Scale Changes the Operating Model

Scale is changing the operating model of data-center development.

Three public signals make the point. On 5 October 2026, atNorth announced a new Salo campus with 75 MW secured for its first phase and a pathway to 230 MW of gross power. The same day, DayOne filed for a proposed US initial public offering, bringing capital-market scrutiny to a platform active across ten markets. On 6 October, W.Media reported that Firmus and CDC had ended their proposed Southgate development partnership after about 42 MW had been deployed.

These are different businesses and different events. I am not suggesting a common cause. The useful inference is narrower: as portfolios grow, the risk shifts from finding one site to controlling many interfaces at once.

That is the core of data center portfolio delivery. A large pipeline is not simply a collection of projects. It is a system of land, power, permits, capital, customers, partners and people moving at different speeds.

Facts, inference and the decision that follows

The facts belong in an evidence register. atNorth has disclosed the location, land area, first-phase power and longer-term campus scale. DayOne has disclosed an IPO filing and a multi-market footprint. Firmus and CDC have publicly described a change in their development relationship.

The inference is that each event creates a governance question. For a new campus, can the first phase reach service without relying on the full end-state? For a listed or listing-stage platform, can investors compare projects on a consistent basis? After a partnership changes, can the sponsor preserve customer, power and delivery commitments while reallocating responsibilities?

The decision is not whether to “believe” the headline. It is whether the operating model is strong enough to convert each promise into a dated, accountable workstream.

This distinction builds on our earlier explanation of why data-center capacity is not one number. A 230 MW campus, a 75 MW secured phase and a 60 MW IT deployment are related figures, but they do not have the same evidence, timing or delivery owner.

Five interfaces decide whether a portfolio can scale

1. Portfolio strategy versus project reality

Management may allocate capital by market, customer or strategic theme. Projects advance through local grids, authorities, contractors and land agreements. The portfolio model must therefore preserve a common language without erasing local differences.

At minimum, every project should report the same definitions for controlled land, contracted power, IT capacity, gross power, planning status, target ready-for-service date and confidence. The evidence standards described in our market-intelligence audit are especially important when the data is used for investment committees rather than marketing.

A portfolio dashboard that shows only headline MW is not a control system. It is a catalogue.

2. End-state ambition versus phase-one independence

The first phase should be underwritten as a project in its own right. It needs a defensible power block, access, permits, cooling concept, procurement plan, commissioning route and commercial case.

Future phases can create valuable expansion rights, but they should not quietly carry costs or dependencies required by phase one. Our development-readiness milestones offer a useful baseline: each milestone needs a source, owner, date and condition of satisfaction.

The same discipline applies to power. A utility allocation, a connection agreement and delivered energisation are different levels of evidence, as set out in our grid-connection due-diligence framework.

3. Sponsor capability versus partner dependency

Multi-market growth usually requires partners: landowners, utilities, design teams, construction managers, operators, customers and capital providers. The mistake is to treat the partnership itself as the delivery mechanism.

A serious operating model maps each critical obligation to a named party and then asks what happens if that party changes. Who retains the grid application? Who controls design information? Can procurement orders be transferred? Which consents are needed to replace a contractor, operating partner or capital source? Are customer commitments portable?

The Firmus–CDC development change is a current reminder that partnerships can evolve while contracted capacity and broader strategy continue. The lesson is not that partnerships are fragile. It is that replacement pathways belong in the original governance design.

4. Capital availability versus executable spend

A financing event increases optionality; it does not remove sequencing risk. The portfolio still needs to decide which sites receive deposits, long-lead orders and management attention first.

That is why capital is not capacity. Capital should be released against evidence gates: land control, connection milestones, permit maturity, customer conditions and procurement readiness. Debt and equity may also test different risks, which we discuss in reading data-center portfolio financing signals.

The practical output is a capital-allocation matrix that shows not only expected return, but also the next irreversible commitment and the evidence required before making it.

5. Commercial demand versus delivery acceptance

A customer commitment may be powerful evidence of demand, but it introduces a second schedule. The building must reach mechanical and electrical completion; the customer must be ready to install, test and accept its systems.

For AI infrastructure, this interface is getting more specialised. Data Center Dynamics reported that Anthropic appointed a former Google data-center design and construction executive to lead EMEA capacity delivery. The appointment is evidence that sophisticated users are building dedicated capacity functions, not simply buying undifferentiated megawatts.

Portfolio governance should connect contractual conditions in a pre-lease agreement with the technical evidence required for commissioning readiness.

A compact control model

I would expect a scaling platform to maintain four linked views.

First, an evidence register: one dated source for every material claim, including who verified it and when it must be refreshed.

Second, an interface matrix: sponsor, utility, partner, contractor and customer obligations, with clear hand-offs and remedies.

Third, a phase plan: the minimum independent scope for phase one, followed by separately evidenced expansion options.

Fourth, a portfolio intervention list: the ten decisions that need senior management action, not the hundred tasks already being handled by project teams.

This is where permitting and stakeholder work stay connected to delivery. A project may have a credible technical plan while still carrying unresolved consent risk. Our guides to permit due diligence and community engagement show why these workstreams must sit in the same decision pack as power and capital.

The model also works for smaller developers. A USTDA-backed feasibility opportunity in Nigeria describes two sites and phased expansion. Even before construction, a two-site programme needs comparable assumptions, separate evidence trails and a clear decision on what is shared versus site-specific.

My view: standardise evidence, not judgement

The best portfolio teams do not force every project into the same answer. They force every project to answer the same questions.

That difference matters. Grid processes vary by country. Permitting evidence varies by authority. Construction markets, cooling choices and customer requirements vary by site. A rigid template can hide those differences. A consistent evidence model makes them comparable.

PowerlandMap is designed around that idea: projects, markets, power milestones and source dates should be viewed together, without collapsing them into a single pipeline number. The purpose is not to replace local development judgement. It is to show where judgement is being applied, which facts support it and which interface needs attention next.

If you are comparing a multi-market pipeline or testing whether a portfolio can absorb its next phase of growth, request access and we can review the evidence structure with you.

*Matthieu Gallego*