How to read the counts. The built layer shows city clusters rather than individual buildings: one circle per market. A circle is sized by published IT capacity in megawatts where a figure exists, and drawn as a small fixed mark where none does. Sources disagree on the national total, and we have not averaged them. Cloudscene counts 529 data centres in Germany; a colocation-portfolio database counts 194 existing and 43 upcoming; trade coverage says roughly 500. The gap comes from scope: every facility including small hosting, against colocation only. Live IT load is shown where a figure is published, which is Frankfurt, Munich and Berlin.

How to read the dots. The dots are not our projects. They mark positions in the market, graded by how settled the information is. Built and planned come from the evidence base and carry its tier. Potential is our own inference from ownership and existing grid connections. It is deliberately the thinnest layer on the map: nobody publishes this information, and it can be gathered only in conversations on the ground, never at a desk.

A site below the transmission level can still be caught by the procedure. Added 9 September 2026. The maturity procedure covers connections to the transmission grid, the extra-high-voltage level of 220 and 380 kV. A load of roughly 5 to 10 MW and more connects at 110 kV, which in Germany is the high-voltage level of the distribution grid; a load of roughly 100 MW connects at 380 kV. A site of 20 to 50 MW therefore sits at distribution level and looks exempt. It usually is not: TransnetBW's own procedure page lists ad-hoc requests from distribution operators to the transmission grid as covered too. Where the local distribution operator has no headroom, it must apply upstream. The site then sits in the queue as one line in someone else's application, without the levers (secured land, permits, equipment orders) that directly improve a ranking in the procedure. Several planned sites on this map, among them Colt DCS at 54 MW and Data Center Partners' Munich site at 17 MW, fall exactly into this size band.

What this layer is not. These are the three or four areas the driver model actually analysed, drawn as ellipses that approximate their extent. They are not a national survey. A blank area under any of the five toggles means the model has not assessed it, and says nothing about the driver there. The ratings are this project's own synthesis (labelled [inference] in the source document) and rest on evidence of only medium confidence. Berlin's growth rate comes from a trade source not checked against a primary. The Berlin allocation figure comes from the operator's own results as reported by ZfK (May 2026), and the Nauen status was checked against independent sources on 21 September 2026.

Why these five areas. Two are Germany's other two federally designated lignite regions (the Rhenish belt, already on Map 1, is the third): Lusatia, where LEAG plays the role RWE plays in the Rhenish belt, and the Mitteldeutsches Revier south of Leipzig, where MIBRAG does. The third, the wider Ruhr, extends the logic of existing power-plant connections to the hard-coal and gas fleet beyond the Rhenish core, in the same crowded Amprion zone as the committed Microsoft cluster. The fourth, the Schleswig-Holstein wind corridor, works by a different mechanism: new transmission capacity is being built there, SuedLink to carry wind power south and NordLink to link with Norway, and no industrial connection is inherited. The fifth, added 8 September 2026, is different again: Rhein-Neckar & Neckarsulm was chosen for its customers, because two named, well-funded buyers (SAP, Schwarz Digits) already sit there, according to our demand research. The "Potential customers" layer below shows that link. The faint grey dots are Map 1's built and planned positions, shown for orientation only; they are not part of this layer.

How to read this map. A favourable marker is active support from an elected mayor or council. For Monheim am Rhein it marks a fiscal-policy benchmark of how far investment-friendliness goes nationally; the town has no known data-centre interest and is on the map for comparison only. An unfavourable marker is a recorded council vote against, or an operator halting work after protest. Contested sits between the two: a citizen initiative and an elected council at odds, with the outcome still open.

Totals: all 14 disclosed figures on this page, €37,238m combined. Private commercial: €5,600m across four projects (the three Map 3 data-centre cases plus the Grünheide expansion). Three of the four, Groß-Gerau, Eisenach and Maintal, are now rejected, paused or stopped, so this is capital proposed rather than deployed. Public: €11,637.7m across eight items: six academic and research programmes (segment 8) plus two subsidy pledges to private projects that have since failed (Magdeburg, Heide). That deserves attention: most of the disclosed "public" figure on this page is subsidy committed to industry rather than spending on public research infrastructure. Mixed: €10,000m, one project (Dresden/ESMC), roughly half private capital and half federal subsidy in the same fab. Regulated: €10,000m, one project (SuedLink), paid for by the grid operators and recovered from every electricity consumer through network charges; it is neither risk capital nor a budget line. None of these four totals is a national census; each is what this project has gathered and sourced so far. For the strategy discussion, the pattern across the five comparable projects matters more than the totals. Two (Dresden, Grünheide) show large, resource-intensive projects moving from announcement to construction or a workable compromise within one to two years, despite real environmental concerns. Two (Magdeburg, Heide) show that political and financial backing, even subsidies in the billions, does not guarantee that a project survives its own execution risk. SuedLink shows that sustained opposition across several states can force a fundamental redesign without stopping a project outright. None of the six data-centre cases on this map has yet reached a resolution by redesign like Grünheide's or SuedLink's; that is worth watching as the cases develop.