The number of data centre projects announced around Milan has grown faster than the number that can actually be connected. This is not a planning failure in the usual sense. Sites exist, capital exists, and permits are obtainable. What is scarce is a connection to the high voltage network at the capacity these buildings need.
Grid connection is a queue, and the queue is physical. New capacity at a substation requires transformers, switchgear, and often a new line, and those have delivery times measured in years because the global order book for that equipment is full. A project that has financing and planning consent can still be waiting on a connection date that sits outside its investment horizon.
The queue also has a fairness problem. Connection requests are typically processed in order of application, which rewards speculative early applications and penalises projects that waited until they were real. Operators with several sites can hold positions in multiple queues, which inflates the apparent demand and pushes genuine projects further back.
There are two credible fixes, and they point in different directions. One is to make queue positions conditional on demonstrated progress, so speculative requests fall away. The other is to move the buildings to where the capacity already is, which in practice means the industrial areas of the Po valley with existing heavy connections and cooling water, rather than the ring immediately around the city.
The second option is better engineering and worse commercially, because latency to Milan matters for a subset of customers and land near the city is where the operators want to be. What decides it will not be planning policy. It will be whichever substation gets its transformers first.