Four Numbers a Data Center Portfolio Does Not Tell You

A large data center portfolio announcement answers one question: the sponsor is prepared to commit capital. It does not automatically answer where that capital will be spent, how much IT capacity each site can support, when power will be available, or which planning risks remain.

Google’s announcement on 9 September 2026 is a useful example. Reuters reported that the company intends to invest at least €13 billion in Finnish AI infrastructure during 2027–2028. The programme includes three new northern data centers, an expansion of the existing Hamina campus, grid improvements, clean-energy measures and battery projects. It is a major commitment—but not a site-by-site capacity schedule.

For investors, developers, utilities and enterprise users, the practical task is therefore not to repeat the headline. It is to turn the announcement into an evidence-based portfolio map.

Portfolio capital is not site-level CAPEX

The first discipline is to preserve the boundary around the reported number. Google’s €13 billion is a portfolio commitment covering several kinds of infrastructure. Public reporting does not allocate the figure between Hamina, Kajaani, Muhos and Vaala, nor between buildings, grid works, energy contracts and batteries.

That distinction matters. Dividing €13 billion by four locations would create a figure that looks precise but has no public basis. Dividing it by an assumed number of megawatts would compound the error, because no site-level IT MW or electrical MW was disclosed in the announcement.

In PowerlandMap’s methodology, reported figures and estimates are kept separate for exactly this reason. A portfolio-level budget is valuable evidence of sponsor commitment. It is not evidence of deliverable capacity at any individual site.

The same rule applies to capacity announcements. A group target, a grid request, a campus electrical envelope and contracted IT load are different measures. They should not be collapsed into one “MW” field.

The four-site Finnish portfolio has different maturity signals

Public evidence already shows that the four Finnish locations are not at the same stage.

Hamina is an operating Google campus and the new programme includes further expansion. It starts from a different baseline from the three northern locations.

Yle reported on 9 September that the new sites are in Otanmäki in Kajaani, Pyhänsivu in Muhos and Nuojua–Järvikylä in Vaala. The public broadcaster also reported visible earthworks in Kajaani and Muhos, while approximately 900 hectares are being considered in zoning at Vaala. Planning decisions in Kajaani and Muhos have been subject to appeals, which may affect timing.

These are material facts, but they do not yet create comparable capacity records. Earthworks indicate physical progress; they do not confirm energisation. A large zoning area indicates development scope; it does not confirm how much land will be used for data halls. An investment programme indicates funding intent; it does not replace a grid milestone.

This is where a structured coverage view becomes more useful than a single news card. Each site needs its own record, sources and milestones, while remaining linked to the portfolio announcement.

Four numbers decision-makers still need

A credible portfolio view should seek four numbers that the headline does not provide.

1. Site-level IT capacity

IT MW describes the computing load that a facility can serve. It should be reported only when the project sponsor, planning documents or another reliable source defines it. Until then, the correct value is “not disclosed”, not an estimate derived from capital.

2. Electrical capacity and its contractual status

A grid request, an offer, a reservation and an executed connection agreement are not interchangeable. Decision-makers need both the electrical quantity and the milestone behind it. They also need the expected energisation date, any reinforcement dependencies and whether on-site generation or storage changes the connection strategy.

The 22-year power agreement reported between Google and Fortum is important at portfolio level. Reuters said it covers up to 50% of output from Fortum’s Loviisa nuclear plant. That supports long-term energy strategy, but it does not by itself disclose the connection capacity of Kajaani, Muhos or Vaala.

3. Phase timing

A single programme window can hide several project schedules. Site preparation, planning appeals, substation delivery, first building completion and ramp-up should be tracked separately. The relevant question is not simply “when will the campus open?” but “what capacity can be delivered, at which site, under which conditions, and by when?”

4. Capacity available to the market

Hyperscale self-build capacity, operator capacity and capacity available to third parties serve different demand pools. Google’s Finnish programme is strategic infrastructure for its own cloud and AI operations. It should therefore be recorded as supply infrastructure without implying that the capacity is available for colocation tenants.

PowerlandMap’s product is designed to preserve these distinctions across markets, projects and evidence updates.

A practical portfolio-planning framework

A useful framework separates the announcement into four layers.

The capital layer records the amount, period, sponsor and stated use of funds. The site layer identifies each location, land position, planning status and construction evidence. The power layer records grid and energy milestones without confusing procurement strategy with physical connection readiness. The delivery layer tracks phases, dependencies and realistic dates.

Only after these layers are assembled should users compare the portfolio with alternative markets or capacity options. A developer may use the comparison to sequence sites. A utility may use it to understand future demand concentration. An investor may use it to test whether capital is matched by development readiness. An end user may use it to judge where capacity is likely to become operational.

Users evaluating a specific market can request access or book a demonstration to see how public supply, power and development signals are connected in the platform.

What to watch next in Finland

The next material updates will not necessarily be larger investment numbers. More useful signals may be planning appeal outcomes, building or environmental permits, grid-connection milestones, substation works, published construction packages, battery specifications and site-level delivery dates.

Any disclosed IT or electrical capacity should be attributed to the correct location and phase. Any portfolio-wide number should remain at portfolio level. If a source later provides a more precise figure, the record can be enriched without rewriting the original evidence.

The same approach applies beyond Finland. NEXTDC’s reported A$1.1 billion convertible-note raise is explicitly linked to an Australian development pipeline, but it still requires project-level mapping before it can support site comparisons. Capital is a trigger for deeper diligence, not a substitute for it.

PowerlandMap view

Our view: the market overvalues headline CAPEX and undervalues evidence structure. Large numbers attract attention, but commercial decisions depend on the relationship between land, power, planning and timing at individual sites.

The best market-intelligence systems do not force every announcement into a capacity estimate. They preserve uncertainty, identify the missing evidence and make the next verification step visible. That is how portfolio news becomes decision-grade intelligence.

PowerlandMap was originally developed as an internal market-intelligence tool within The Blob Company and has evolved into an independent SaaS product. Explore the platform, review the methodology, compare pricing, or request a demo for a market or portfolio relevant to your decisions.

*Matthieu Gallego* *Founder, PowerlandMap*