The Infrastructure Behind AI | Edition 3
Billions are being committed to AI infrastructure. But investment alone does not guarantee operational capacity — or a commercial return.
The scale of investment in artificial intelligence is extraordinary. Technology companies are committing vast sums to computing infrastructure, while developers, utilities, engineering contractors and equipment manufacturers are responding to a rapidly expanding pipeline of capital projects.
Yet beneath the excitement surrounding AI lies a fundamental question.
Can the physical infrastructure be delivered quickly enough, reliably enough and economically enough to justify the investment?
It is a question that goes well beyond data centre construction. It reaches into electricity generation, transmission networks, planning approvals, substations, specialist equipment, construction capacity, commissioning and ultimately the commercial assumptions underpinning the investment itself.
As the scale and complexity of these programmes increase, so does the importance of understanding what can actually be delivered, rather than simply what has been announced. And that is where Planning & Project Controls become particularly important.
The investment numbers are difficult to ignore
Consider some of the commitments being made.
In its second-quarter 2026 results, Meta forecast capital expenditure of between $130 billion and $145 billion for the year, including principal payments on finance leases.
Alphabet reported $80.6 billion of capital expenditure in the first six months of 2026, more than double the equivalent period in 2025.
These figures are not exclusively for constructing data centres. They include computing equipment, technical infrastructure and other capital investment. Nevertheless, they illustrate the enormous scale of resources being committed to AI and its supporting infrastructure.
The energy implications are similarly significant.
The International Energy Agency’s Energy and AI report projected that global data centre electricity consumption could more than double to approximately 945 terawatt-hours by 2030, driven substantially by AI.
Research published by Lawrence Berkeley National Laboratory in June 2026 estimated that data centres could account for 11.8% of US electricity consumption by 2030, with scenarios ranging from 9.5% to 15.3%.
These are projections rather than certainties, but they demonstrate the magnitude of the infrastructure challenge.
There is also an important distinction between investment commitments, construction starts and operational capacity. A project can be funded without being ready to build. It can be under construction without having secured its required electricity supply. And it can be physically complete without being commercially operational.
The value of the investment ultimately depends on bringing usable capacity into service, not simply on spending the capital.
When the critical path sits outside the construction site
Our previous article examined how access to electricity is becoming one of the defining constraints on AI infrastructure development. But the wider delivery challenge is more complicated.
Imagine a major data centre campus progressing broadly to schedule. The buildings are taking shape. Mechanical and electrical installation is advancing. Equipment is arriving and contractors are reporting progress against their individual programmes.
Yet the substation required to energise the campus is six months behind schedule. Perhaps a transformer has been delayed, a transmission upgrade is awaiting approval, or a third-party utility programme has slipped.
The construction project may appear successful. The investment programme is not. This is not simply a scheduling problem. It is a failure to manage the dependencies that determine operational readiness.
The IEA identified the seriousness of this challenge in its 2025 research, estimating that around 20% of planned data centre projects could be at risk of delays because of grid constraints, unless those risks are addressed. It also highlighted that new transmission lines in advanced economies can take four to eight years to deliver.
A delayed grid connection may affect more than the opening date. It can influence financing costs, contractor commitments, equipment warranties, resource allocation and when the development begins generating revenue.
For the owner or investor, the most important milestone may not be construction completion. It may be the date on which the facility is energised, commissioned, accepted and able to provide the capacity on which its business case depends.
When permission, people and equipment become the critical path
These risks are not theoretical. Recent developments in the United States demonstrate how ambitious AI infrastructure plans can encounter obstacles beyond the construction site.
In Virginia, the proposed Prince William Digital Gateway development, potentially accommodating up to 37 data centres, was abandoned after its rezoning approvals were overturned following legal challenges concerning statutory public-notice requirements.
Compass Datacenters withdrew in April 2026, with its president acknowledging that the company could no longer proceed. On 2 July, fellow developer QTS withdrew its final legal appeal, effectively bringing the entire project to an end.
This was not simply a construction delay. A proposed multibillion-dollar development was stopped altogether, demonstrating how planning compliance, legal challenges and community opposition can ultimately determine whether a major infrastructure investment proceeds at all.
In North Carolina, Raleigh City Council introduced a six-month moratorium on new major data centre development applications in October 2026, allowing the city to review regulations concerning water consumption, noise, environmental impacts and public resources.
Raleigh is not alone in confronting these questions. Other North Carolina communities have also introduced temporary restrictions while reconsidering how major data centre developments should be managed.
Neither example demonstrates a shortage of investor interest. Both illustrate how regulatory compliance, community concerns and planning uncertainty can materially affect the development timetable — or determine whether projects proceed at all.
Meanwhile, the pressures extend well beyond planning approvals.
According to JLL’s 2026 Global Data Center Outlook, 57% of data centre projects globally experienced construction delays of three months or more during 2025.
The same research found that average data centre equipment lead times in the US remained at approximately 42 weeks, 83% above 2019 levels.
Turner & Townsend’s Global Construction Market Intelligence 2026 adds another dimension. Across the 112 markets examined, more than 70% reported skills shortages, while 87% reported shortages in specialist mechanical, electrical and plumbing trades.
The report also identified data centres as the most constrained construction sector globally for contractor capacity, with more than 70% of markets reporting tightening or overstretched resources.
These are global findings rather than US-specific statistics, but they highlight a problem facing American developers competing for the same specialist skills, equipment and delivery expertise.
For investors and project owners, the question becomes increasingly difficult.
How reliable is a completion date when so many critical activities depend on approvals, utilities, suppliers and specialist resources beyond the direct control of the construction team?
A project may have sufficient capital, a capable main contractor and a strong commercial case. Without those dependencies being properly understood and managed, it may still fail to deliver when the investment requires it.
