Data Centres: Power, Risk and Resilience 2026 - Highlights

inspiratia convened industry experts and commentators in London on 18 June for a wide-ranging discussion on the future of data centres and the power systems underpinning them. 

The strong positive response to our inaugural data centre event reflects just how significantly digital infrastructure is intensifying the pressures on Europe's energy transition and how urgently the market is grappling with the constraints that will shape its next phase of growth. 

Daniel Atzori, inspiratia's Head of Intelligence, opened the day with welcoming remarks and set the stage with an overview of the themes that would run throughout the day. 

• He argued that the sector is at a crunch point, with data centre execution becoming harder as grid access and permitting increasingly determine which projects are actually energised.
• Capital now favours the most bankable paths to build and is shifting from megawatts announced to megawatts actually delivered.
• This is not an exodus from FLAP-D, rather a build-out of secondary hubs and spillover corridors with clear planning and low-carbon power. 

The day opened with a panel on data centres as system-critical infrastructure, looking at how Europe can absorb AI and hyperscale demand without derailing decarbonisation or destabilising the grid, whether data centres should be treated as strategic infrastructure in national energy planning, and what the UK and EU can learn from each other on grid access and policy. 

Speakers included:  

• Søren Juel Hansen - Nordic Solar | Director Energy Storage 
• Steven Jack - JLL | Head of Energy & Infrastructure Advisory EMEA 
• Laura Hoffmann-Ostenhof - Gridcog | Central Europe Industry Lead 
• Maureen Paul - Jacobs | Senior Director and Global Principal 
• Kiran Arora - Eversheds Sutherland | Partner 
• Moderated by: Alessandro Pecorari - Global Infrastructure Investor Association | Policy and Public Affairs Manager  

The key takeaways are: 

• The European grid is not broken; it was built for a world with stable demand. Adapting it to meet AI-driven load growth is, as one panellist put it, like building a plane while you are flying it.
• The core problem is coordination, which is solvable but will take time. UK data centre demand is projected to rise from 5TWh to 20-30TWh by 2030, requiring an additional 6-10GW of capacity.
• Northern wind power is being curtailed while demand goes unserved in the south. UK wind curtailment cost £1.5 billion last year, while data centres in constrained areas cannot get a grid connection at any price.
• Flexible grid connections let projects connect much sooner, and that early access is worth far more to a data centre than the occasional revenue hit of being curtailed. The curtailment risk does not vanish, but it can be priced, and that is what makes a project financeable.
• Training versus inference is the real distinction that matters for grid planning, not hyperscale versus co-location.
• Co-location helps an individual project connect, but it does not fix the wider grid problem. At the system level, it is essentially neutral, and the next load that cannot co-locate runs straight into the same constraints.
• Energy strategy is now a board-level business question, not a technical one.
• Over-centralising planning decisions suppresses the market signals that normally guide efficient investment and location choices.
• Regulators divide into two types: proactive, UK, Germany, Ireland, Lithuania, actively pushing TSOs and markets, and reactive, Denmark, Greece, waiting on century-old incumbents. Lithuania and the Baltics were flagged as surprisingly fast movers.
• Hyperscaler capex is now around $300 billion. The question is no longer whether they build, but how. The number one investor concern has shifted from access to capital to stranded assets. 
• At least 60-70% of growth will land in tertiary locations; the demand is simply too large to be absorbed by established hubs alone.

The second panel of the day explored electricity demand growth and grid constraint, examining how AI-driven load is reshaping transmission and distribution planning across Europe and the UK, and whether existing network operators can keep pace with demand clusters emerging faster than the grid was built to handle. 

