Systems in depth
One section per system: its state at evidence close, its exposures and what the operational series will watch.
Contents
- ·Edition home
- ·Executive summary
- 1Technology rests on a few physical inputs
- 2Key findings
- 3Systems and sectors at baseline
- 4Risks and capability
- 5Local grid constraint that overlaps a chip or mineral supply constraint
- 6Readings from the other desks
- 7The 2027 agenda
- 8Systems in depth
- AMethodology
- BData appendix
- CReferences and record
- ·Complete edition, for print
8.1Common causes
One cause can strain many layers at once: fast demand growth, a concentrated supply chain, a policy change or a hazard. The shared cause matters more than the layer.
| Instance | Mechanism in action | Evidence | Status |
|---|---|---|---|
| Demand growth | data-centre use grew about 12% a year to 2024 | 1 | Reported |
| Concentrated supply | one country holds the top refining share for most minerals tracked | 3 | Reported |
| Policy change | new chip, AI, data and cyber rules and funding | 43816 | Reported |
| Export measures | a 2025 US measure on AI chips and its rescission were reported; effect not bound | 28 | Inferred |
Pathways in this mechanism
- AIC → ENE load growth
- MIN → TRD concentrated supply
- FIN → TRD policy funding
Evidence to bind in 2027
- common cause tagging of shortages
- frequency of concurrent causes
Related editions
- GLB Global Systems
- SR-02 Critical Infrastructure and Lifeline Systems
- SR-03 Global Supply Chains, Trade and Chokepoints
- SR-04 Critical Minerals and Strategic Materials
- XT-06 Advanced Semiconductors and AI Compute
- XT-08 Cloud, Edge and Distributed Computing
8.2Domino effects
A limit in one layer removes an input from the next. A grid limit slows new compute, and a refining limit slows chips.
| Instance | Mechanism in action | Evidence | Status |
|---|---|---|---|
| Power to compute | clustered load tightens local grids | 112 | Reported |
| Minerals to chips | refining concentration limits inputs to chips and equipment | 3 | Inferred |
| Chips to services | chip supply limits server and network capacity | 4 | Inferred |
| Cloud to users | shared cloud services pass a fault to many users | 9 | Inferred |
Pathways in this mechanism
- AIC → ENE → INF
- MIN → TRD → AIC
- DAT → INF
Evidence to bind in 2027
- sequence timing of capacity limits
- service-level dependence data
Related editions
- GLB Global Systems
- SR-02 Critical Infrastructure and Lifeline Systems
- SR-03 Global Supply Chains, Trade and Chokepoints
- SR-04 Critical Minerals and Strategic Materials
- XT-06 Advanced Semiconductors and AI Compute
- XT-08 Cloud, Edge and Distributed Computing
8.3Inter-systemic feedbacks
Technology infrastructure depends on what it powers and also presses back on it. Compute raises power demand, and power limits compute.
| Instance | Mechanism in action | Evidence | Status |
|---|---|---|---|
| Compute to grid | clustered demand tightens local grids | 12 | Reported |
| Grid to compute | connection limits slow new capacity | 1 | Inferred |
| Capital to capacity | private investment funds new sites and chips | 11 | Inferred |
| Efficiency to load | hardware and software efficiency changes the load path | 1 | Reported |
Pathways in this mechanism
- AIC → ENE
- ENE → AIC → INF
- FIN → AIC
Evidence to bind in 2027
- connection-time data
- cluster-level reliability
Related editions
- GLB Global Systems
- SR-02 Critical Infrastructure and Lifeline Systems
- SR-03 Global Supply Chains, Trade and Chokepoints
- SR-04 Critical Minerals and Strategic Materials
- XT-06 Advanced Semiconductors and AI Compute
- XT-08 Cloud, Edge and Distributed Computing
8.4Shared vulnerabilities
Different layers share the same weak points: concentrated refining, concentrated chip supply, shared cloud services and clustered sites.
| Instance | Mechanism in action | Evidence | Status |
|---|---|---|---|
| Refining | a single country leads refining for most minerals tracked | 3 | Reported |
| Cloud services | wide use makes shared providers a shared dependency; shares not bound | 9 | Inferred |
| Clustered sites | nearly half of US capacity sits in five regional clusters | 1 | Reported |
| Orbital links | satellite and timing signals underpin networks; no series opened | 25 | Expert input |
Pathways in this mechanism
- MIN → TRD concentrated refining
- DAT → INF shared cloud
- AIC → ENE clustered load
Evidence to bind in 2027
- provider concentration by region
- dependence on timing and positioning signals
Related editions
- GLB Global Systems
- SR-02 Critical Infrastructure and Lifeline Systems
- SR-03 Global Supply Chains, Trade and Chokepoints
- SR-04 Critical Minerals and Strategic Materials
- XT-06 Advanced Semiconductors and AI Compute
- XT-08 Cloud, Edge and Distributed Computing
8.5Thresholds and tipping
A system can absorb stress until a threshold is crossed. Grid headroom, refining diversity and delivered chip capacity are such thresholds for technology infrastructure.
