By now, the grid connection story is familiar. Interconnection queues have stretched to years in some of Europe’s core hubs. In the UK, the queue nearly tripled in seven months before an emergency reform. The idea that grid capacity, not chip supply, is the real constraint on data center growth has moved from contrarian to conventional wisdom in under two years.
That story is true. It’s also incomplete. A grid connection offer is a right to build, not a delivered substation, and not the same as getting power. Between that offer and energization sits a full electrical build-out: transformers, switchgear, protection systems, cabling, all specified, manufactured, delivered and commissioned. Even the UK’s recent queue reform only clears one link in that chain. Clear every queue in Europe tomorrow, and most projects still wouldn’t move at the pace their timelines assume.
The equipment wall
Start with the hardware. Medium-voltage switchgear now averages around 44 weeks, and contractor-side reporting for data center-class specifications runs considerably higher: standard switchboards near 52 weeks, power circuit breaker switchboards past 84 weeks, and medium-voltage gear approaching two to three years in some configurations. Large power transformers average well over two years, generator step-up units longer still, and lead times on high-voltage units from tier-one manufacturers have been reported as high as five years. Wood Mackenzie has assessed an estimated 30% supply deficit in power transformers against current demand. Europe is not exempt: ABB and Siemens Energy, two of the largest transformer manufacturers in the world, have both reported high-voltage lead times of 48 to 60 months, with new orders effectively targeting 2030-2031 delivery.
The consequence shows up in build rates as clearly as in quotes. Bloomberg reported that of the near 12 GW of US data center capacity slated to come online in 2026, only about a third was under active construction, with a shortage of transformers, switchgear and batteries cited as a primary reason. This is not a data center-specific problem, either. Utilities, industrial electrification projects and renewable interconnections are drawing on the same finite pool of transformer and switchgear manufacturing capacity, which is part of why the backlog keeps growing rather than clearing. France’s mid-2026 capacity commitments alone, including SoftBank, Ardian and Nebius, add up to close to 4 to 6 GW, implying 30 to 40 large high-voltage transformers drawn from that same strained European manufacturing base.
The material wall: even cable isn’t fully immune
Cable lead times are shorter than transformers, typically weeks rather than years, so they rarely make headlines on their own. But material and trade dynamics are adding friction in specific geographies. In the US, a 50% tariff on imported semi-finished copper products took effect in August 2025, covering wire and other forms used directly in power distribution. The International Copper Study Group has projected a global refined copper deficit of some 150,000 metric tons for 2026.
To be precise about what this does and doesn’t mean: copper scarcity by itself is rarely why a specific project misses its date. The bigger constraint remains transformer and switchgear lead times. The tariff dynamic described above is specific to the US market, but the underlying refined copper deficit is global, and European buyers draw from the same constrained supply chain. Tariffs, price volatility and regional supply concentration add cost and planning complexity on top of an already stretched equipment timeline, particularly for projects sourcing cable and components from a single region or a single supplier. In a system already short on slack, that’s one more variable that can turn a tight schedule into a missed one.
The people wall
The least visible constraint may be the most binding one. A 2023 Uptime Institute survey found that 58% of global data center operators already faced difficulty sourcing talent for open roles, before the current wave of construction accelerated. More recent industry surveys put the number closer to 90% citing staffing as a significant obstacle to building or expanding. In the US specifically, electrical engineering enrollment has fallen roughly 90% relative to computer science since the 1980s, and industry surveys describe a retirement ratio of about three senior engineers leaving the field for every one or two new graduates entering it. The UK shows the same strain from a different angle: Skills England’s 2026 sector assessment found that 68% of priority engineering occupations are already in critical or elevated demand, with electrical and electronics engineers named among the most pressured groups.
The shortage compounds on itself in a specific way: the same power systems engineers needed to design a data center’s electrical architecture are also the engineers’ utilities need to build out grid capacity in the first place. Data centers and the grids meant to supply them are competing for the same, shrinking pool of specialized talent. An engineering firm stretched across too many simultaneous projects works slower and has less room to redesign around a delayed transformer, a substitute switchgear unit, or a change in cable routing, exactly the kind of adaptation this environment now demands routinely.
