London, UK — In the relentless gold rush of the artificial intelligence era, a profound paradigm shift is sweeping through the digital infrastructure landscape. For decades, data centre development was dictated by a singular gospel: ruthless cost optimization. Procurement teams honed their leverage to squeeze margins, negotiate volume discounts, and drive down capital expenditures.
However, at the Capacity Europe 2026 conference in London, industry leaders delivered a sobering reality check. During the headline panel, "Scaling Under Scarcity: Delivering Projects Amid Shortages," held on October 14 on the AI Build Stage, top engineering and procurement executives made it clear that cost is no longer king. In a market crippled by unprecedented supply chain bottlenecks, labor deficits, and explosive AI-driven demand, manufacturing capacity has become the ultimate currency.
If you cannot secure a production slot, your capital is worthless, your designs are obsolete, and your ready-for-service (RFS) deadlines will be missed. Data centre builders are now forced to order multi-million-dollar generators two years in advance, commit to suppliers before architectural blueprints are finalized, and rethink the very fundamentals of how digital infrastructure is conceived, procured, and deployed.
Executive Overview: The Death of Traditional Procurement
The convergence of generative AI, large language models (LLMs), and hyper-scale cloud expansion has created an unprecedented infrastructure crunch. As tech giants and enterprise firms race to secure compute power, the physical supply chain underpinning the internet has hit a brick wall.
The symptoms of this systemic scarcity are visible across the globe:
- The Two-Year Wait: Essential electrical infrastructure, such as backup generators and high-voltage transformers, now face lead times stretching up to 24 months.
- The "Procure First" Mandate: Operators are locking in manufacturing capacity blindly, agreeing to framework contracts before knowing the exact technical specifications of what they are purchasing.
- Time as Currency: Industry leaders are redefining "savings" not in monetary terms, but in days—calculating that paying a premium to accelerate deployment yields far greater business value than haggling over unit prices.
- The Human Deficit: Beyond hardware, a severe shortage of skilled mechanical, electrical, and plumbing (MEP) tradespeople and engineering talent threatens to choke project delivery through the end of the decade.
This comprehensive report examines the structural shifts transforming data centre development, drawing insights from key industry figures at Greenscale, Start Campus, Telehouse, and Kentik.
Detailed Chronology: How Scarcity Upended the Development Lifecycle
To understand how the data centre sector reached this tipping point, one must examine the timeline of operational adjustments forced upon developers over the past several cycles.
Phase 1: The Recognition of the Bottleneck (Early 2024 – Late 2025)
For years, grid connectivity was recognized as the primary gating factor for digital infrastructure. However, as hyperscalers accelerated their AI clusters, equipment manufacturing quietly reached saturation. Grid operators and hyperscalers began sounding alarms regarding high-voltage transformers (HVTs) and gas-insulated switchgear (GIS). What started as a localized component delay snowballed into a global hardware famine.
Phase 2: The Shift to Blind Commitments (2026)
By Capacity Europe 2026, the strategy had evolved from reactive expediting to preemptive hoarding. Miriam van Kooperen, Senior Director of Supply Chain at Greenscale, articulated the industry’s anxiety during the London panel:
"We’re worried about cost and lead times, where I believe we should be worried about capacity."
Operators realized that waiting for final engineering designs before engaging manufacturers was a fatal error. Factories were operating at full capacity, and hesitation meant losing a production slot to a rival hyperscaler capable of gobbling up hundreds of units in a single sweep.
Phase 3: Redesigning the Workflow ("Procure First, Design Around It")
Faced with a landscape where AI deployment methodologies shift roughly every six months—while a fast-tracked data centre takes 18 months to build—traditional sequential project management collapsed.
Enterprises like Start Campus, currently developing a massive hyperscale campus in Portugal, completely inverted their operational models. Nelson Pinto de Magalhaes, Director of PMO and Procurement at Start Campus, explained the firm’s new mantra: "Procure first, design around it."
Because factories cannot be compressed to produce more units on demand, procurement teams have taken the driver’s seat. They secure long-lead items based on standardized frameworks, forcing design and engineering teams to adapt their blueprints to fit the hardware that is actually obtainable.
