A Data Center Project Status Needs a Change Log

A data center project status should never be a label without a history.

For an investor, developer or occupier, “planned”, “under construction” and “live” are not editorial descriptions. They are decision states. Each one changes which costs can be committed, which delivery assumptions can enter a model and which questions must still be answered. If the status is overwritten without recording the evidence and date, a map can look current while its investment logic is already stale.

My view is simple: every material status change should be treated as an evidence event. The current label matters, but the change log matters more. It shows what moved, why it moved, what did not move, and which claims remain unsupported.

That approach is especially useful when projects evolve through approvals, revised operator disclosures, construction starts and certifications that do not arrive in a neat sequence. A public announcement can be genuine and still be insufficient for a commercial decision.

The decision: commit only against the highest proven state

I use four working states for a first-pass review:

• Planning: a named project and location supported by a current developer or authority source. This does not prove a permit, construction, grid rights or a delivery date.
• Approved or permitted: a specific approval or permit tied to the project. This does not prove construction start, financing, energisation or capacity availability.
• Under construction: current operator or authority evidence of physical works. This does not prove commissioning, customer-ready capacity or occupancy.
• Live: operator evidence of operation, ideally supported by certification or another independent record. This does not prove vacant capacity, contracted power headroom or expansion timing.

The rule is deliberately conservative. A project can have more detail than the minimum and still remain in the same state. The purpose is not to understate progress. It is to prevent one encouraging fact from silently proving five others.

This complements the PowerlandMap development-readiness framework: land, power, permits, construction and commercial readiness should be tracked separately. It also extends the evidence logic described in Can Data Center Market Intelligence Be Trusted?. A project record should retain both the claim and the uncertainty.

A five-record audit across four markets

On 20 October 2026, I reviewed five PowerlandMap project records across Thailand, Czechia, Poland and Cyprus. This was a bounded quality-control sample, not a representative market survey. The unit of observation was the individual project record. Sources were reopened, current operator or authority statements were compared with stored fields, and capacity, status and timing were reconciled separately.

The sample produced four different outcomes:

• one capacity revision and confirmed construction state
• one construction-state advance without commissioning evidence
• two projects kept below construction because their sources did not support a higher state
• one operational facility added with capacity left unknown

That distribution is the point. A good review does not push every record forward. Sometimes the correct update is a stronger limitation.

Thailand: approval gave way to a newer construction disclosure

GSA Data Center 02 in Chonburi was initially associated with a Thailand Board of Investment approval describing 35 MW of IT load. A newer GULF investor presentation, supported by the company’s 2026 AGM minutes, identifies GSA02 as under construction, with a current scope of up to 38 MW and scheduled commercial operation in the first quarter of 2027.

The update therefore changed more than one field. The status advanced to under construction. The current design scope became 38 MW. The earlier 35 MW approval figure was retained as historical evidence rather than deleted. The quarter was stored as operator guidance, not converted into a guaranteed commissioning day.

Nothing in those documents proves energised capacity, customer availability or occupancy. The project can be under construction and still require separate diligence on grid delivery, commissioning and commercial allocation.

Czechia: construction can start while the capacity boundary remains unclear

MasterDC’s Kanice announcement describes an initial 4 MW energy centre, a planned site input of up to 25 MW, 80–100 racks and an autumn 2026 operating target. On 20 October, the operator’s current website stated that construction of the specialised AI facility was starting.

That was enough to advance the lifecycle from development to under construction. It was not enough to record 25 MW as IT capacity. The 4 MW and 25 MW figures describe electrical scopes in the operator material. The 5 MWp photovoltaic figure is generation. They should not be added, converted or compared as though they were the same boundary.

This is the same discipline applied in From Grid MW to Live IT Capacity: the number is only useful after its physical and commercial boundary is clear.

Poland: removing false precision is a material update

Microsoft’s Ożarów Mazowiecki project page says the campus is in early planning and undergoing environmental-impact assessment before a building permit can be issued. A prior internal modelled estimate of 54 MW was therefore removed. The source does not publish site-specific capacity, and “unknown” is more accurate than a precise-looking proxy.