What happens if the financial assumptions change?
There is another dimension to the infrastructure boom that deserves attention. Much of the current investment is being justified by expectations of sustained growth in demand for AI computing capacity.
Those expectations may prove correct. AI adoption is already influencing investment across a wide range of industries. However, the relationship between capital expenditure and future financial returns remains uncertain.
What happens if computing efficiency improves faster than expected? What if customers become more selective about the price they are prepared to pay for AI services? What if funding becomes more expensive, or investors demand clearer evidence of returns before committing to the next phase of expansion?
None of these possibilities necessarily means that demand for AI infrastructure disappears. But they could change the timing, scale, sequencing and economics of individual projects.
For developers and delivery partners, a slowdown or reprioritisation could create significant challenges. Procurement commitments may already have been made. Contractors may have mobilised. Long-lead equipment may have been ordered. Infrastructure upgrades may be progressing against an assumed energisation date.
Projects could face deferment, redesign, phased delivery or cancellation. Equally, organisations with the strongest commercial cases and most credible delivery programmes may be better positioned to attract investment when capital becomes more selective.
This is where programme governance, schedule assurance and risk management connect directly with financial decision-making.
The question is not simply whether the project can be built. It is whether it can be built at the right time, for the right cost, against a business case that still makes sense.
From managing projects to managing investment risk
Planning and Project Controls are frequently associated with schedules, cost reports and progress dashboards. Those activities matter, but they are only part of the discipline.
On major AI infrastructure programmes, the real value lies in providing decision-makers with a reliable view of what is achievable, where the risks sit and what actions can protect the outcome.
That requires an understanding of the complete delivery chain. Generation and grid infrastructure must align with site development. Design decisions must be reflected in procurement plans. Long-lead equipment must arrive when installation teams can use it. Construction completion must align with testing, commissioning and operational acceptance.
And those dependencies cannot always be controlled by the same organisation. An effective integrated programme needs to identify the interfaces, challenge assumptions and establish ownership of the critical activities.
It must also recognise that the critical path can change. An equipment shortage may initially be the dominant risk. Once resolved, the constraint may move to specialist installation capacity, an external approval or the commissioning programme.
Strong Project Controls professionals do not simply report these changes after they happen. They help anticipate them, evaluate their consequences and give management time to respond.
That capability is particularly valuable when investment decisions have to be made before every element of the delivery programme is certain.
The experience question
There is a people dimension to this challenge which is sometimes overlooked. As more organisations compete to develop data centres, power generation, transmission infrastructure and advanced manufacturing facilities, they are also competing for people with experience of delivering complex capital programmes.
Finding somebody who knows Primavera P6 or can produce a professional-looking dashboard is not necessarily the same as finding somebody who can challenge an unrealistic programme. The difference often lies in experience and judgement.
Can the individual recognise when a contractor’s schedule lacks credible logic? Can they understand how a late design decision might affect procurement, installation and commissioning? Can they identify risks concealed within apparently satisfactory progress reports? And can they communicate those risks effectively enough to influence senior management decisions?
These are not simply technical questions. They are questions about the ability to protect very substantial capital investment.
Technology, including AI, will undoubtedly improve aspects of planning, reporting and analysis. But better information does not automatically produce better decisions. Experienced people are still needed to interpret the information, test assumptions and understand the practical realities of delivery.
As capital programmes become more complex and interconnected, that judgement becomes increasingly valuable.
The next competitive advantage may be delivery certainty
There is no shortage of ambition in AI infrastructure. The investment is real, the technological developments are significant and the opportunities for the companies involved extend well beyond the data centre sector.
But large-scale infrastructure delivery has never been determined by capital availability alone. Projects depend on complex relationships between financing, engineering, procurement, construction, utilities, people and regulatory approvals. Each additional dependency introduces another potential source of risk.
The cancellation of Virginia’s Prince William Digital Gateway and the introduction of development restrictions in North Carolina demonstrate that planning and regulatory risks can do far more than disrupt schedules. They can determine whether projects proceed at all.
Evidence from JLL and Turner & Townsend also demonstrates that shortages of equipment, contractor capacity and specialist skills are already affecting the delivery environment.
Some organisations will navigate those challenges more successfully than others. Those with well-integrated programmes, realistic schedules, effective risk management and experienced delivery teams should be better placed to make informed decisions, preserve flexibility and protect their investment.
Those relying primarily on ambitious completion dates and fragmented reporting may discover problems much later than they should.
The AI infrastructure boom will undoubtedly create opportunities for technology companies, investors, utilities, developers and contractors. But the real measure of success will not be how much money is announced or how many developments begin construction.
It will be how much dependable capacity can be brought into operation, when it is needed, and at a commercially sustainable cost.
The winners in AI infrastructure may not simply be those who invest the most. They may be those who can turn investment into operational capacity with the greatest certainty.
The Infrastructure Behind AI is a series exploring the physical infrastructure, investment and project delivery challenges supporting the growth of artificial intelligence.
Colin Lapthorn | Osirian Consulting
Osirian has specialised in Planning, Scheduling and Project Controls recruitment for complex capital programmes for over 20 years, supporting organisations where experienced delivery professionals make a material difference.
Sources and further reading
- Meta — Second Quarter 2026 Results
- Alphabet — June 2026 Form 10-Q
- International Energy Agency — Energy and AI
- Lawrence Berkeley National Laboratory — US Data Center Energy Usage Report 2025 Update
- Virginia Business — Prince William Digital Gateway Data Center Project Officially Dies 6 July 2026
- Prince William County — Digital Gateway Planning Status
- City of Raleigh — Data Centers FAQ and Regulatory Updates
- JLL — 2026 Global Data Center Outlook
- Turner & Townsend — Global Construction Market Intelligence 2026