Supporting Context & Metrics: The Anatomy of a Supply Squeeze
The challenges facing data centre developers are multi-layered, encompassing heavy equipment lead times, grid connection delays, and severe human resource constraints.
The Hardware Deficit
The most acute physical bottlenecks remain concentrated in power generation and distribution.
- Generators: Lead times average up to 2 years. Ozgur Duzgunoglu, Head of Engineering and Design at Telehouse, recounted an instance where his team faced a Friday deadline to order generators that would not arrive for 24 months, simply to protect their production slots against predatory hyperscaler orders.
- High-Voltage Transformers: Sourcing HVTs now routinely requires 12 to 18 months, depending on regional manufacturing capacity.
- Liquid Cooling Units: To keep pace with high-density AI chips, operators are increasingly relying on off-site prefabrication for coolant distribution units (CDUs), shifting installation from laborious on-site assembly to rapid plug-and-play deployment.
The Power Paradox and Legal Fallout
Securing physical equipment is only half the battle; getting power to the site remains a high-stakes gamble. In Frankfurt, acquiring grid power can take five to seven years, forcing developers to carefully weigh regional availability against customer demand for central connectivity hubs like London.
The pressure has even triggered severe legal maneuverings. Oracle recently issued a force majeure notice regarding Project Jupiter—its massive Stargate AI campus in New Mexico—anticipating potential delays in its local power supply ahead of its targeted 2028 operational launch. Other operators are resorting to running backup diesel generators continuously to bring facilities online before permanent grid connections are established.
The Talent Crunch
As noted by Steve Walton, CEO of 3C Datalec, the shortfall of engineers and specialized tradespeople is every bit as restrictive as power and equipment shortages. Clients, contractors, and suppliers are locked in a fierce, circular competition for the same finite pool of MEP professionals. Furthermore, as developers push into new geographic regions rich in green power, they frequently encounter local labor markets entirely devoid of data centre construction experience.
Official Statements & Industry Perspectives
The panels and interviews at Capacity Europe 2026 highlighted a unified front among forward-thinking infrastructure leaders:
- Miriam van Kooperen (Greenscale): Emphasized the transition from product-based to performance-based specifications. Rather than specifying a single brand of equipment that may face indefinite delays, operators are specifying performance parameters so they can swap vendors dynamically if preferred suppliers stall.
- Nelson Pinto de Magalhaes (Start Campus): Challenged the traditional metrics of procurement success, reframing financial savings into temporal gains: "What if everything that we called savings now is days? You are judged on whether you hit your ready-for-service date, not on which component arrived late."
- Ozgur Duzgunoglu (Telehouse): Highlighted the financial logic of prioritizing speed over cost: saving a single day brings revenue forward, accelerates the return on borrowed capital, and offsets minor cost premiums paid for expedited delivery.
- Justin Ryburn (Kentik): Addressed the downstream impact on smaller players who lease cages in multi-tenant facilities. With data centre providers increasingly selling space by the floor, niche operators are being forced to re-architect their software to run directly on hyperscale clouds that already possess locked-in capacity.
Future Outlook: Navigating Toward 2029
As the digital infrastructure community looks ahead to the remainder of the decade, the consensus is clear: relief is not arriving anytime soon.
Looking toward 2029, the primary constraints on AI expansion will remain inextricably linked to power and people.
- The Persistent Power Wall: Grid infrastructure upgrades take years to materialize, ensuring that power scarcity will dictate market entry strategies well into the next decade.
- The Battle for Human Capital: The scarcity of MEP talent and specialized commissioning engineers will separate successful operators from stalled projects. Those who build robust training pipelines and foster collaborative regional supply chains will thrive.
- Architectural Agility: Standardized building blocks, performance-based procurement, and modular prefabrication will transition from "nice-to-haves" to existential survival mechanisms.
Ultimately, Capacity Europe 2026 signaled the permanent maturation of the data centre industry. In an age where artificial intelligence promises to reshape global economies, the companies that win will not be those that bargain the hardest for a discount, but those that secure their place at the front of the manufacturing queue.
