London, UK — In the high-stakes gold rush of the artificial intelligence era, a profound paradigm shift is rewriting the rules of industrial development. For decades, the digital infrastructure sector operated on a predictable gospel of commercial optimization: cut costs, squeeze suppliers, negotiate hard on unit prices, and engineer every penny out of a bill of materials.
Today, that playbook is obsolete.
At the flagship Capacity Europe 2026 conference in London, industry leaders gathered on the AI Build Stage for a panel titled "Scaling Under Scarcity: Delivering Projects Amid Shortages." The consensus delivered by procurement and engineering heavyweights from Greenscale, Start Campus, Telehouse, and Kentik was stark, unequivocal, and entirely disruptive to traditional business logic: manufacturing capacity, not price, now dictates whether multi-billion-dollar AI infrastructure projects cross the finish line on time.
As hyperscalers and enterprise operators race to secure compute power, data centre builders are increasingly forced to order critical components up to two years in advance. They must make binding financial commitments to suppliers long before final architectural blueprints are drawn. In this hyper-accelerated market, speed has replaced cost as the ultimate competitive currency. Every day saved in delivery translates directly into accelerated revenue and maximized return on capital, rendering nominal equipment savings irrelevant if a facility misses its ready-for-service (RFS) window.
Executive Overview
The rapid ascent of generative AI and massive large-language-model (LLM) training clusters has created an unprecedented demand shock across the global supply chain. Data centres—once built on predictable 24-to-36-month timelines—are now subject to severe structural bottlenecks. High-voltage transformers, gas-insulated switchgear, generator engines, and specialized cooling distribution units have transformed from off-the-shelf commodities into scarce allocations controlled by a handful of global manufacturers.
During the London panels, industry veterans laid bare the realities of building under pressure:
- The Two-Year Lead-Time Reality: Essential power infrastructure components, such as backup generators and high-voltage substation equipment, now require ordering cycles extending up to 24 months.
- The "Procure First, Design Around It" Doctrine: Operators are abandoning traditional sequential workflows. Procurement teams now secure manufacturing slots blindly, forcing engineering divisions to adapt facility designs around whatever hardware can actually be delivered.
- The Human Capital Crisis: Equipment scarcity is only half the battle. A critical shortage of skilled mechanical, electrical, and plumbing (MEP) tradespeople, designers, and commissioning engineers threatens to stall projects even when components arrive on site.
- Grid Bottlenecks and Contractual Risk: Long delays in securing municipal and national grid connections have turned power delivery into a legal hazard, prompting emerging risk-mitigation measures like force majeure filings on multi-billion-dollar mega-campuses.
Detailed Chronology: How the Scarcity Crisis Unfolded
To understand how the data centre industry arrived at this precipice, one must examine the compounding pressures that began accelerating in the wake of the generative AI boom.
Phase 1: The AI Demand Shock (2023–2024)
When commercialized generative AI models captured global enterprise markets, the demand for GPU-dense computing clusters skyrocketed overnight. Hyperscalers—including Microsoft, Google, AWS, and Oracle—began gobbling up real estate and ordering servers at an exponential rate. Initially, the primary bottleneck was silicon: securing high-end NVIDIA and AMD accelerators. However, as compute clusters scaled from megawatts to hundreds of megawatts, the physical infrastructure required to power and cool these chips quickly became the primary chokepoint.
Phase 2: The Equipment Backlog (2025)
By 2025, the supply chain for heavy electrical equipment cracked under the pressure. Manufacturers of high-voltage transformers and switchgear found their order books completely filled for years in advance. Traditional procurement strategies failed. Operators accustomed to aggressive vendor bidding wars suddenly found themselves pushed to the back of the queue. Grid operators, already straining under the weight of industrial electrification and renewable energy transitions, began quoting connection timelines stretching past five to seven years in major hubs like Frankfurt and London.
Phase 3: The Operational Pivot at Capacity Europe 2026 (October 2026)
Marking the milestone 25th anniversary era of the conference, Capacity Europe 2026 crystallized the industry’s collective realization: the bottlenecks are permanent structural features of the mid-2020s landscape. On October 14, 2026, industry leaders articulated a new operational reality. Procurement is no longer about commercial negotiation; it is about industrial survival and timeline defense.
Supporting Context & Metrics: Inside the Supply Chain Squeeze
The numbers and operational constraints shared by panellists in London illustrate the sheer magnitude of the hurdles facing modern digital infrastructure builders.
Component Lead Times at a Glance
- Backup Generators: 18 to 24 months. Procurement teams have reported facing strict Friday afternoon deadlines to authorize multi-million-dollar generator orders that will not materialize on site until two full calendar years later.
- High-Voltage Transformers: 12 to 18 months, heavily dependent on global factory allocations and raw material availability (such as grain-oriented electrical steel).
- Grid Interconnection: 3 to 7 years in saturated Tier-1 markets like Frankfurt, Amsterdam, and central London, forcing developers to look toward secondary and tertiary markets where power is more readily accessible.
