Treating both an application for a grid connection and the completion of actual power delivery as having secured power makes it difficult to assess the operating risks of an AI data center properly.
Revenue losses and contractual liabilities arising when the operating date promised to customers does not align with the grid connection date may remain with the project.
Companies should reflect the stage of securing power and acceptable delays in their investment conditions, and link these to asset valuation and investor reporting.
As demand for AI computing grows, data center investment is expanding into a long-term undertaking that requires power, cooling and telecommunications networks to be secured together. According to a white paper citing the International Energy Agency’s (IEA) 2026 outlook, global data center electricity consumption is expected to rise from 485 terawatt-hours (TWh) in 2025 to about 950 TWh in 2030. Electricity consumption at AI-focused data centers was expected to triple over the same period.
Data center electricity demand grew by 17% in 2025. Although data centers are expected to account for about 3% of total electricity demand in 2030, the white paper did not conclude from this figure alone that there would be a global power crisis. It placed greater emphasis on the concentration of data centers in particular electricity markets and the resulting conflicts with regional generation, transmission and substation capacity.
Regardless of how much computing demand increases, investors still need to establish whether individual projects can begin operating when required. Power generation and grid reinforcement require land, permits, equipment and skilled personnel, and these processes may follow a different schedule from data center construction. Planning capital deployment and customer occupancy requires checking the readiness of infrastructure outside the building as well.
Global investor services provider IQ-EQ, led by CEO Mark Pesco, and the Sustainable Real Estate Forum, referred to below as the Forum, published the white paper “Powering intelligence: Can data centres and the energy transition coexist?”
The white paper says that turning AI demand into actual investment returns requires scrutiny of power delivery timing, contractual rights, operational resilience and delay costs together. Rather than asserting that energy risks have been mispriced at all data centers, it warns that some conditions affecting asset values lie beyond an individual facility’s control. It recommends reflecting power status in investment eligibility, asset valuation and investor notifications, and planning energy and data center development together.
The document is not a report based on its own survey, but a white paper bringing together analysis from investment, insurance, legal and professional organizations and an operator. Contributors included IQ-EQ, the Forum, Zurich Insurance, Norton Rose Fulbright, the Royal Institution of Chartered Surveyors and Start Campus.
Construction takes 2 to 3 years, while grid connections take 7 to 10 years: delay liabilities remain with the project
Citing IEA material, the white paper explained that data center construction takes 2 to 3 years, while the associated energy infrastructure requires more time. According to 2025 material from energy research organization Ember, new data centers at major European hubs, including Frankfurt, London, Amsterdam, Paris and Dublin, may wait an average of 7 to 10 years for a grid connection, with some facing waits of up to 13 years.
The meaning of having secured power also differs by stage. A connection application, provisional capacity allocation, signed connection agreement, funding for grid reinforcement, contracted generation output, and completion of actual supply and testing each involve different rights and execution risks. The white paper noted that presenting these as a single yes-or-no measure of whether power has been secured makes it difficult to identify remaining work and conditions, or the effects of schedule changes.
Tamas Mark, global head of real assets at IQ-EQ, takes the view that “fiber-optic networks, land, tax efficiency and customer demand alone are insufficient to assess a location’s investment appeal. Investors should verify whether power can be delivered on the required schedule and make energy due diligence a condition of investment approval.”
Norton Rose Fulbright, which provided the legal analysis, explained: “There is a timing gap between customer contracts and grid contracts. Leases with large cloud providers specify dates for delivering capacity, and failure to meet them can result in damages, contract termination and lost revenue. By contrast, grid connection terms in continental Europe are often set by regulation, leaving little room for negotiation, and compensation for delays is rare.”
Some risks can be transferred to generators, power purchase agreement (PPA) counterparties, or engineering, procurement and construction contractors. However, if liabilities arising from customer contracts exceed what the counterparty can bear, energy project costs may rise or financing may become difficult. Signing a contract should be distinguished from concluding that delay risk has been adequately transferred.
Management is still required after power capacity has been allocated. According to the white paper, contracted data center demand in the UK grid connection queue was about 73 gigawatts (GW) in mid-2025, and reforms assess project readiness. In the Netherlands, contracted capacity that goes unused for around 2 years can generally be withdrawn, making it difficult to assume that spare capacity reserved for future expansion will remain available indefinitely.
The European Union’s (EU) proposed Cloud and AI Development Act sets a target of tripling data center capacity within 5 to 7 years and calls for available grid capacity, storage and clean generation to be considered when designating acceleration zones. The white paper noted that the proposal does not impose equivalent connection deadlines on grid operators. Designating an acceleration zone and confirming an actual power delivery schedule remain separate matters.
Power failures can spread to cooling and controls: insurance, carbon and skilled personnel also require review
Larger projects also increase the extent of the effects of delays or failures. Roger Baumann, global real estate chief operating officer and head of product development at Zurich Insurance, said the average value of data center projects the company underwrites had risen from $150 million 5 years ago to $3 billion currently. These figures refer to Zurich Insurance’s underwriting portfolio, not the average for the entire market.
Zurich Insurance’s review of major global data center outages over the past 10 years found that 82% of the incidents analyzed originated in electrical or power system failures. More than half involved cascading failures in which the initial problem spread to a loss of cooling or operational instability, and more than one-third involved human or procedural factors. These percentages cannot be added together because causes may overlap.