Speakers included: 

• Alastair Martin - Flexitricity | Chief Strategy Officer 
• Francesc Filiberto - Nuveen Clean Energy Infrastructure | Head of Business Development 
• Matthew Brown - CMS | Partner 
• Phil Harris - PJM | Former Chairman 
• Jonathan Horne - M.P.E. Power System Consultants | Director 
• Moderated by: Biljana Stojkovska - Hitachi Energy | Senior Expert Market Innovation  

The key takeaways are: 

• The demand queue is far larger than the system itself; around 100GW of projects are waiting for connection, compared with 50GW peak load. Only 8-10GW is likely to be built by 2030, but the grid impact is large. NESO projections show electricity demand rising to 60-70TWh.
• AI training workloads can swing by around 30% within a second (0.9 to 0.6 per unit), a sharp break from historically steady demand. These rapid fluctuations create operational challenges, including voltage flicker and on weaker grids can drive oscillations with synchronous generators, threatening stability. A Finnish case showed loads jumping from 100% to 50% every two seconds in a square‑wave pattern.
• Sudden disconnection from the grid is a serious risk - losing 500 MW+ of load instantly hits both frequency and voltage. Most issues are manageable with money and good design. 
• “We are moving to a digital economy.” We have gone from an industrial age to an information age to a digital economy and “speed of power” is the main metric. Load is becoming elastic and can compete with generation in a five-minute dispatch window.
• This is a systemic connection problem prompting solutions like private micro-grids and flexible demand  
• The Spanish blackout showed the danger of big grids with little flexibility, heavy renewables, low inertia and a system designed to export renewables, not absorb big loads.
• The panel agreed that the system needs clear rules, not the absence of rules. The real problems are hidden, ambiguous, and slow rules that hold everything up. 
• Private wire is happening regardless of regulation as a long-term bridging solution to slow grid connections  
• A call to work in collaboration between TSOs, OEMs and data centres – as the wind industry did 20 years ago.
• Get large loads into security-constrained economic dispatch at the five-minute interval.
• The grid's need has shifted; what is now valuable is the slower flexibility from thermal stores.

The third panel turned to FLAP-D and beyond, examining how grid constraints are pushing digital infrastructure investment past the established hubs of Frankfurt, London, Amsterdam, Paris and Dublin toward emerging markets such as the Nordics, Vienna, Milan and Barcelona, and whether these newer locations are structurally advantaged or simply repeating FLAP-D's early-stage challenges. 

Speakers included:  

• Adam Nethersole - Kao Data | VP of Marketing
• David Kippen - Cordiant | Managing Director
• Andrea Semmola - Bird&Bird | Senior Associate
• Justin Fitzhugh - Telis Energy | CFO
• Moderated by: Beatrice Fourniere - CIBC | Executive Director, UK Corporate Banking, CIBC Capital Markets

The key takeaways are:

• The spillover out of FLAP-D is happening fast. Those markets are just too crowded, investors and developers are looking for favourable regulation.
• Power is the binding constraint. One campus, built seven years ago when people thought West London would never be constrained, has now sold all its power, and the queue for more is 2032. Operators only build with an end customer already secured, typically known 2 years before construction.
• The shift to tier 2 locations is due to power, but also due to other advantages such as: Barcelona's cable landing station, Milan's Genoa connection, Manchester/Leeds as Nordic gateways.
• Training is migrating from the US/Dublin through Manchester and Leeds straight to the Nordics, with results brought back for latency-sensitive inference. What was a latency dealbreaker in 2021 is now resolved by network architecture (split training/inference/agentic). All roads lead to Norway.
• Milan is a great benchmark of starting as a tier 2 market and becoming a tier 1 market. Italy is a strong tier 1 candidate. 
• Other hubs: Spain (transformed by cheap renewables, 4–7c/kWh vs UK/Germany at 9–13c/kWh), Poland (user-friendly regulation, fast-growing economy, the Baltics (cheap power, low construction costs), Czech Republic, Belgium.
• Renewable developers are pivoting into the DC market as off-takers or green-hub developers. 
• Projects can still be rejected if they do not deliver clear community benefits, even when they already have grid approval. 
• Scotland is a viable location in the UK, its cold/wet climate is ideal for free-air cooling, abundant stranded wind power, and consumers pay £42/year (rising to £135 by 2030) to curtail Scottish wind for lack of grid connection south. 
• Italy’s data centres are treated as strategic assets, so foreign investment must be notified to the government 
• Project finance will be much harder than for renewables.  
• The market is softening from a seller's to a buyer's market, with banks being more cautious in lending into neocloud deals.  
• Cutting hollow-core fibre latency by 60–70% means a Manchester data centre can serve London. 
• There is no single winner; the market is becoming distributed and decentralised, with Spain, the Nordics, the Baltics, Poland and Italy all named as regions to watch. 