| Instance | Mechanism in action | Evidence | Status |
|---|---|---|---|
| Grid headroom | load that doubles by 2030 meets local limits first | 112 | Inferred |
| Chip capacity | a 20% target against an 11.7% forecast | 7 | Reported |
| Refining diversity | rare earth refining concentration fell modestly while other minerals rose | 3 | Reported |
| Tipping in networks | cascades can become abrupt in interdependent networks | 40 | Inferred |
Pathways in this mechanism
- AIC → ENE threshold
- FIN → TRD capacity threshold
- MIN → TRD diversity threshold
Evidence to bind in 2027
- grid headroom by region
- cascade thresholds at operating scale
Related editions
- GLB Global Systems
- SR-02 Critical Infrastructure and Lifeline Systems
- SR-03 Global Supply Chains, Trade and Chokepoints
- SR-04 Critical Minerals and Strategic Materials
- XT-06 Advanced Semiconductors and AI Compute
- XT-08 Cloud, Edge and Distributed Computing
Notes
Numbers are the edition’s reference numbers, so a note reads the same in every chapter. The full list, with every source and its licence, is in Annex C. Annex C, references
- 1IEA (ed.), Energy and AI (World Energy Outlook Special Report). International Energy Agency, Paris, April 2025. Observation year 2024. Accessed 2026-11-03.back
- 3IEA (ed.), Global Critical Minerals Outlook 2026 (revised version, July 2026). International Energy Agency, Paris. Observation year 2025. Accessed 2026-11-03.back
- 4WSTS (ed.), Global Semiconductor Market grows 26% in 2025 to USD 796 billion (release page; sales of USD 795.6 billion, up 26.2%). World Semiconductor Trade Statistics, Dec 2025 data, release dated March 2026 by file name. Observation year 2025. Accessed 2026-11-03.back
- 43US Congress (ed.), CHIPS and Science Act of 2022, Public Law 117-167, Division A, Title I, section 102 (statute text, US Government Publishing Office; CHIPS for America Fund appropriations for fiscal years 2022 to 2026 of USD 39 billion for manufacturing incentives under section 9902 and USD 11 billion for research; totals are sums of the annual amounts). Observation year 2022. Accessed 2026-11-03.back
- 8European Commission (ed.), European Chips Act (policy page; Regulation (EU) 2023/1781; three pillars; 19 State aid decisions with over EUR 33 billion of public and private investment). European Commission, Brussels. Observation year 2026. Accessed 2026-11-03.back
- 16European Commission (ed.), AI Act (policy page; Regulation (EU) 2024/1689; in force 1 August 2024; general application 2 August 2026; omnibus in force 27 July 2026). European Commission, Brussels. Observation year 2026. Accessed 2026-11-03.back
- 28UNVERIFIED, not opened: US Department of Commerce, Bureau of Industry and Security, rescission of the 2025 AI diffusion export measure, May 2025, cited from memory; the Federal Register page redirected to an access gate. Observation year 2025. Accessed 2026-11-03.back
- 2IEA (ed.), Global Energy Review 2026 (revised version, May 2026). International Energy Agency, Paris. Observation year 2025. Accessed 2026-11-03.back
- 12Central Statistics Office (ed.), Data Centres Metered Electricity Consumption 2024 (release of 10 June 2025; 22% of metered electricity in 2024, 21% in 2023, 5% in 2015). CSO, Cork. Observation year 2024. Accessed 2026-11-03.back
- 9Eurostat (ed.), Cloud computing: statistics on the use by enterprises (Statistics Explained; 52.74% of EU enterprises with 10 or more persons used paid cloud services in 2025). Eurostat, Luxembourg. Observation year 2025. Accessed 2026-11-03.back
- 11Stanford Institute for Human-Centered AI (ed.), The 2026 AI Index Report (report page; private AI investment in 2025 of USD 285.88 billion in the United States, USD 20.92 billion in Europe and USD 12.41 billion in China). Stanford University, Stanford. Observation year 2025. Accessed 2026-11-03.back
- 25UNVERIFIED, not opened: ESA (ed.), ESA Space Environment Report 2025, cited for orbital infrastructure context; the page returned an access error and no figure is used. European Space Agency, Paris. Observation year 2025. Accessed 2026-11-03.back
- 7European Court of Auditors (ed.), The EU’s strategy for microchips, special report 12/2025 (news item read; the 20% target and the Commission forecast of 9.8% in 2022 and 11.7% in 2030; the full report was not opened). European Court of Auditors, Luxembourg. Observation year 2025. Accessed 2026-11-03.back
- 40UNVERIFIED, not opened: Buldyrev, S.V. et al., Catastrophic cascade of failures in interdependent networks, Nature 464, 1025 to 1028, 2010, cited from memory. Springer Nature. Observation year 2010. Accessed 2026-11-03.back