A power interruption can reduce cooling performance and destabilize temperatures, while control system responses or operator intervention can amplify the failure. The white paper explained that inspections should cover not only individual components such as generators and protection devices, but also the interfaces between power, cooling, controls, communications and operating procedures. Design, commissioning, isolation of faulty sections, diversification of cooling paths and personnel training all require review.
Taking out insurance does not eliminate all these risks. Coverage may be limited for interruptions without physical damage or external dependencies beyond a developer’s control. The white paper explained that loss scenarios, controls, insurance terms and residual risks should be identified from the early design and construction stages so that lenders and investors can also assess the extent of their exposure.
The Royal Institution of Chartered Surveyors (RICS) emphasized carbon measurement and the execution capabilities of skilled personnel alongside access to capital. It recommended comparing costs and carbon together during early design, when material quantities, equipment specifications and construction methods can still be changed, and specifying the scope of coverage for buildings, electrical and cooling equipment, replacement cycles and customers’ information technology equipment. It explained that carbon figures based on different assessment periods or scopes should not be compared directly.
The construction and real estate surveys cited by RICS were not surveys exclusively of data centers. The white paper used them as a basis for examining skilled personnel, productivity measurement and the gap between AI trials and routine workplace adoption. It explained that, beyond securing budgets, data center developers need to check the availability of technical personnel required for design, installation, testing and handover, and how contractors coordinate their work.
As an example of actual operations, the white paper introduced the SINES Data Campus in Sines, Portugal, operated by data center developer and operator Start Campus. Its first building, SIN01, operated without interruption during the Iberian Peninsula grid blackout on April 28, 2025. However, the white paper described this as a record of a single event under specific conditions, rather than extending it into proof of uninterrupted operation in all future circumstances.
The facility reuses the water intake and discharge infrastructure of a nearby closed power plant for seawater cooling. The white paper also specified that the performance disclosed by the operator concerned the first building and was compiled internally without external verification. The campus’s power usage effectiveness (PUE) of 1.1 and water usage effectiveness (WUE) of 0 are design targets, distinguished from operating results. The benefits of reusing existing facilities should likewise be assessed only to the extent that remaining service life, maintenance and permits have been checked.
From investment approval to asset valuation: 5 management criteria reflecting the status of power access
IQ-EQ recommended managing power risks throughout an asset’s holding period, rather than checking them only once at investment approval. Its recommendations link rights, responsibilities and reporting criteria across 5 areas.
1. For capital and ownership, investors should establish which legal entity holds the land, buildings and grid connection rights, and where debt and development and operating obligations sit. This includes co-investors’ rights and responsibilities among funds, developers and operators. If rights are divided among several special purpose vehicle (SPV) entities, investors should identify contractual counterparties, cash movements and each entity’s obligations.
2. For the start of power supply, investors should distinguish currently usable capacity from conditional capacity, and identify remaining work and the parties responsible. The recommendations call for fund documents to specify minimum requirements for securing grid access rights and the maximum acceptable delay in supply. Investment due diligence should review the legal status of connection rights, queue position, contractual certainty, local grid constraints and realistic delivery schedules.
Net asset value (NAV) assessments should distinguish assets where actual power supply has started, assets with contracted capacity awaiting supply, and assets still in the connection queue. Because risks differ by power supply stage even when construction progress is the same, the recommendations call for these differences to be recorded in valuation assumptions and verification records. Supply delays can change not only revenue but also the timing of capital calls and financing costs.
3. For contractual dependencies, investors should identify who bears power price fluctuations, grid reinforcement costs, supply delays and delays in customers taking up capacity. Renewable energy procurement agreements should also be distinguished according to whether they provide accounting recognition, financial price settlement or physical delivery. Investors should also examine whether rights and obligations remain in force after refinancing or a change of ownership.
4. For resource and resilience performance, investors should review power usage effectiveness, water usage effectiveness, carbon usage effectiveness (CUE), the share of renewable energy procurement, and grid availability and reliability together. Power usage effectiveness is total facility energy consumption divided by information technology equipment energy consumption; a low value does not necessarily mean that the electricity also has low carbon intensity.
5. To improve investor transparency, capital providers should be able to understand how changes in power conditions affect cash flow and value. The recommendations call for notifying them when supply schedules exceed agreed thresholds or contracted power capacity cannot be secured, when power prices or availability change significantly, or when renewable energy procurement commitments cannot be met. Annual reporting alone may be insufficient for assets under development or those facing rapidly changing grid conditions.
Citing the IEA’s 2025 estimates, the Forum explained that about 20% of planned data center projects worldwide were at risk of delays because of grid constraints. The figure is not a count of the share of projects already delayed. It therefore recommended preparing a regional infrastructure delivery specification alongside the asset’s business plan.
The specification should record required grid works, the parties responsible for costs, key equipment schedules and the effects of delays. If batteries or adjustments to computing workloads provide grid flexibility, it should specify the capacity actually provided, response time, duration, operating constraints and authority to activate the service. Technically possible functions should be distinguished from services that can be delivered under contract.
Peter Panayi, founder and CEO of the Forum, emphasized: “Turning AI demand into sustained cash flow requires the capability to actually complete and operate infrastructure. Pursuing data centers and the energy transition together requires linking design, financing, contracts and operations, and continuously checking the external conditions that determine power supply.”
Original Korean article: gttkorea.com
English brief by Korea Wind Intelligence, koreawindintel.com
