Grid Connections: The Real Data Center Bottleneck

For two years the Italian data center debate was a debate about permits. Uncertain timelines, fragmented procedures, no dedicated rulebook. In 2026 that problem was addressed: a ten-month single procedure, a commissioner lane above €1 billion, a Framework Act awaiting the Senate.

Meanwhile the bottleneck moved. It is no longer paper. It is copper.

1. The time series that explains everything

The end-2025 figure comes from Politecnico di Milano's Digitalization and Decarbonization Report: more than double the previous year, nearly thirteen times 2023 levels. The others come from surveys of requests filed with Terna.

One number for scale: peak demand of Italy's entire electricity system sits around 60 GW. Data center connection requests alone are now worth one and a half times the national peak. Nobody, inside Terna or outside, believes they will all be built — and that is precisely the root of the problem.

2. Virtual saturation: the technical name for a traffic jam of ghosts

A connection request occupies capacity on paper long before a construction site exists. If the project is never built, that capacity stays booked anyway — blocking whoever comes next with a real project. This is virtual saturation: a grid saturated with facilities that do not exist.

The phenomenon worsened as request sizes grew. In the October 2025 survey, 42% of applications fell in the 50-100 MW band, power levels requiring 220 and 380 kV connections — that is, directly onto the national transmission grid rather than distribution.

Article 7 of DL 21/2026 — the "DL Bollette" — attacks the knot with three levers:

1. ARERA must revise procedures within 180 days, allocating capacity exclusively to projects that already hold an authorization
2. Terna must disclose the capacity genuinely available, node by node
3. Automatic expiry of pre-existing connections tied to facilities not yet authorized

The consequence is a logical inversion that flips the market: no longer book first, permit later, but permit first, connect after. From "first-to-request" to "first-to-permit". Anyone who built a position by accumulating connection requests on optioned land will find that position is now worth nothing.

A deadline that went almost unnoticed points the same way: ninety peremptory days from confirmation of the connection request to file complete environmental impact documentation, on pain of forfeiture. Ninety days, for an EIA file, is short. Anyone without the study already in the safe will not fit inside it.

3. The geography of the problem

More than 70% of requests are concentrated in the North, with around 20 GW in Lombardy alone and particular density in metropolitan Milan, Brianza, and the Brescia and Cremona areas. The province of Milan already counted 73 active data centers at the end of 2025, out of roughly 209 facilities nationwide.

Here the arithmetic turns unpleasant. Concentrating tens of GW of new load in an already dense area means new substations, grid reinforcements, and construction timelines measured in years that do not depend on the developer. No extraordinary commissioner can compress the build time of a 380 kV line.

4. The asymmetry nobody exploits

Now look at the same problem from the other end of the country.

Southern Italy systematically produces more renewable energy than it consumes. Apulia in particular is a structural exporter: solar and wind generate more than local load can absorb, and the energy has to travel north along the backbones toward consumption centers. When those backbones are congested, generation is curtailed — renewable energy already available that goes unproduced because there is nowhere to put it.

A data center is a load. A large one, constant, flat twenty-four hours a day. Placing it where the energy is already produced and has no outlet does exactly three things:

• absorbs locally the generation that would otherwise be curtailed
• reduces the flows to be moved along the South-North backbones
• does not compete for grid capacity at an already saturated node

This is the industrial case for the digital South, and it is not an argument about regional pride: it is an argument about dispatch. The Article 7 reform strengthens it, because it rewards whoever shows up holding a permit at a node where capacity genuinely exists.

5. What a developer who gets it does

1. Check the node's real capacity before the land, not after: with transparency imposed on Terna, the data becomes consultable
2. Invert the sequence: environmental authorization as the project's first construction site, not its last formality
3. Have the EIA study ready before starting the ninety-day clock
4. Assess Southern nodes as seriously as Milan, looking at real latency to the markets served rather than at a mental map
5. Design for flexibility: modulation capability and storage are negotiating arguments, not merely technical ones

The ten months of the single procedure are real progress. But a permit does not produce electrons. The next race will not be won where the bureaucracy is fastest: it will be won where the power already is.

Note: connection request figures come from public surveys referring to different dates and are by nature fast-moving. The most recent figures cited come from press sources and should be read as orders of magnitude.