Best countries to live in a post-AGI world

Before you act on this ranking, check fiscal fragility. The ranking scores safety after AGI. Fragility on the way to AGI scores who has the debt, deficits, interest bills and unrest that could break them during the transition. The US tops both lists, and 7 of the default top 15 fail a simple fiscal check.

The ranking above asks who is safe after AGI arrives. It doesn't ask who breaks on the way there, and I think that's the bigger near-term risk.

The mechanism.

1. AI hits office work first. The IMF estimates that about 60% of jobs in advanced economies and 40% in emerging markets are exposed to AI (IMF, Jan 2024). The ILO's narrower measure puts generative-AI exposure at 34% of jobs in high-income countries and 11% in low-income ones (ILO WP140, May 2025).
2. Those salaries are the tax base. Across the OECD, personal income tax and social security contributions are 49.2% of all tax revenue (23.7% + 25.5% in 2023; OECD Revenue Statistics 2025). If AI profits show up as corporate income in a few countries, while payroll falls everywhere, revenue drops just as spending on benefits, retraining and pensions rises.
3. A state that starts with high debt, a big deficit and a large interest bill has no room to absorb that. Bond markets then force a choice between austerity, money printing and higher taxes on whatever is still taxable. Each of these feeds populism and unrest, and countries that are already less peaceful have less margin.
4. The opposite case: low debt, net financial assets (sovereign wealth funds), budget surpluses, and resource or compute rents let a government pay for the transition without asking bond markets first.

None of this says when. It says who has the least room if the transition is fast.

Most likely to crack

Most resilient

Where this conflicts with the ranking above, honestly.

• The US is first on the default Combined ranking and first for fiscal fragility. Both are true: it controls the technology (Route A) and it has the least fiscal room of any large economy here. My reading is that the US is the best place to own AI winners from, but the worst-placed large state to pay for the transition. Expect bond stress, higher taxes or inflation, and political turbulence there before anywhere else at the top of the AI race.
• Seven of the Combined top 15 fail a simple fiscal check (table below): the US, St Kitts and Nevis, Barbados, St Vincent, the Maldives, Grenada and Cape Verde. Route B ("nothing worth taking") doesn't protect a country from a debt crisis or the loss of its tourism and service income.
• The UK scores well on Route A (53.1) but is sixth for fragility. France and Belgium also pass liveability easily but sit in the fragile eight.
• The averaged-routes leaders mostly hold up. Singapore, South Korea and Switzerland are in the resilient group and Germany is mid-table (16th). The US and UK are the exceptions.

My picks after both tests. Places that do well on the post-AGI ranking and have fiscal room: Switzerland, Singapore, Denmark, Norway, Ireland and the Netherlands (23rd of 35 for fragility), plus Luxembourg, Malta and Cyprus from the small-state check (debt of 28–51% of GDP and deficits under 3%). I've moved the US from a "safe" pick to "own its assets, don't rely on its state", and I treat the high-debt islands as no longer safe havens. The interactive ranking itself is unchanged, because its weights are yours to adjust. This section is the cross-check.

How the fragility score works. For each of 35 large or commonly considered countries I take seven inputs and convert each to a percentile rank within the set (0 = least fragile, 100 = most fragile, ties averaged). Percentiles rather than raw values stop one outlier, such as Norway's net assets of 152% of GDP, from flattening everyone else. The score is the weighted mean:

Rules and limits.

• Where an input is missing, it's shown as n/a and the score re-weights over the inputs the country has. China lacks net debt and the office-job share; Greece, India, Argentina and the UAE lack net debt; Canada, New Zealand, Saudi Arabia and Japan lack a usable office-job split in ILOSTAT (Japan's release doesn't separate technicians).
• Singapore reports gross debt of 166% of GDP, but by law the government can't spend what it borrows under the Government Securities Act: the proceeds are invested, and assets exceed liabilities (Singapore MOF). So I show the figure but don't score it. With no IMF net debt or interest data either, Singapore is scored on four inputs and is low-confidence.
• Net interest flatters countries with large interest receipts, notably Japan (0.8% of revenue on this measure despite debt of 199% of GDP). In Japan's case the risk is rising bond yields feeding through as debt rolls over, which this snapshot can't see.
• The GPI also penalises militarisation and external wars, not just domestic unrest. That's why Israel and the US score badly on it.
• Not captured: how concentrated corporate tax is (Ireland's depends on a few multinationals), currency or reserve-currency status (which buys the US time), and political will. The weights are my judgement. I tested three alternatives. With equal weights, the top eight and bottom eight are identical. With fiscal-heavy weights (debt and interest 85%), 7 of each 8 stay. With social-heavy weights (ageing, office jobs and peace 65%), 7 of the top 8 and 6 of the bottom 8 stay. The US comes first in every version.
• Thirty-five countries is a sample, not the world. Small states that top the Combined ranking are checked separately above.

Route A (frontier/sovereign AI) is the weighted mean of three indicators, each log-scaled to 0–100 against the leading country:

• frontier models trained (Epoch AI),
• notable models since 2023 (Epoch AI),
• H100-equivalent cluster compute located in the country (Epoch AI GPU clusters).

Route B (little worth taking) is 100 minus the weighted mean of each country's percentile rank on:

• share of world lithium, cobalt and graphite production (USGS via OWID),
• electricity generation in TWh,
• renewable water per person,
• land area.

Liveability is the weighted mean of:

• HDI × 100 (UNDP),
• the rule-of-law index × 100 (V-Dem via OWID),
• a conflict score: 100 minus 20 for each year with an active UCDP conflict in 2021–2025, with a floor of 0.

Thesis score is either the better of A and B (the thesis says either route works) or their average. Combined = Liveability and the thesis score, weighted as you choose.

Limits

• Route B is a proxy for "few AI-relevant resources". It also rewards being small and poor, which is why it is combined with liveability.
• Missing data: countries with gaps are scored on the indicators they have. Hide countries missing more than one indicator (on by default) drops them; untick it to see every country.
• Epoch counts are by the developer's or operator's headquarters, which flatters the US and China.
• The indicators cover the present. They say nothing about how AGI actually unfolds, alliances, nuclear protection, climate or immigration rules.
• The weights are value judgements; the defaults are a starting point, not a finding.