Data center power demand has risen sharply over the past three years, driven by the likes of AI infrastructure and the higher rack densities it requires. Facilities that were once designed around 6 to 12 kilowatts per rack are now being planned around loads several times higher , with AI clusters introducing large, concentrated demand blocks that must be provisioned upfront by grids that were designed to accommodate predictable industrial and commercial loads.
The International Energy Agency has warned that procurement timelines for core grid components now routinely extend beyond two years for cables and up to four years for large power transformers. Even when regulatory approval moves quickly, actual, physical delivery might not follow until years later. This disconnect between construction timelines and grid expansion timelines has left transmission operators unable to respond at the pace hyperscalers like Amazon need to bring their new facilities online in a timely fashion.
Meanwhile, Europe’s data center electricity consumption continues to grow exponentially. Estimates by the European Commission place EU data center demand at roughly 96 TWh in 2024, with projections rising to 168 TWh by 2030 . That growth is heavily front-loaded in AI-capable facilities, which concentrate power demand in ways traditional grid planning models don’t handle well.
The most severe bottlenecks can be found in nations like Italy, where grid operators have received tens of gigawatts’ worth of connection requests tied to speculative or early-stage data center projects. Similar patterns have been observed in Spain and parts of Northern Europe, with capacity being reserved several years in advance and without any sort of policing or enforcement against companies that sit on secured capacity without then proceeding to define or achieve construction milestones.
This capacity squatting creates a situation where grid capacity exists on paper but is, for all intents and purposes, unavailable. Projects that are ready to build are forced to wait behind projects that may never materialize, turning access to the grid into a race of who can get to it first, versus whose projects are actually ready to get moving.
In the United Kingdom, efforts are underway to reform this system, with energy regulator Ofgem moving away from strict first-come-first-served allocation toward a “first-ready” model that prioritizes projects with land, financing, and permits in place. It’s understood that similar reforms are being discussed in Italy and Spain, but bureaucracy is slow, and existing queues will take years to unwind.
All of this is influencing where hyperscalers are willing to invest. Reuters , citing analysis from energy think tank Ember, has warned that poor grid planning could push Europe’s data centers away from established hubs like Frankfurt, London, Amsterdam, and Paris, where waiting times are the longest.
Oracle reportedly delays several new OpenAI data centers because of shortages
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/amazons-european-data-center-projects-stalled-by-grid-delays#main
- https://www.tomshardware.com
- China broadens its crackdown on cryptocurrencies — expands ban to include real-world asset tokenization, crypto ads, and providing network traffic for crypto ac
- [Daily Due Diligence] NVDA NVDA
- CEOs of NVIDIA and Lilly Share ‘Blueprint for What Is Possible’ in AI and Drug Discovery
- Crypto disaster sends $44 billion in bitcoins to random users — exchange accidentally makes funds vanish thanks to a typo
- Retro Apple Mac mod implements thermal printer floppy swap — machine also benefits from a Mac Mini brain transplant
Informational only. No financial advice. Do your own research.