KI-Nation 2030: A Schildbürger’s Masterplan

Contradictions, Energy Reality, LNG, Nordstream, and the Planwirtschaft of the German AI Taskforce

Industry notes AI & provenance
true
2026-11-03

The Masterplan and the Shoulder-Tapping

On 2 November 2026, at the Digital-Gipfel in Berlin, the German government published the Ergebnisbericht der KI-Taskforce: “Deutschland als KI-Nation 2030”.1
Eighteen cross-ministry key measures, measurable targets (AI in 80 % of large firms, 55 % of SMEs and 90 % of federal authorities; AI data-centre capacity +300 % vs 2025), and the firm assurance that the energy will be covered “nachhaltig und klimaneutral” - somehow.

The document itself records the investment gap:

Nvidia’s Jensen Huang had already offered Chancellor Merz some free advice in Berlin on 19 October 2026:

“There’s nothing to regulate if the industry doesn’t thrive.” - and, interrupting the Chancellor a few minutes later: “Give Germany a chance.”3

What followed was another “Arbeitskreis”, an AISI Deutschland (AI Safety and Security Institute) to be built out, and a coordination document.
Textbook Planwirtschaft - “made in Germany”: diagnose the gap, convene a body, list the measures, and pat each other on the shoulder for having produced a masterplan.
Shall we have a quick reality check? An opinion piece.

Official Energy Numbers from the Ministries and Sector Studies

Current baseline (Bitkom / Borderstep, November 2025)

Source: Bitkom / Borderstep Institut, Rechenzentren in Deutschland: Aktuelle Marktentwicklungen (Update 2025), 10 Nov 2025.5

The Taskforce’s +300 % target for AI data-centre capacity sits on top of this base - and is, incidentally, exactly the fourfold growth Bitkom already expected the market to deliver on its own. Asked by the Bundestag, the government replied that it has “keine Informationen” on the future energy demand of AI-capable data centres; a BMWE-commissioned study puts all data centres at about 42 TWh in 2035.6 Earlier transmission-operator scenario work pointed to 39-88 TWh by 2037/2045.7
Which leaves the question: can Germany even support such a “climate-damaging” technology!?

Simple arithmetic / (R-magic) on the Taskforce ambition

Show code 
baseline_twh_2025 <- 21.3
ai_share_2025 <- 0.15
ai_twh_2025 <- baseline_twh_2025 * ai_share_2025

# +300 % AI capacity (Taskforce language) = 4x
ai_growth_factor <- 4
ai_twh_2030_upper <- ai_twh_2025 * ai_growth_factor

# Conservative total DC if only the AI portion multiplies and the rest stays flat
total_twh_2030_conservative <- (baseline_twh_2025 - ai_twh_2025) + ai_twh_2030_upper

# Bitkom total capacity path (~5 GW by 2030 from 2.98 GW)
capacity_growth_factor <- 5000 / 2980
total_twh_2030_bitkom_scale <- baseline_twh_2025 * capacity_growth_factor

cat("AI-related TWh 2025 (approx.):", round(ai_twh_2025, 1), "\n")
AI-related TWh 2025 (approx.): 3.2
Show code 
cat("AI-related TWh 2030 at +300 %:", round(ai_twh_2030_upper, 1), "\n")
AI-related TWh 2030 at +300 %: 12.8
Show code 
cat("Conservative total DC TWh 2030:", round(total_twh_2030_conservative, 1), "\n")
Conservative total DC TWh 2030: 30.9
Show code 
cat("Increase vs 2025 baseline:", round(total_twh_2030_conservative / baseline_twh_2025, 2), "x\n")
Increase vs 2025 baseline: 1.45 x
Show code 
cat("Bitkom-scale total DC TWh 2030 (capacity ratio):", round(total_twh_2030_bitkom_scale, 1), "\n")
Bitkom-scale total DC TWh 2030 (capacity ratio): 35.7
Show code 
cat("BMWE all-DC estimate 2035: ~42 TWh\n")
BMWE all-DC estimate 2035: ~42 TWh

Note: AI capacity runs at higher utilisation than the average rack, so applying the capacity share to energy (15 % of TWh) understates AI’s real consumption. The numbers above are a floor, not a ceiling.

