently available. For a customer with a fixed service date, the first phase and the handover evidence matter more than the end-state marketing number.

This is why I read the Equinix CH5 announcement as a positive investment signal but not as a capacity disclosure. The company has supplied useful facts that deserve to be preserved exactly. The missing MW and delivery details remain unknown rather than estimated from the US$215 million investment or the full-build area.

The decision test

A serious screening process should force every project through five separate questions.

1. What exactly has been announced?

Record the figure and its scope without translation. For CH5, that means an initial US$215 million investment and more than 190,000 square feet of colocation space at full buildout. It does not mean US$215 million has already been spent, that the full space is commissioned, or that a specific IT load is available.

2. What capacity basis is documented?

Identify whether the number is IT load, utility power, campus power, a grid request, a contracted connection, a phase or an estimate. If no capacity number is disclosed, leave it unknown. Do not infer it from floor area, generator count or investment unless a transparent estimation method and range are explicitly presented.

3. What is the current development stage?

Separate an announcement, permit, power reservation, construction start, commissioning milestone and operational service. These are not synonyms. Each stage transfers a different risk from the developer to the customer, lender or utility.

4. What can be delivered by the required date?

The relevant capacity is the block that can be handed over under the customer's technical conditions, not the largest number in the press release. The diligence needs phase dates, commissioning dependencies, cooling compatibility and evidence of the power allocation behind the offered block.

5. What would change the conclusion?

A good record identifies the next proof point. For CH5, a future technical disclosure, phase commissioning notice or product specification could establish the missing capacity basis. Until then, the investment announcement supports confidence in the project's seriousness, not a megawatt conversion.

The economic mechanism

The difference between capital and capacity changes who bears the risk.

If investors value a project mainly from headline capex, they may reward expenditure before the asset's deliverable output is clear. If developers market ultimate capacity without phase evidence, customers may spend diligence time on sites that cannot meet their service date. If utilities treat every request as equally mature, network planning can become distorted by projects with very different financing and construction readiness.

A capacity-evidence approach does not penalise early projects. It makes their milestones investable. Land control, permits, contracted power, energisation dates and phase financing can be tracked independently. Each completed step narrows uncertainty and can justify a different valuation or commercial conversation.

The approach also protects operators whose public disclosures are intentionally conservative. An operating facility with no published MW figure should not be assigned zero capacity. It should be classified as operational with capacity unknown, then qualified through the appropriate commercial channel. Unknown is a prompt for diligence, not a negative score.

A practical comparison

Consider two fictional projects.

Project A announces a US$500 million campus and 400,000 square feet, but no MW figure or phase date. Project B announces a 24 MW IT first phase, documents the grid connection and targets handover in 30 months, but does not publish total capex.

A capital-led ranking puts Project A first. A delivery-led screening may put Project B first for a customer seeking 16 MW in three years. The ranking changes because the decision changes. Neither project is inherently superior; each provides different evidence against a specific requirement.

The counter-argument is that the market needs fast comparables, and capex per MW is one of the few ratios available. That ratio can be useful when both numerator and denominator share the same scope. It becomes misleading when a full-campus investment is divided by first-phase IT load, when utility power is compared with IT load, or when estimated capacity is mixed with direct disclosure. The answer is not to abandon comparables. It is to label the basis and refuse false precision.

What this changes for the market

For investors, the next generation of data-centre underwriting should separate money committed, construction completed, power contracted and capacity commissioned. These milestones do not move together.

For operators, disclosure quality becomes a commercial asset. A clear first-phase specification can be more valuable to a buyer than a larger ultimate-campus headline.

For advisers, the evidence gaps are the work programme. They reveal which technical, grid, permitting and financial questions can change a decision.

For policy makers and utilities, maturity-weighted pipelines are more informative than raw application totals. The IEA's demand outlook makes the point urgent: the constraint is not merely whether capital wants to build, but whether the energy system can convert that intent into reliable service.

My conclusion is straightforward. Capital proves commitment. Floor area describes physical scope. Capacity requires its own evidence. The market should stop translating one into another by habit.

Sources and method

The CH5 facts come from Equinix's 29 October 2026 announcement and the company's CH5 facility page. The energy-demand context uses the IEA's Key Questions on Energy and AI.

The PowerlandMap extraction includes published, non-demo records in `mi_supply_current` as at 1 November 2026. Counts are database records, not deduplicated campuses, available inventory or market share. Direct capacity means a populated `it_mw` or `campus_mw` field. Estimated-only means `estimated_it_mw` is populated while `it_mw` is not. The 35.1% figure is 27 divided by 77 and rounded to one decimal place. No confidential data is used. No missing value is treated as zero.