LONDON — In the high-stakes, hyper-accelerated world of artificial intelligence infrastructure, a profound paradigm shift is underway. For decades, the data centre industry operated on a foundational gospel: ruthlessly negotiate the lowest price, squeeze margins through competitive bidding, and let cost drive procurement strategy. Today, that playbook is obsolete.
According to industry leaders speaking at the "Scaling Under Scarcity: Delivering Projects Amid Shortages" panel during Capacity Europe 2026 in London, manufacturing capacity has supplanted cost as the ultimate arbiter of project success. In the AI era, builders are no longer asking how much equipment costs; they are panicked about when—or if—it will arrive.
Data centre developers are now forced to order multi-million-dollar generators up to two years in advance, commit to suppliers before architectural designs are even finalized, and fundamentally rewrite their contracts just to secure a coveted manufacturing slot. As hyperscalers and enterprise tech giants scramble to secure compute power, missing a ready-for-service (RFS) deadline carries financial penalties that dwarf any savings negotiated by a procurement department.
Executive Overview: The Supply Chain Reality Check
The rapid ascension of generative AI, large language models (LLMs), and compute-heavy enterprise workloads has created an unprecedented demand shock for digital infrastructure. While public discourse often fixates on silicon shortages—such as advanced GPUs from NVIDIA and AMD—the physical infrastructure required to house and power these chips is facing an equally catastrophic bottleneck.
High-voltage transformers, gas-insulated switchgear, massive backup generator engines, and specialized cooling distribution units have become the new gold. Lead times for critical power hardware now stretch from 12 to 24 months.
During the Capacity Europe panel on October 14, 2026, engineering and procurement executives from across the digital infrastructure ecosystem—including Greenscale, Start Campus, Telehouse, and Kentik—painted a sobering picture of an industry operating at maximum stress. Moderated by Simon Allen, global ambassador at Nomad Futurist, the session dismantled traditional notions of project management, replacing them with a new corporate mantra: Procure first, design around it.
Detailed Chronology: How Scarcity Redefined the Build Cycle
The evolution of data centre delivery has accelerated dramatically over the past several years, driven by the exponential growth trajectories of AI deployments.
Phase 1: The Traditional Model (Pre-AI Surge)
Historically, data centre construction followed a linear, predictable path. Real estate was acquired, detailed architectural and engineering designs were finalized down to the last bolt, competitive tenders were issued to equipment vendors, and construction commenced. Equipment orders were synchronized with construction milestones, typically requiring lead times of six to nine months. Cost optimization was paramount, and delays were absorbed by comfortable project buffers.
Phase 2: The Grid and Hardware Crunch (2023–2025)
As AI clusters began consuming unprecedented amounts of electricity—often requiring dozens or hundreds of megawatts per facility—the power grid began to buckle. Grid operators across North America and Europe faced massive connection backlogs. Simultaneously, manufacturing lines for heavy electrical equipment became oversaturated. Lead times for high-voltage transformers stretched past a year, turning power procurement into a speculative venture.
Phase 3: The Present Era of Extreme Lead Times (2026 and Beyond)
Today, the timeline mismatch has reached a crisis point. As Nelson Pinto de Magalhaes, director of PMO and procurement at Start Campus, pointed out during the panel, the methodologies and hardware requirements of AI are evolving roughly every six months. Conversely, even an aggressively fast-tracked hyperscale data centre takes approximately 18 months to build.
To bridge this chasm, operators are breaking all traditional rules. Miriam van Kooperen, senior director of supply chain at Greenscale, highlighted the absurdity of the current environment: "We’re worried about cost and lead times, where I believe we should be worried about capacity."
Ozgur Duzgunoglu, head of engineering and design at Telehouse, shared a harrowing anecdote underscoring the urgency: his team recently faced a Friday deadline to place orders for emergency generators that would not arrive for another two years. Procurement teams are forced into a relentless cat-and-mouse game with manufacturers, continuously monitoring production slots to prevent better-funded hyperscalers from jumping the queue and displacing their orders.
Supporting Context & Metrics: The Anatomy of a Mega-Bottleneck
To understand why traditional procurement models have collapsed, one must examine the specific hardware and logistical constraints crippling the sector:
- Generators: Essential for ensuring uninterrupted uptime, heavy-duty backup diesel and gas generators now carry lead times of up to 24 months.
