Datacentre operators spent the last cycle competing on land and connectivity. That competition has moved on. Power is now the industry’s scarce input and by extension, its main competitive advantage. The more important story is no longer the scarcity itself, but how capital is organising around it.
The constraint everyone already knows about
EY-Parthenon’s report Hyperscale meets hyperlocal: Value maximization in the data centre boom set out the mechanics of the constraint: transmission capacity booked out for years in Tier I markets, planning restrictions in core European metros, and grid connection queues that now run well beyond a typical development cycle. None of that is news to anyone reading a datacentre deal sheet. What has changed is how operators, hyperscalers and capital providers are responding and that shift in behaviour, rather than the constraint itself, is where the opportunity now sits.
Where operators once banked land and speculative grid connections independently, the market is now converging. Datacentre operators, utilities, renewable generators, power-focused investors and hyperscalers are now assembling around the same scarce asset. For nearly every power generator and utility, datacentres have become a priority sector; the scale of demand makes standing aside difficult to justify.
“With datacentre demand continuing to accelerate, power generators and utilities are increasingly viewing the sector as a key source of future growth and long-term value creation.”
— Dwayne Gibson-White – Partner, Digital Infrastructure Leader, EY-Parthenon
The US offers a preview. Cloud providers there have moved into generation through direct ownership and joint ventures with power producers, and the US market is several years further into the AI build-out than Europe. That wave of demand is now reaching European markets. Growth may not match US rates, but EY-Parthenon still expects European datacentre demand to increase three- to four-fold over the next decade.
Structuring for power: build, own and play
The practical result of convergence is a widening list of deal structures built to secure power, such as private wire connections running directly from generation to campus, utility co-investment, and renewable developers with no intent to operate datacentres actively seeking datacentre demand. The question worth asking of any new structure is the same one asked of any infrastructure financing: who creates the value, who carries the risk, and how are the two allocated?
Private wire shows how the answer shifts with the asset. In metro facilities, EY-Parthenon sees operators taking ownership of the private wire as part of the build and recovering the cost through conventional yield-based pricing. For larger remote campuses, the build more often sits with the power company or utility, on terms the market is still defining. The underlying bargain is consistent: the tenant gains access to power and certainty of base load, while the power provider secures an anchor offtaker that de-risks its own build.
Renewable developers face a similar spectrum, from a straight PPA through co-development to an equity stake. Joint ventures are already a feature of the market, with the operator contributing construction expertise, track record and tenant relationships, and the developer bringing generation capability outside the operator’s remit. Guaranteed offtake remains the traditional route. The natural extension is for the developer to retain equity in the datacentre itself, moving from selling power to offering powered land.
“The traditional model has been to secure long-term demand through an offtaker agreement. Increasingly, however, utilities are looking to retain equity in the datacentre itself, transforming a power supply arrangement into a powered land proposition.”
— Rob Harris-Deans, EY-Parthenon
The harder question is duration. Utilities building transmission are underwriting assets with lives far longer than the contracts that justify them. Hyperscale contracts typically run 10-15 years, with extensions of around five years expected, but grid investment timelines still outlast the lease. To date the mismatch has been manageable. Churn among cloud providers from existing facilities has been very low, held down by the reputational risk and cost of migration and by power scarcity itself, which has given utilities confidence in renewal or in the ability to re-let. That confidence is being tested as the market moves towards AI-focused remote campuses leased to neo-clouds on five- to eight-year terms.
“To date, churn from cloud players in existing facilities has been incredibly low. But as we move to AI-focused remote campuses with shorter neo-cloud contract terms, the mismatch between grid investment and lease length is coming into sharper focus.”
— Rob Harris-Deans, EY-Parthenon
Outsourcing is a regulatory and land-bank story, not a cost story
The report’s regional outsourcing split remains one of its most telling findings. Outsourcing runs at roughly 10-20% in North America against 80-90% across MENA, with Europe at 50-60% and APAC at 50-70% in between. The instinctive read is cost. EY-Parthenon’s read is speed, access and local expertise.
Hyperscalers with long histories of land and power banking in the US can self-build. Europe differs in kind as well as degree. Rather than a single, largely homogeneous market, it is a patchwork of national regulators, planning regimes and grid operators, which puts a premium on local expertise. Metro land and power access reinforces the point: cloud providers and neo-clouds cannot realistically bank land and power across every European metro, and rely on third-party operators to do it for them. The market consistently describes operator-led delivery as materially faster than self-build, and with demand at its current scale. With this faster speed-to-market, higher return on assets and greater operational flexibility, EY-Parthenon’s report predicts reliance on outsourcing to persist rather than unwind.
“Given the scale of opportunity across European markets, hyperscalers and neo-cloud providers cannot realistically secure and develop every site themselves. As a result, they are increasingly reliant on third-party operators to source, power and deliver capacity.”
— Dwayne Gibson-White, Partner, Digital Infrastructure Leader, EY-Parthenon
Higher outsourcing in MENA, APAC and LATAM reflects a different mix: the need to work with local providers to meet sovereignty requirements, and the challenge of finding the right skills in less established datacentre supply chains.
That advantage compounds as workload placement becomes more sophisticated. Latency-sensitive inference stays close to the end user, while latency-agnostic training and inference is increasingly routed to lower-TCO markets such as the Nordics and Iberia, though rarely exclusively, as a shift in data sovereignty rules could strand a single-market bet. The resulting footprint is a deliberate hedge: cheap-power markets for base-load workloads, and metro and neo-cloud capacity in AI-positive economies such as the UK, France and Germany.
Pricing is diverging by tenant, not just by market
The report’s renewal analysis points to uplifts of 10-13% at contract renewal in established US markets, with this uplift potentially being pushed higher as spot market rates are rising due to underlying build cost inflation. The brake on further increases is migration risk, and the fact that a hyperscaler negotiating hard on one facility is also weighing the others it might build with the same operator.
A second, sharper dynamic is emerging beneath that. Neo-clouds, working to 12-18-month procurement horizons rather than the three- to four-year lead times hyperscalers plan against, are paying a premium for capacity. EY-Parthenon sees that premium reflecting both urgency and counterparty risk: neo-clouds need capacity now, their credit quality does not match a cloud service provider’s, and their shorter contracts leave operators more exposed at expiry. It remains an early market, and the premium varies by contract and location, so it is the direction of travel rather than any single figure that matters.
Dedicated instruments that insure against neo-cloud counterparty risk are still emerging. The nearest analogues sit in GPU (Graphics Processing Unit) financing, where revenue-share structures send customer revenue directly to the lender rather than through the neo-cloud, and in special purpose vehicles that move lenders up the capital stack, backstopped by the end-user contract rather than the neo-cloud’s balance sheet. Those structures point to where the next wave of innovation is likely to come from.
“The closest we have seen is in GPU financing, where customer revenue goes directly to the lender rather than via the neo-cloud, and SPVs that move the lender up the capital stack so they are backstopped by the end-user contract.”
— Rob Harris-Deans, EY-Parthenon
What to watch
The throughline is consistent: power has stopped being a line item and become the asset that determines who wins a market. EY-Parthenon’s Hyperscale meets hyperlocal report mapped where that constraint bites; the next question is who controls the answer. That reframes the diligence question for investors from “does this operator have land and connectivity?” to “who in this operator’s ecosystem controls the power, and on what terms?”. The structures answering that question, including private wires, renewable-developer partnerships, utility co-investment and, in time, mechanisms that ring-fence neo-cloud counterparty risk, are still being written. That is exactly why they are worth understanding now, before they standardise.
