We connect public EuroHPC facts with official infrastructure aggregates from Germany’s Federal Network Agency. Explore compute systems, AI factories and regional network assets—source-bound, without estimated capacity or a synthetic score.
Also covered: 1 public compute system, 2 AI factories, 16 states and 400 districts.
Compare absolute source values only within the same administrative level. Larger and more populous regions will often have larger totals; this is not a quality or capacity ranking.
| Region | Source value | Rank | Share of federal value |
|---|---|---|---|
| 174,889.08 km | 1 of 16 | 19.6% | |
| 146,524.24 km | 2 of 16 | 16.4% | |
| 125,352.6 km | 3 of 16 | 14.1% | |
| 90,310.61 km | 4 of 16 | 10.1% | |
| 61,248.55 km | 5 of 16 | 6.9% | |
| 60,351.34 km | 6 of 16 | 6.8% | |
| 45,095.9 km | 7 of 16 | 5.1% | |
| 38,631.24 km | 8 of 16 | 4.3% | |
| 36,481.07 km | 9 of 16 | 4.1% | |
| 31,246.61 km | 10 of 16 | 3.5% | |
| 26,327.87 km | 11 of 16 | 3% | |
| 20,802.32 km | 12 of 16 | 2.3% | |
| 14,323.55 km | 13 of 16 | 1.6% | |
| 8,234.4 km | 14 of 16 | 0.92% | |
| 7,374 km | 15 of 16 | 0.83% | |
| 4,120.9 km | 16 of 16 | 0.46% |
Federal, state and district rows are separate aggregates. Never add them across levels—that would count the same infrastructure more than once. A missing value means “not reported by the source”, not zero infrastructure.
Since 13 July 2026, we have archived new source states in an auditable form. Once a second materially changed source period exists, we can show genuine changes by source period—unchanged weekly checks will not become a synthetic trend.
The private collection preserves auditable source states. The public site receives only a strictly bounded field-level projection with exact units and provenance.
A rule-based collector checks the approved EuroHPC and Gigabit sources weekly. Changed source states are archived immutably.
Only published facts on a fixed allowlist enter the board. Raw files, run metadata, contacts and internal storage paths remain private.
Federal, state and district aggregates stay separate. Missing values remain missing; kilometres and item counts are never blended.
The data proves neither bandwidth nor free compute capacity, private GPU holdings, data-centre sites or grid headroom.
Absolute totals are not a fair efficiency comparison without area, population or network length. We therefore deliberately publish no infrastructure or AI-readiness score, and no map without separately audited boundaries.
The board is produced during the website build from the private archive. Raw and normalised archive objects have no public access path.
We translate workload, privacy, latency and cost into a robust cloud, EU or on-prem architecture—without pretending catalogue signals are availability commitments.
These tools complement the current result.