On 18 September 2026 the Federal Ministry for Digital Affairs and State Modernisation (BMDS) published its half-year review of the National Data Centre Strategy. The cabinet adopted the strategy on 18 March 2026. It bundles 28 measures in three fields: energy and sustainability, location and land, and technology and sovereignty. The goal is to at least double data centre capacity by 2030 and at least quadruple AI and high-performance computing capacity. Six months in, 25 of the 28 measures have started, 4 are complete and 5 are in the parliamentary process.
This is a respectable delivery rate for a government document. The decisions that determine whether Germany reaches those multiples have not yet been taken.
The case for the government's reading
The strongest defence of the review is that a strategy is a coordination device, and coordination takes time. Data centres sit across federal energy law, state planning law and municipal land use. A document that gets 25 workstreams moving in half a year, publishes a measure-by-measure status page dated 17 September, and commits to annual updates is more accountable than most national digital strategies. The headline funding is also real. The BMDS cites about €3 billion for European IPCEI projects in cloud, AI and data infrastructure. Federal and state funding for national high-performance computing is also moving.
What has actually been finished
The finished items are mostly low-friction. The BMDS lists the sovereign AI cloud for federal administration, available as part of the Deutschland-Stack, and the continuation of state subsidies that reduce EEG surcharge costs. Neither required Parliament to change the rules private operators work under. Continuing an existing subsidy is worth having, but it does not make power cheaper than it is today.
The measures that would change investment decisions are not done:
- Grid access. The grid connection package was adopted by cabinet in July and is now in parliamentary procedure. Tender procedures for transmission and distribution operators, which would let project maturity count for more, wait on it. The model contract for flexible connection agreements is still a first internal draft going to stakeholders from the fourth quarter of 2026.
- Efficiency and waste heat. The Energy Efficiency Act amendment, meant to make waste-heat rules more practical and ease requirements for existing facilities, was adopted by cabinet in June and is also in parliament.
- Municipal incentives. The trade-tax allocation key, which would let host municipalities share in the revenue, is part of the 2027 income tax reform draft that cabinet adopted on 2 September. It has been launched, not enacted.
- Siting. Preferred areas are still at the stage of a survey of the states. The municipal criteria catalogue is a draft with the states, and the practice check on faster permits with North Rhine-Westphalia was only scheduled for 9 October.
So "started" covers a wide range, from a signed law to a workshop concept. A count of 25 of 28 tells readers little about whether a developer can get a grid connection in 2027.
Why the multiples are hard to judge
The targets are set as multiples rather than absolute figures. Germany's installed IT connection capacity is about 3,000 MW, according to Computerwoche's reading of the strategy. That article also notes that many measures only commit the government to "examine", "identify" or "prepare" something. It adds that the strategy sets no price target for electricity, and that the same article finds no binding obligation or consequence if a measure slips.
Industry reaction at adoption was similar. Bitkom called the strategy an important signal but "kein großer Wurf", and eco said it names the right challenges but stays vague on solutions. The German Datacenter Association argued that local acceptance needs binding incentives for municipalities. The tax allocation key is the first answer to that, and it is the item still in the legislative queue.
Law-firm analysis points the same way. Morgan Lewis describes the strategy as a non-binding policy framework whose results depend on follow-on legislation and administrative action. It lists limited grid capacity, high electricity costs, scarce land in hotspots and lengthy approvals as the binding constraints.
A pro-innovation reading
Our view is that Berlin has picked the right levers. Compute capacity is built where power is available, permits are quick and local communities see a benefit. A strategy that targets grid queues, permitting and municipal revenue is better than one built around subsidies for chosen champions or a mandate for sovereign hosting.
The risk is that the process becomes the achievement. Three tests will say more than the 25-of-28 figure:
- Whether the grid connection package passes with tender rules that reward ready projects over speculative queue positions, without new grid-use conditions that make flexible connection a de facto requirement.
- Whether the efficiency amendment eases requirements for existing facilities as intended, rather than adding reporting duties on top of the current regime.
- Whether the state criteria catalogue and the acceleration zones in the proposed Cloud and AI Development Act turn into approvals measured in months rather than guidance documents.
Proportionate regulation helps here. Efficiency and waste-heat rules serve legitimate aims, and municipalities have a fair concern about land and noise. But these goals are met better by clear, predictable rules than by uncertainty about what a site will cost. The sovereign AI cloud for federal administration is a fair use of government as a customer. It should not be read as evidence that private capacity is being built.
The next honest progress report should publish absolute megawatt figures for contracted, connected and under-construction capacity, so the 2030 multiples can be tracked against something. Until then, the best summary of the six-month review is that the easy items are done and the hard ones are in the Bundestag.