The Innovation Cycle vs. Construction Velocity
One of the most profound friction points highlighted during the panel is the temporal mismatch between technology and construction. As Nelson Pinto de Magalhaes, director of PMO and procurement at Start Campus, pointed out, the architecture and thermal density requirements of AI deployment shift roughly every six months. Conversely, even a fast-tracked hyperscale data centre campus takes a minimum of 18 months to break ground and operationalize.
This creates a high-stakes guessing game: operators must commit to physical footprints and cooling topologies today that may need to support entirely different hardware specifications by the time the facility opens its doors.
Official Statements and Industry Insights
The panel discussions at Capacity Europe featured sharp, candid assessments from leaders on the front lines of global digital infrastructure deployment.
"We’re worried about cost and lead times, where I believe we should be worried about capacity."
— Miriam van Kooperen, Senior Director of Supply Chain, Greenscale
Van Kooperen emphasized that the modern supply chain environment demands a radical rethinking of procurement strategy. Rather than holding out for the best commercial terms, supply chain directors must secure production slots with preferred vendors. To achieve this, Greenscale has moved toward securing manufacturing capacity "without knowing exactly what it is we will be buying," setting strict contractual cut-off dates where designs must be frozen to keep factories humming.
"Procure first, design around it."
— Nelson Pinto de Magalhaes, Director, PMO and Procurement, Start Campus
Pinto de Magalhaes detailed how Start Campus—currently developing a massive hyperscale campus in Sines, Portugal—has upended traditional corporate hierarchies by integrating procurement and design teams under procurement leadership. Recognizing that "you can’t really compress a factory to produce more," Start Campus treats time as the ultimate metric of savings. "What if everything that we called savings now is days?" he asked the audience, noting that customers judge operators strictly by whether ready-for-service (RFS) dates are met.
"You can be faced with a deadline on a Friday to order generators that arrive two years later… If you don’t keep chasing manufacturers, a hyperscaler can step in and order hundreds more units, pushing you out."
— Ozgur Duzgunoglu, Head of Engineering and Design, Telehouse
Duzgunoglu highlighted the precarious position of mid-sized and enterprise operators who must compete against deep-pocketed hyperscalers capable of buying entire factory outputs in a single stroke. To mitigate this, Telehouse utilizes flexible, standardized designs with interchangeable sub-packages that can be deployed across disparate global markets—from Frankfurt and Paris to Thailand—while maintaining relationships with four or five primary equipment suppliers to avoid single points of failure.
"Paying marginally more to bring infrastructure online faster can deliver real value to the business."
— Justin Ryburn, Field CTO and VP of Global Solutions Engineering, Kentik
Representing an enterprise operator running bare-metal infrastructure in leased cages, Ryburn noted that the scarcity crisis is trickling down. With data centre providers increasingly selling space by the entire floor rather than individual cages, smaller players are forced to innovate—ranging from re-engineering software architectures to run natively within existing hyperscaler environments to utilizing continuous diesel generation on-site to bridge the gap before grid connections are finalized.
Future Outlook: Power, People, and 2029
Looking toward the horizon of 2029, the panellists agreed that the challenges facing the digital infrastructure sector will only evolve in complexity. Two primary bottlenecks will continue to dictate market winners and losers: power availability and human capital.
1. The Looming Power and Grid Crisis
Power remains the single most formidable barrier to expansion. As grid operators grapple with surging electricity demands from data centres, heavy industry, and electric vehicles, the risk of project failure due to delayed grid connections has escalated into major legal territory. Recent high-profile maneuvers—such as Oracle issuing a force majeure notice regarding Project Jupiter, its massive Stargate campus in New Mexico, over potential power delivery delays for its 2028 launch—underscore that grid uncertainty is now an existential contractual risk.
To bypass traditional grid lag, operators are exploring innovative stopgaps, including running backup generators continuously to bootstrap facility operations before transitioning to permanent grid power once infrastructure catches up. Furthermore, strategic geographic relocation—moving developments to regions with robust, unburdened power supplies—is becoming a primary lever for operational success.
2. The Talent Shortage and MEP Deficit
While hardware and power dominate headlines, the human element remains dangerously constrained. Industry leaders warn that a severe shortfall of skilled engineers, project managers, and specialized mechanical, electrical, and plumbing (MEP) tradespeople is creating a fierce cross-industry bidding war. Clients, contractors, designers, and equipment suppliers are all competing for the exact same limited pool of certified professionals.
As Steve Walton, CEO of 3C Datalec, noted in parallel commentary at Capacity Europe, the engineering deficit is every bit as restrictive as a transformer shortage. Furthermore, as developers move into new, remote geographic locations to find available land and power, they frequently encounter local labor markets entirely devoid of specialized data centre construction experience.
Conclusion
The narrative emerging from Capacity Europe 2026 is clear: the AI data centre boom cannot be engineered solely through financial capital or aggressive commercial bargaining. In an industrial landscape defined by scarcity, the winners will be those who master the art of time-driven procurement, embrace flexible performance-based engineering, and secure both manufacturing slots and skilled human talent years ahead of the curve. The era of building fast and cheap is dead; the era of building smart and securing capacity has officially begun.
