Scenarios 2027 / Scenarios
Rated at Standard Severe. Likelihood type: systemic. Source of scores: ginc-desk-v0.2.
The graduate hiring round that normally takes 40,000 entrants takes 9,000. The accountancy, law and software firms that were the country's largest employers of 22-year-olds now buy the same work from an API billed in a foreign currency. Tax receipts from employment fall faster than the finance ministry's models allowed, because the models were trained on recessions, and this is not one: output is up. The public-service digital team discovers it cannot run its own systems without a provider that changed its terms of service in a quarter. Universities lose a third of applicants to the degrees that paid for the rest. The welfare system is asked to carry a population it was not designed for: qualified, young, unemployed and online. Unions, professional bodies and three political parties converge on a demand for a tax on compute; the compute is abroad. The nations that come through are those with a domestic model capability, a way to tax value where it is consumed, an education system that can retrain at scale and a state that uses the technology better than its citizens' employers do.
The dateline is illustrative, not a forecast. The narrative is hypothetical; the historical anchors below are real events.
Frontier AI displaces a large share of knowledge work within three years while models, compute and standards sit with a few foreign providers. Tax bases, graduate labour markets, education systems and the state's own administrative capability are loaded at once. A market-concentration variant covers a sudden repricing of AI-linked assets and the dependency of national digital infrastructure on providers whose decisions are made elsewhere. The Allianz Risk Barometer 2026 ranks AI second (up from tenth); the WEF ranks adverse AI outcomes fifth on the ten-year horizon.
| Event | Date | What happened | Calibrates |
|---|---|---|---|
| DeepSeek R1 release | 27 January 2025 | Nvidia lost about US$590 billion of market value in a day on a cheaper frontier-class model | market-concentration variant |
| Generative AI adoption | 2023–2026 | Fastest consumer technology adoption on record; measurable effects on entry-level hiring in exposed occupations by 2025 | diffusion speed |
| Allianz Risk Barometer 2026 | January 2026 | AI the fastest-rising business risk, second globally | business perception |
| WEF Global Risks Report 2026 | January 2026 | Adverse outcomes of AI fifth on the ten-year horizon, the largest climb between horizons | long-horizon severity |
| Eurasia Group Top Risks 2026 | January 2026 | 'AI eats its users': extractive business models as a stability risk | political channel |
| Previous general-purpose technologies | 1990s–2000s | ICT diffusion took two decades to show in productivity; displacement concentrated by region | counter-anchor on speed |
Shown at their preset values. Parameters are not adjustable in this release and nothing on this page is computed from them. Custom settings run (Phase B) but are labelled 'non-standard run' and excluded from comparisons.
| Parameter | Default | Range or options | Note |
|---|---|---|---|
| Share of occupations with over half of tasks exposed | 30 per cent | 15–50 | — |
| Diffusion to displacement | 3 years | 2–7 | — |
| Domestic frontier-model capability | none | options: some / frontier | — |
| Compute import dependence | high | low–high | — |
| Market-concentration shock | off | — | On at Extreme |
| Provider terms-of-service shock | on | — | — |
High: capability band shifts expected under current plans. Medium: band shifts under 'none' policy response only. Low: strain without band shift. Loads are judgement-based until the Atlas connects. Domains link to the Atlas.
| Domain | Load | Channel |
|---|---|---|
| Hard | ||
| Defence and security | Medium | dependence on foreign models for defence and intelligence functions |
| Strategic infrastructure | Medium | data centres, power demand, connectivity |
| Critical technology | High | domestic model, compute and data capability |
| Soft | ||
| Government effectiveness | High | tax design, regulation, state use of AI, procurement dependence |
| Human capital | High | graduate labour market, retraining, education finance |
| Influence and cohesion | High | distributional conflict, information environment, standards influence |
| Economic | ||
| Macro-financial | High | tax base, asset concentration, productivity |
| Industry, trade and supply | Medium | services trade, firm concentration |
| Energy and resources | Medium | compute power demand |
| Region | Exposure | Rationale |
|---|---|---|
| North America | High | Origin; concentration and displacement |
| Europe | High | Dependence without domestic frontier providers |
| China | Medium | Domestic capability; its own displacement |
| Indo-Pacific | High | Services exporters; Japan and Korea's ageing labour markets as mitigant |
| South Asia | High | IT services and outsourcing exposure |
| Gulf and Middle East | Medium | Importers with capital |
| Africa | Medium | Lower exposure; leapfrog potential; dependence |
| Latin America and Caribbean | Medium | — |
| Russia and Eurasia | Medium | — |
| Indicator | Source | Threshold |
|---|---|---|
| Entry-level vacancy and graduate employment series by occupation | — | — |
| Share of tasks automated in exposed sectors | ILO, OECD | — |
| AI capex and model price per token | — | — |
| Public procurement of foreign AI services | — | — |
| Education application trends | — | — |
| Equity concentration in AI-linked stocks | — | — |
| Sovereign compute capacity | — | — |
From the GINC 250: trends rated Very high or Critical for this scenario. All S13 trend scores.
| Rating | Band or level | Why |
|---|---|---|
| Likelihood, two years | 3Possible | Displacement at the 30 per cent exposure level within two years is unlikely; the market-concentration variant is likelier and sits at Extreme. |
| Likelihood, ten years | 5Highly likely | — |
| Systemic impact | 4Severe | On capability grounds rather than GDP loss, which the scenario may raise. |
| National impact | 4Severe | — |
| Confidence | low | The technology's trajectory is contested. |
Source of scores: ginc-desk-v0.2. Confidence refers to the rating, not the scenario. Calibration sources are listed with the anchors above and on the methodology page.
Contested assumptions for the panel to resolve.
No signed commentary in this build.
0.2.0 · active0.2.0 · 2 October 2026 · Entered the Library at v0.2 with GINC desk scores.GINC (2027). Scenario S13 Rapid AI diffusion shock, Scenario Library v0.2. scenarios.ginc.org/library/rapid-ai-diffusion-shockContent and data are published under CC BY 4.0.