The fourth panel turned to corporate PPAs and power procurement strategies, examining how PPA structures are evolving amid rising costs and grid congestion, and weighing the trade-offs among virtual PPAs, on-site generation, and co-location with renewable assets. 

Speakers included:  

• Semih Oztreves - Zenobe | Chief Commercial Officer(Network Infrastructure) 
• Alex Gilbert - Places for London, TfL | Head of Energy & Electrification 
• Jenny Murray - Bird & Bird | Partner 
• Kari Tikkanen - Aukera Energy | Head of Revenue 
• Moderated by: Ed Porter - Modo Energy | Global Director of Industry  

The key takeaways are:

• PPAs have moved from pure price-risk transfer to explicitly pricing delivery circumstances like negative-price hours, curtailment and capture-rate decline, and the market is now customer-led rather than take-it-or-leave-it. 
• AI training clusters were reframed as "AI factories", which shifts PPAs toward capacity and reliable power delivery rather than just energy volume.
• The "data centre flexibility" narrative is largely overstated.
• Both behind-the-meter and virtual PPAs are favoured: a long-term asset plus on-site generation/storage plus a virtual PPA used as an actively managed financial hedge.
• Islanding data centres was called a missed opportunity, as connecting large AI loads to the grid would spread generation and network costs across a wider demand base.
• A renewables PPA does not make a data centre green, and without 24/7 certificates, such environmental claims are "largely a con".
• The panel called for greater urgency from NESO, the government and especially the DNOs, which need to specify the flexibility services they require and what they will pay for them.

Oliver Carr, inspiratia's Head of Data Analysis, delivered a presentation titled "Balancing Constraint with Opportunity," offering a sneak peek into our forthcoming thought leadership report on the European data centre market.

The penultimate panel examined how investors are navigating a more complex risk landscape shaped by power constraints, market volatility, and policy uncertainty. The panel looked at how financing models are evolving to balance resilience, sustainability, and returns, how lenders and equity providers are pricing power and technology risk, and what structures are emerging to bring institutional capital into the asset class. 

Speakers included: 

• Jonathan Dames - CMS | Partner 
• Miguel Galatas Perez-Juez - nTeaser | Co-founder 
• Miguel Solana - Alter5 | Co-founder 
• Jonathan London - Akereos Capital | Head of Markets 
• Moderated by: Daniel Garcia - Glas | Country Head, Spain  

The key takeaways are: 