Even under these modest assumptions the additional load is material.
The higher paths (Bitkom’s 5 GW, the IEA’s global doubling of data-centre demand by 2030,8 the grid operators’ 39-88 TWh) push it further.
The report itself offers no demand figure in TWh and not one word on baseload.

The Energy Contradiction: “Nachhaltig und klimaneutral” vs LNG and Nordstream Reality

Irrational “German Atom-Angst” and our political master brains shut down - and then blew up - the last German nuclear power stations in April 2023 (so that’s out of the window), and sun and wind apparently never send an invoice…
So what’s left?

Nordstream and the LNG pivot

Nord Stream 1 and 2 were sabotaged in September 2022, ending Russian pipeline gas to Germany in practice - with thanks to our best friends and allies.
The federal government has repeatedly stated that a restart of Nord Stream 2 is not under consideration; certification remains suspended; and EU law now phases out remaining Russian gas imports by 2027 regardless of sanctions.9

Replacement: Norwegian pipeline gas plus LNG terminals (Wilhelmshaven, Brunsbüttel, Mukran; Stade still pending).
In 2025 LNG accounted for 10.3 % of German gas imports (106 TWh of 1,031 TWh), and about 96 % of that LNG came from the US.10
On 1 November 2026, the legal reference date, storage stood at just 61.0-61.2 % against a legal average target of about 70 % (80 % for most sites, 45 % for six) - the lowest 1 November level in at least ten years, and 14 points below 2025.11
The ministry assures us supply is secure. Cui bono?

Price reality

The Taskforce language (“verlässliche nachhaltige und klimaneutrale Energieversorgung”, renewable expansion that “berücksichtigt ambitionierte Ausbauszenarien”) treats all of this as a coordination problem.
Yet, it is a physical and a price problem!
US hyperscalers sign long-term nuclear offtake; Berlin still frames the matter as a renewables-plus-efficiency exercise while gas stays the marginal fuel.
And the Energieeffizienzgesetz’s requirement that data centres cover 100 % of their electricity from renewables from 1 January 2027 - bilanziell, via certificates - does not deliver a single firm electron at 3 a.m. in a January Dunkelflaute.

Who else is in the race? A small price overview

Country Price per kWh Basis / year
Germany 20.6 €ct Eurostat, 500-1,999 MWh/yr, H1 2026, excl. VAT
EU-27 average 18.0 €ct Eurostat, 500-1,999 MWh/yr, H1 2026, excl. VAT
France 15.7 €ct Eurostat, 500-1,999 MWh/yr, H1 2026, excl. VAT
Sweden 11.8 €ct Eurostat, 500-1,999 MWh/yr, H1 2026, excl. VAT
Finland 9.3 €ct Eurostat, 500-1,999 MWh/yr, H1 2026, excl. VAT
Norway 11.7 US-ct GlobalPetrolPrices, business 1 GWh/yr, avg 2023-26
USA 8.5 US-ct EIA, all industrial customers, 2025
China 10.9 US-ct GlobalPetrolPrices, business 1 GWh/yr, avg 2023-26
India 12.2 US-ct GlobalPetrolPrices, business 1 GWh/yr, avg 2023-26
UAE 11.0 US-ct GlobalPetrolPrices, business 1 GWh/yr, avg 2023-26
Saudi Arabia 7.0 US-ct GlobalPetrolPrices, business 1 GWh/yr, avg 2023-26
Qatar 3.6 US-ct GlobalPetrolPrices, business 1 GWh/yr, avg 2023-26

Sources: Eurostat nrg_pc_205 (H1 2026); EIA Electric Power Annual (2025); GlobalPetrolPrices.com, business rates, 2023-2026 average (incl. taxes; regulated tariffs in the Gulf, China and India).
Germany on the GlobalPetrolPrices basis: 28.4 US-ct.
Figures are indicative, not like-for-like: hyperscalers negotiate high-voltage, high-load-factor contracts, and the IEA notes that heavy industry in Inner Mongolia or Texas pays well below national averages.16

So, with the deadliest politeness: whom exactly does Germany expect to invest here, and why?
Not the Gulf states, which sell power at a quarter to a half of our price and are building sovereign AI campuses with that cheap gas and sun.
Not China, which pairs tariffs at about half of ours with 295 bn USD of state money.
Not the Nordics or France, a short fibre hop away, with hydro and nuclear at roughly half to two-thirds of the German price.
And not the Americans, who - as Huang kindly reminded us - have their own cheaper electrons and a regulator that waits until there is something to regulate.
Perhaps the investor the masterplan has in mind is the one who values Planungssicherheit above all: certainty that the power will be expensive, the permit slow, and the storage 61 % full.