Switch Datacenters’ WAW1 site page presents a 50 MW IT design, a 22 MW first phase and 90 MVA utility feeders. It says construction is expected to start in 2027. The review moved 50 MW from current IT capacity to reported design IT capacity, kept 90 MVA as a separate utility input and removed an unsupported ready-for-service date.

Neither correction makes the projects less real. Both make them more usable. An underwriting model can now see what is design, what is utility scope and what is still a future milestone.

Cyprus: live status did not justify a megawatt estimate

CL8 states that it has operated a carrier-neutral facility in Limassol since 2018. The operator description is supported by the Uptime Institute record for CL8 LIM Data Center Phase 1, including Tier III design and constructed-facility awards.

That combination supported a live status. It did not support a capacity number. No reviewed source disclosed aggregate electrical MW, IT MW, occupied load or commercially available headroom. The record was therefore added as operational with capacity unknown.

A live site with unknown capacity is more valuable than a fictitious number. It tells the user which fact is proven and which conversation still has to happen.

Why the change log matters economically

A status error can change how a shortlist is priced. Consider an illustrative 10 MW IT phase. If a model assumes a project will be customer-ready and it is delayed, the gross contribution deferred can be expressed as:

`IT kW × monthly contribution per kW × delay months`

The table below is a sensitivity, not a market benchmark. It excludes utilisation ramp, taxes, financing, capex, termination rights and the possibility that revenue is recovered later.

The same calculation produces €1.2m, €2.4m and €3.6m after three months at €40, €80 and €120/kW-month. After six months, the figures are €2.4m, €4.8m and €7.2m. After twelve months, they are €4.8m, €9.6m and €14.4m.

This is gross contribution deferred, not automatically lost revenue or net present value. The lesson is narrower: a lifecycle label can move a model by millions before any physical asset changes.

The previous PowerlandMap analysis, What a Six-Month Data Center Delay Costs, develops the timing question in more detail. Here, the practical control is to prevent the status field from importing an unsupported delivery date into that model.

The counter-argument: too much conservatism can hide momentum

There is a reasonable objection. If every project remains “planning” until construction is undeniable, a dataset may react too slowly and understate genuine progress. Markets move through conditional milestones, and users often need an early signal.

I agree with the need for early signals. I do not agree that the answer is a more optimistic status.

The better design is to record both the current proven state and the leading indicators: approval granted, contractor appointed, equipment ordered, construction announced, utility agreement signed, commissioning target stated. Those events can improve confidence without being collapsed into “live”.

This preserves speed without sacrificing auditability. It also makes contradictions visible. A 2025 approval, a 2026 operator presentation and a later construction update can coexist as a sequence rather than compete for one undated label.

What I would ask before accepting a higher state

For any project moving through the pipeline, I would ask for the evidence that closes the next gate:

1. From planning to approved: the permit or approval, its scope, conditions and project identity. 2. From approved to under construction: dated evidence of physical works tied to the facility, not only site preparation elsewhere on a campus. 3. From construction to live: energisation, commissioning or operator confirmation that the facility is operating. 4. From live to available: the specific commercial capacity, timing, configuration and conditions offered to customers.

I would also keep power boundaries separate throughout. A 90 MVA feeder, 50 MW design IT envelope and 22 MW first phase are related, but they are not additive. Data Center Capacity Is Not One Number explains why this separation belongs in every comparison.

A practical method for current market intelligence

PowerlandMap’s contribution is not a claim that every public record is complete. It is a method for keeping project identity, evidence, capacity boundary, lifecycle and uncertainty together.

For this five-record sample, the method was:

• extraction date: 20 October 2026
• population: five named project records in four countries selected from the current quality-control queue
• evidence: operator, authority and certification sources opened directly
• unit of observation: project record, not company, campus portfolio or market
• exclusions: no private CRM facts, no client data, no inferred occupancy and no conversion of electrical input to IT load
• limitation: the sample is operational QA and cannot support a market-wide error rate

The decision rule is straightforward: advance the status only when the evidence supports it, retain prior facts as dated events, and leave material unknowns visible.

If you are comparing a shortlist, the most useful output is not the project with the biggest headline. It is the project whose next milestone, power boundary and evidence trail can be tested. PowerlandMap’s market studies apply that logic to site, power and market comparisons where the decision requires more than a pin on a map.