- High-Voltage Transformers: The vital bridge between local grids and energy-hungry AI servers require 12 to 18 months from order to delivery, dependent heavily on specialized raw materials like grain-oriented electrical steel.
- Grid Connections: In dense European hubs like Frankfurt, securing reliable grid power can take an astonishing five to seven years, forcing operators to explore creative—and expensive—interim workarounds.
- The Innovation Velocity Mismatch: AI software and chip iterations update every 6 months, while physical civil and electrical infrastructure takes 18 months minimum to materialize.
Furthermore, this hardware crunch is compounded by severe labour shortages. Steve Walton, CEO of 3C Datalec, noted that the engineering and mechanical, electrical, and plumbing (MEP) talent pool is stretched paper-thin. Contractors, designers, clients, and suppliers are all raiding the same finite talent pool, driving up labour costs and threatening project execution.
Official Statements and Industry Insights
The panelists at Capacity Europe 2026 offered profound insights into how their organizations are adapting to this unforgiving landscape.
Rethinking Procurement and Design Integration
Start Campus, developer of a massive hyperscale campus in Portugal, has instituted an extreme structural pivot to survive the shortages. The company has integrated its procurement and design teams under a unified command structure where procurement takes the lead.
"‘Procure first, design around it’ is now one of our mantras," explained Nelson Pinto de Magalhaes. "You can’t really compress a factory to produce more."
Pinto de Magalhaes argued that customers evaluate operators based entirely on whether the Ready-For-Service (RFS) date is met, not on which individual component was delayed. By utilizing phased RFS models, facilities can go live incrementally while heavy construction continues in adjacent phases. In this framework, time is monetized directly: "What if everything that we called savings now is days?"
Ozgun Duzgunoglu of Telehouse echoed this sentiment, emphasizing that every day shaved off a build schedule accelerates revenue generation and improves capital return metrics. Under these conditions, a slight price premium for expedited delivery is viewed as a minor operational cost compared to the revenue lost by missing a market window.
Shifting from Product-Specific to Performance Specifications
Greenscale’s Miriam van Kooperen revealed how her philosophy on standardization has evolved under pressure. While supply chain managers once begged design teams for rigid product standardization to streamline maintenance, Van Kooperen now advocates for performance-based specifications.
By defining what a piece of infrastructure needs to achieve rather than naming a single proprietary brand, operators can dynamically pivot to alternative suppliers if their primary vendor hits manufacturing capacity limits.
However, this flexibility is not without its trade-offs. Pinto de Magalhaes noted that while performance specs broaden the available supplier pool, they can shift technical risk toward the integration interfaces and commissioning phases—stages that are already compressed by tight deadlines.
Future Outlook: Power, People, and 2029
Looking three years down the road toward 2029, the consensus among industry experts is that relief is not imminent. The structural headwinds facing the digital infrastructure market will continue to evolve, shifting from simple hardware acquisition to complex geographical and human resource challenges.
The Power and Grid Crisis
Power availability will remain the absolute ceiling for AI expansion. Justin Ryburn, field CTO and VP of global solutions engineering at Kentik, pointed out that smaller players are already feeling a different kind of squeeze. With data centre providers increasingly selling capacity by the floor rather than the cage, boutique operators are weighing whether to re-architect their software stacks to run natively on hyperscaler clouds that already possess secured power.
The risks of late power are no longer theoretical. Oracle recently issued a force majeure notice regarding Project Jupiter—its massive Stargate campus in New Mexico—preemptively hedging against the possibility that local energy infrastructure will fail to meet its targeted 2028 operational launch.
In the United States and parts of Europe, some operators are resorting to running backup generators continuously to bring data halls online before grid interconnects are finalized, transitioning to municipal power only once the grid catches up.
The Geographic Arbitrage of Power
To bypass grid paralysis in saturated Tier 1 markets like London, Frankfurt, and Paris, developers are increasingly forced into geographical arbitrage. Moving projects to secondary and tertiary markets where greenfield power is accessible removes the grid bottleneck, but introduces a new problem: a total lack of local data centre construction expertise and MEP trades.
The Rise of the "Chief Time Officer"
When moderator Simon Allen half-jokingly suggested that modern data centre operators might soon require a "Chief Time Officer," the panel noted that procurement leaders are already stepping into that exact role. As manufacturing lines fill up years in advance and billions of dollars in AI investments hang in the balance, mastering time—not cost—has become the ultimate competitive advantage in the digital infrastructure economy.