Musk’s version of the same arithmetic

Elon Musk put the bottleneck plainly back in February 2024:

“A year ago, the shortage was chips; neural net chips. Then, it was very easy to predict that the next shortage will be voltage step-down transformers. […] Then, the next shortage will be electricity. They won’t be able to find enough electricity to run all the chips.”17

And that is why his companies are now taking the data centres off the planet altogether:

“In 36 months, but probably closer to 30 months, the most economically compelling place to put AI will be space. […] The only place you can really scale is space. Once you start thinking in terms of what percentage of the Sun’s power you are harnessing, you realize you have to go to space. You can’t scale very much on Earth.”18

Note the order of escape routes: first cheap firm power on the ground, then orbit. Germany offers neither - though, to be fair, in space there is no grid-connection queue and no Arbeitskreis either.

Obvious Contradictions in One Place

  1. Sovereignty rhetoric vs customer reality. The report and Chancellor Merz speak of “digitale Souveränität” and of avoiding dependence on third parties.
    The physical prerequisites (cheap firm power, capital at scale, fast permits) are not in place.
    The default outcome is continued purchase of frontier training and bulk inference from external providers.

  2. +300 % AI capacity, zero energy figure. A quantitative compute target sits beside a government that told the Bundestag it has no information on future AI energy demand. The BMWE’s 42 TWh (2035) and the Bitkom capacity path never made it into the masterplan.

  3. “Klimaneutral” vs gas as the residual fuel. Post-Nordstream gas is LNG-heavy, US-sourced and price-volatile, and storage started this winter well below target. Certificates are not firm, cheap electrons at the socket.
    Baseload, nuclear offtake and a price gap of ~2.5× to the US: not discussed.

  4. Regulation before industry. Huang said it to Merz’s face: nothing to regulate if the industry does not thrive.
    The answer: a Taskforce, quarterly State Secretary meetings, and an AISI Deutschland layered on top of the EU AI Act.
    Process multiplies; the capability base stays thin.

  5. Capital gap closed by paperwork!? 3.9 bn vs 286 bn (and China’s 295 bn of state money) is not bridged by 18 measures and an “Effizienz-Check” of existing funding instruments - especially as the report expressly makes no funding decisions at all.19

  6. Industrial niche vs frontier ambition. Germany’s real strengths (manufacturing data, engineering, Mittelstand, “physical AI”) are duly acknowledged.
    Yet the goal is still framed as a sovereign KI-Nation with frontier access and European alternatives.
    The energy and capital arithmetic points to a niche site, not to a frontier trailblazer status.

Why Customer and Bystander Status is the Default Path to 2030++

Put the contradictions together and the forecast writes itself:

Germany will take part as a sophisticated customer, integrator and niche supplier until the physical constraints change.
The masterplan documents the gap and answers it with the instruments Germany already knows; the history of Planwirtschaft suggests results will lag the paperwork.
Whatever does emerge will most likely do so despite the programmes, not because of them.

The shoulder-tapping is available immediately. The electrons and the capital are not.

Germany got what it voted for - and we’ll gladly sail the Narrenschiff helmless over the event horizon into the singularity.