• Financing structures are shifting decisively from real estate-style to project finance-style: detailed cash flow modelling, layered account regimes, distribution lock-ups, and comprehensive insurance requirements are now standard, driven by the complexity of the construction market as much as lender preference.
• Fixed-price turnkey construction contracts no longer exist for data centres. Developers and lenders must now absorb target-price arrangements, indexation, force majeure for cost as well as time, and change-of-law risks being pushed back onto employers. This requires contingent facilities to be built into financing structures from the outset.
• Technology obsolescence is the underwriting risk with no clear precedent. A grid delay of two to three years can strand a data centre with the wrong generation of GPUs, making assets unable to serve the workloads the market requires. Unlike power assets with 30-year lives, the consequences of timing failure are near-immediate.
• There is a clear funding gap in the mid-market. Hyperscalers are well-served by large project finance banks; smaller and mid-sized data centres serving industrials, banks, and governments are not. Development finance and co-development agreements, hybrid debt-equity structures allowing developers to retain a stake, are the primary instruments filling that gap.
• Portfolios are emerging as the main route to institutional capital. Refinancing mini-perms into portfolio or fund financings allows lower-quality assets to be bundled with stronger ones, and gives long-term institutional lenders the scale and diversification they need to get comfortable with the asset class 
• Export credit agency-wrapped financings, NWF and UKEF, both actively providing guarantees of up to 80% of debt, represent the next significant wave of institutional participation, materially changing the risk profile of individual transactions.
• Co-location with renewables and BESS is now close to standard across European markets. On-site PV self-consumption is economically compelling given grid connection costs; BESS reduces both the cost premium associated with corporate PPAs and exposure to intermittency. The most creative structures, wind hybridised with data centre load, second-cycle heat offtake from cooling systems, are where margin is being made.
• Data centre demand flexibility is a sleeping giant. Load-shifting, throttling inference during high-price periods, and routing workloads to cheaper markets are already happening at hyperscaler scale. As utility-style pricing of compute matures, flexibility will become a core part of the financing picture rather than an afterthought.
• Sovereign and institutional demand for edge and distributed data capacity, hospitals, governments, and defence, is creating a structurally distinct market segment that will be publicly or quasi-publicly driven, largely independent of commercial return expectations.
• Geography is diversifying. Latency constraints that anchored capacity to Frankfurt, Dublin, London, and Amsterdam are weakening. Secondary and tertiary locations, northern England, Scotland, the Nordics and southern Spain are increasingly viable, with land availability, cooling conditions, and grid headroom compensating for distance from consumption centres. SMR deployment and data centre expansion are expected to develop in parallel as mutually reinforcing anchors for these emerging hubs.

The day closed with a panel on regulation, planning, and market design, examining how UK market reforms and flexibility schemes compare with EU regulatory approaches, what practical lessons each can draw from the other's planning and permitting frameworks, and how evolving regulations are affecting project feasibility, timelines, and risk allocation. 

Speakers included: 

• Rosa Rotko - Mott Macdonald | Project Director Energy Transformation 
• James Hill - Flint Global | Director 
• Patrick Hayes - Slough Borough Council | Executive Director of Housing & Property 
• Jeffrey Altman - Finadvice | Senior Advisor 
• Moderated by: Maya Chavvakula - Inspiratia | Head of News

The key takeaways are: 

• The west London grid is congested, and upgrading older racks to higher-density AI workloads can triple or quadruple a site's demand overnight, each upgrade triggering a fresh grid capacity assessment.
• The public framing of "data centres vs housing vs Heathrow" is a false trilemma. The real policy gap is insufficient generation capacity for all three; picking winners between legitimate infrastructure users avoids the harder question of how to decarbonise and expand the grid.
• NIMBY pressure operates at two levels simultaneously: local residents with immediate amenity concerns, and national campaign groups waiting to challenge planning decisions through judicial review. Both need to be managed, and the two require different responses.
• UK regulatory reform is genuinely moving, Ofgem and NESO are progressing queue reform, and a new independent transmission owner/operator licence that could allow data centres to self-build or commission their own high-voltage connections, but political policy has not kept pace.
• The government's AI growth zones framework contains a near-circular contradiction: DESNZ requires 95% renewable connection to qualify, but the reason developers seek growth zone status is precisely that they cannot secure a grid connection in the first place.
• Cheap power, available industrial land, fast permitting, and ready grid access are the four factors driving location decisions. The EU's broader framing of data infrastructure as a matter of digital sovereignty is shifting the terms of competition further.
• Data sovereignty and undersea cable vulnerability are hardening the case for domestic capacity independently of cost. The resilience and national security arguments for keeping data onshore are increasingly cross-cutting across government, financial services, and defence.
• The UK has two genuine trump cards: its world-leading nuclear regulatory framework, which attracts international SMR vendors for generic design assessments, and the emerging independent transmission licence, which could unlock private capital for grid buildout alongside data centre development.
• Location selection and community consultation are decisive. Planning outcomes turn on political management as much as technical merit. Waste heat recovery piped to nearby homes, local employment commitments, and early community engagement have all proven effective in converting opposition into acceptance.