Redispatch 2.0 and the German squeeze

Germany spent €2.8 billion on redispatch in 2024. Fifteen times the figure from a decade ago. And nobody called it a grid crisis.

That's the interesting thing. The Netherlands had a National Grid Congestion Action Programme in December 2022. Ireland had CRU/21/124. Both got press cycles, both got named policy moments. Germany has had a permanent soft-constraint regime running since October 2021 and it never got a moment, because the mechanism looks like a back-office reconciliation exercise rather than a policy announcement.

The mechanism, briefly. German wind sits mostly in the north. Demand sits mostly in the south and west. Transmission between the two is chronically overloaded. When too much northern wind shows up, the national market-clearing price says run the turbines, but the physical grid can't move that electricity south. Redispatch 2.0 lets TSOs override the market, curtail northern wind, and ramp southern gas plants to compensate. The curtailed wind gets paid. The ramped-up gas plants get paid. The grid operator passes the cost through to consumers.

In 2024 that mechanism curtailed 9.4 TWh of renewable generation. 3.5% of the country's annual renewable output, generated and then thrown away because the grid couldn't carry it.

The number is coming down. First three quarters of 2025 saw total grid management costs at €2.2 billion, with redispatch volumes down roughly 30% year on year as new transmission upgrades landed. The trend is improving. From a very high base, slowly.

I know this framing is going to annoy the German grid operators, and I'm partly playing devil's advocate on behalf of their PR teams, but Germany has effectively run in permanent soft-constraint without ever saying so. The market stays open. Applications keep getting taken. And whatever physical reconciliation is required afterward, the grid user pays for.

What does this mean for data centres in practice? It depends heavily on where the site is, and most buyers don't think about this enough when they commission.

A site in Schleswig-Holstein or Mecklenburg-Vorpommern — where wind is abundant but southbound transmission is constrained — sits in the zone where curtailment is most frequent. A site near Frankfurt or Munich sits in the zone where redispatch ramp costs land hardest. The grid charge on the bill is averaged nationally. The underlying physical reality is extremely local.

Then there's Section 19(2) of StromNEV, the legacy regime that gave large industrial consumers (including data centres) grid fee discounts for running predictable, baseload-style loads. The regime was designed for a world where constant consumption was easier for the grid to plan around. In a grid dominated by variable renewables, constant consumption is actively harder to accommodate. A study Neon Neue Energieökonomik published for TenneT in late 2024 concluded that Section 19(2) is now "a strong flexibility barrier" and needs reform. That conclusion is mainstream among German energy economists. The political timeline is slower than the economic consensus.

Which leads to the reform that will actually matter. In December 2025 BNetzA published a discussion paper on dynamic grid fees — charges that vary based on real-time local grid capacity. Users in congested zones pay more. Users in uncongested zones pay less. The first phase targets large standalone storage facilities from 2029 at earliest. The longer intention is to phase the regime in across all producers and consumers.

If that happens, and BNetzA has committed to stakeholder discussions from January 2026, it would be the most significant restructuring of the German electricity market since liberalisation in the late 1990s. Lion Hirth at Hertie School called it "the biggest reform since liberalisation" on LinkedIn. That wasn't hyperbole. The implications for any operator running a 20-year infrastructure investment on the assumption of flat grid charges are substantial.

A small digression. It's striking, looking at the German case alongside the Irish and the Dutch, how much the policy response tracks the national character of the regulator. Ireland produced a single clarifying regime document. The Netherlands produced an action programme with named objectives. Germany produced a fifteen-year engineering problem resolved through roughly 900 grid operators implementing a distributed digitalisation protocol. Each country's grid constraint solution reflects the state it was going to produce anyway. Worth watching, when France gets there, whether the solution comes via EDF or Enedis and what that tells you about the resulting operating environment.

Back to the operator question.

For a DC in Germany today, the short version is this. Grid charges are transitioning from fixed line items to real-time operational variables. A site that shapes its load in response to hourly grid conditions will pay meaningfully less than a site running flat. A site that can't respond at all will pay a premium that wasn't in the business case when the build was committed.

This is the exact inversion of the Section 19(2) structure. Constant consumption used to earn discounts. Dynamic consumption will earn discounts. Anyone who committed a flat 24/7 hyperscale build in 2022 on the old incentive structure is going to find the economics have shifted between commit and energisation.

There's a second thing happening underneath worth naming. Germany installed 842 MW of grid-scale battery storage in 2025, almost twice the previous year. BNetzA expects 41 GW of installed BESS by 2037 — roughly double what was projected two years ago. The arbitrage that drives those numbers is the same dysfunction redispatch inefficiency creates: cheap surplus northern wind during the day, expensive southern gas-set prices in the evening, 575 hours of negative prices in 2025. Data centres co-locating BESS capture that arbitrage directly instead of watching it flow to merchant battery operators elsewhere on the grid.

Germany didn't close its grid. It quietly made operating on the grid more expensive, and is in the process of making that expense dynamic and location-specific. The Dutch learned this loudly. The Germans are learning it through the bill.