Selected Sources

R version 4.6.2 (2026-10-29)
Platform: x86_64-pc-linux-gnu
Running under: Ubuntu 24.04.5 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.12.0 
LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.12.0  LAPACK version 3.12.0

locale:
 [1] LC_CTYPE=de_DE.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=de_DE.UTF-8        LC_COLLATE=de_DE.UTF-8    
 [5] LC_MONETARY=de_DE.UTF-8    LC_MESSAGES=de_DE.UTF-8   
 [7] LC_PAPER=de_DE.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=de_DE.UTF-8 LC_IDENTIFICATION=C       

time zone: Europe/Berlin
tzcode source: system (glibc)

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods  
[7] base     

other attached packages:
[1] knitr_1.51

loaded via a namespace (and not attached):
 [1] digest_0.6.39     R6_2.6.1          fastmap_1.2.0    
 [4] xfun_0.57         cachem_1.1.0      distill_1.6      
 [7] memoise_2.0.1     htmltools_0.5.9   rmarkdown_2.31   
[10] lifecycle_1.0.5   cli_3.6.6         fansi_1.0.7      
[13] downlit_0.4.5     vctrs_0.7.3       sass_0.4.10      
[16] jquerylib_0.1.4   withr_3.0.2       compiler_4.6.2   
[19] rstudioapi_0.18.0 tools_4.6.2       bslib_0.11.0     
[22] evaluate_1.0.5    yaml_2.3.12       otel_0.2.0       
[25] jsonlite_2.0.0    rlang_1.2.0

Olaf Schmalfuß, with the OSDS AI team · OS DataServices

Follow us on X: https://x.com/DataMercsDotNet

  1. BMDS, Deutschland als KI-Nation 2030, 2 Nov 2026 (PDF linked under Sources).↩︎

  2. Ergebnisbericht, section “Ausgangslage”, citing its own footnotes 2-4.↩︎

  3. Quotes as reported by Business Insider, Oct 2026 and the event transcript. The two lines were said at different moments; the draft had merged them into one quote.↩︎

  4. The 4 % share is our own estimate (21.3 TWh against roughly 500-520 TWh gross consumption); Bitkom does not state it.↩︎

  5. Bitkom press release; study PDF.↩︎

  6. Bundestag Drucksache 21/1263, answer to question 2.↩︎

  7. Not re-verified for this revision: from the Netzentwicklungsplan Strom 2037/2045 scenario framework; please check the exact figures before publishing.↩︎

  8. IEA, Energy and AI, April 2025: data-centre electricity demand projected to roughly double to about 945 TWh by 2030.↩︎

  9. REPowerEU phase-out regulation (2026). Exact cut-off dates (LNG vs pipeline, short vs long-term contracts) not re-verified for this revision.↩︎

  10. Bundesnetzagentur, Rückblick Gasversorgung 2025, 9 Jan 2026.↩︎

  11. tagesschau.de, 1/2 Nov 2026 (AGSI data, 61.2 %); gasspeicher.app (61.05 %, 2025: 75.13 %).↩︎

  12. Eurostat, nrg_pc_205, H1 2026, released 30 Oct 2026.↩︎

  13. EIA, Electric Power Annual, Table 2.4, final 2025 data.↩︎

  14. Bundesnetzagentur/SMARD, 2025 electricity market data.↩︎

  15. Bitkom study 2025, p. 4.↩︎

  16. IEA chart, final electricity price for large industrial customers, 2019-2024.↩︎

  17. Bosch Connected World 2024, dial-in Q&A, as transcribed (“lightly edited”) by New Atlas, 1 Mar 2024.↩︎

  18. Cheeky Pint podcast with John Collison and Dwarkesh Patel, transcript; reported by Fortune, 6 Feb 2026. At Davos in January 2026 he gave the reasoning: in orbit “any given solar panel will do five times more energy” - no night, no weather, no atmosphere - and cooling is a radiator pointed into 3-kelvin shade (pv magazine, 26 Jan 2026).↩︎

  19. Ergebnisbericht, closing section; see also move-online.de, 3 Nov 2026.↩︎

Citation

For attribution, please cite this work as

Schmalfuß (2026, Nov. 3). OS DataMercs: KI-Nation 2030: A Schildbürger's Masterplan. Retrieved from https://www.datamercs.net/posts/2026-11-03-ki-nation-2030-a-schildbrgers-masterplan/

BibTeX citation

@misc{schmalfuß2026ki-nation,
  author = {Schmalfuß, Olaf},
  title = {OS DataMercs: KI-Nation 2030: A Schildbürger's Masterplan},
  url = {https://www.datamercs.net/posts/2026-11-03-ki-nation-2030-a-schildbrgers-masterplan/},
  year = {2026}
}