Why this compounds, rather than adds up
None of these three constraints operate in isolation. A transformer delay is manageable if an engineering team has the bench strength to redesign around a substitute spec on short notice. It becomes a lost year if that team is already stretched across a dozen other projects. A single-region material sourcing plan is manageable if there’s slack elsewhere in the schedule to absorb a tariff-driven delay. It becomes critical-path risk when it lands on top of a switchgear order that was already running eighteen months late. Treat the queue, the equipment, the materials and the engineering capacity as four separate line items, and each one looks survivable. Treat them as one system, which is what they actually are, and it becomes clear why so many announced projects are sliding.
Turning electrification into a competitive edge
The operators pulling ahead aren’t the ones with the earliest queue position. They’re the ones who treated equipment, materials and engineering capacity as design-stage decisions rather than downstream procurement and locked in each of them early enough to matter.
The growing complexity of AI facilities means that equipment suppliers must contribute more than products. Customers increasingly need application engineering, installation support, and a detailed understanding of how materials behave under direct current, higher temperatures, and sustained electrical loads.
Chief Business Development Officer Data center & Grid & Connect large projects, Nexans
In practice, that means treating manufacturing capacity and engineering bandwidth as scarce resources to secure early, the same way a developer already treats land and permits. A partner with committed production capacity across multiple countries absorbs a supply shock differently than one sourcing from a single plant, and an engineering team with deep bench strength can redesign around a substitute unit in weeks instead of months, neither of which is a line item you can buy at the last minute once a project is already behind. Nexans works this way across 39 countries: engineering depth from substation to rack, a manufacturing footprint that absorbs demand spikes without pushing every customer to the back of the same queue, and application engineering teams that supplement a stretched client-side design team rather than shipping a spec sheet and moving on, treating electrification as a lifecycle discipline rather than a single product sale.
What this looks like on the ground
None of this is theoretical. Three recent Nexans-supported projects, in the UK, the Netherlands and the US, show what solving these constraints actually looks like in practice. Client details stay confidential, but the pattern doesn’t.
On an 80 MW UK data center project, the contractor’s schedule depended on cable stock being available faster than standard lead times allowed. A dedicated cable-stock partnership closed that gap, fire-safety-rated cable meeting Euroclass Cca requirements under CPR (Construction Products Regulation) met the compliance bar, and supply chain support kept the programme on schedule. Nexans delivered more than 260 km of low-voltage cable into the facility, concentrated in the cooling infrastructure that keeps AI-density racks running.
On a 14 MW project in the Netherlands, the deciding factors were different: visibility into the project pipeline built through relationships across the wider OEM ecosystem, competitive pricing, and product availability at the moment it was needed. The low-voltage package included fire-rated, reduced-bending-radius cable specified for exactly the tight routing constraints that AI-density retrofits create.
In the US, the pattern looked different again: early wins built through targeted distributor alignment, an existing product portfolio used to solve an immediate supply chain gap, and manufacturing capacity agile enough to respond to high-volume, fast-decision demand. In a supply-constrained market, what won the business wasn’t the newest technology. It was technical support, responsiveness, and the ability to actually deliver at volume.
None of that shows up in an interconnection queue chart. It shows up in whether a project actually energizes on time.
What specifiers should ask now
For engineering firms and hyperscaler technical teams shaping projects today, the queue is no longer the only question worth asking, and the most useful questions are about what to build differently, not just where the risk sits:
- How do we secure switchgear and transformer capacity early enough, through committed production slots rather than open-market quotes, that equipment lead time stops sitting on the critical path?
- How can we structure cable and material sourcing across more than one region, so that a single tariff shift or supply disruption doesn’t become a project-wide delay?
- What kind of engineering partnership gives our team the bench strength to redesign around a delayed or substitute unit in weeks, instead of losing months waiting on the original spec?
None of these questions has a single right answer. But asking them at design stage, alongside the grid connection timeline rather than after it, is what increasingly separates projects that hit their energization date from those that don’t.
The wall after the wall is the opportunity
Two years of headlines have trained the industry to watch the grid queue. The next constraint won’t announce itself as clearly, spread as it is across equipment backlogs, material sourcing and engineering capacity rather than one visible number. Some operators are already ahead of it, treating those three constraints as design-stage decisions instead of downstream surprises, and the UK, Dutch and US projects above show what that looks like in practice. The rest will find out the hard way.
The queue made headlines. The operators solving what’s behind it are the ones already plugging in.
