World GDP grows 10% in a year

Where things stand

IMF series. I pulled the IMF DataMapper series directly (https://www.imf.org/external/datamapper/api/v1/NGDP_RPCH/WEOWORLD, fetched 2026-10-17, already showing the October 2026 WEO):

• History, 1980–2025: the highest value is 6.7% (2021, the post-COVID rebound). The lowest is −2.7% (2020). World growth has never reached 10% in this series.
• Current projections: 2026 3.1%, 2027 3.4%, 2028 3.3%, 2029 3.2%, 2030 3.1%, 2031 3.1%. The October 13, 2026 WEO (CNBC, 2026-10-12/13) raised 2026 to 3.1%, with AI investment offsetting the Middle East war energy shock. It projects the US at 2.3% in both 2026 and 2027.

Which years can count. Only completed years 2026–2030 qualify. The first estimate for each year appears in the January WEO Update of the next year, with revisions in April, July and October. So the risk falls almost entirely in the first quarters of 2028–2031. Nothing can qualify by 2026-12-31, and the 2026 figure (about 3%) is effectively settled.

How hard 10% is. IMF purchasing-power-parity shares for 2026 (PPPSH) are: China 19.9%, US 14.6%, EU 13.8%, India 8.6%. Emerging and developing economies make up 61%. If the rest of the world grows about 3.5%, world growth of 10% needs the US and China together to grow about 22% or more, or advanced economies plus China to grow about 14% or more. Even a US-only boom like 1942 (about 19%) would leave the world near 6%.

Reference class and paths

• Base rate: zero years at 10% or more, in either the IMF data or longer historical series.
• Path 1, catastrophe then rebound: a temporary collapse in 2027–29 (for example a pandemic far worse than COVID) followed by a rebound of 10% or more. COVID produced only a 6.7% rebound. I put this at about 0.2%.
• Path 2, AI-driven explosive growth (dominant): this needs AI that surpasses humans at most work by about 2028, plus an industrial build-out across the US, China and others that measured world output picks up within one or two years.
  • For faster growth: Amodei's essay "We Must Pace the Frontier" (2026-09-12) says AI improving its own successors has been speeding up progress "since roughly this summer." Anthropic's October 2026 survey finds that the median staff member's answers imply 17.1% US growth in 2030 (US adults: 5.4%). Its own "extreme" scenario reaches 15% US growth by 2030. The AI Futures Project's Q1 2026 update puts Kokotajlo's median for AI that can replace a lab's software engineers at mid-2028 (Eli's at mid-2030), and AI 2027 shows "stratospheric" growth within about a year of superintelligence.
  • Against:
    • In the Forecasting Research Institute's 2026 study, the median economist and superforecaster forecast about 2.4–2.5% unconditional US growth to 2030. Under a "rapid AI progress" scenario, AI experts' 90th percentile was 6.5%.
    • The July 2026 expert panel (LEAP) sees 10%+ growth only as a tail outcome by 2050.
    • The AI Futures Project's own AI 2040 default has full automation of AI research only in 2030.
    • The heads of Anthropic, OpenAI and xAI are now publicly backing a slowdown.
    • Power and grid limits, national-accounts measurement lags, and the heavy weight of emerging economies all work against a 10% world figure.

I roughly put the chance of superintelligence by the end of 2028 at about 9% and by the end of 2029 at about 19%. The chance that world growth of 10% or more is then reported by 2030 is about 45% if it arrives by 2027, 20% if in 2028, and 6% if in 2029. I discounted by about 15% for worlds where catastrophe or conflict stops normal reporting. That gives about 2.2% from AI, about 0.2% from rebounds and a small allowance for other routes.

Hazard series

I computed this in Python, with first-quarter jumps at each January update:

• Quarterly hazard about 0.03% in Q1 2028 (2027 data), 0.28% in Q1 2029, 0.62% in Q1 2030 and 1.1% in Q1 2031, plus small later-quarter revisions.
• Cumulative: 0.05% by end-2028, 0.39% by end-2029, 1.1% by end-2030, 2.4% by end-2031.

Strongest case against

Prediction markets and AI-optimist forecasters are far higher:

• Manifold prices growth above 10% in any single year from 2024 to 2029 at 9% (17 traders). It prices growth above 30% before 2030 at about 7%.
• On Metaculus, the stale question on when the world economy will be 30% above every previous year has its 25th percentile at September 2031.

If superintelligence really is likely by 2028, 5–10% would be defensible. I put more weight on the institutional, economist and superforecaster consensus and on the physical limits, and these thin, long-dated, play-money markets tend to overprice long shots. The other side is that some superforecasters guess near 0.5%. The 2026 figure is fixed at about 3%, so the near quarters are essentially zero.

Sources:

• IMF DataMapper NGDP_RPCH and PPPSH APIs (fetched 2026-10-17)
• CNBC, 2026-10-12: https://www.cnbc.com/2026/10/12/imf-world-bank-annual-meetings-october-bangkok-asia-.html
• Forecasting Research Institute, "Forecasting the Economic Effects of AI" (2026): https://forecastingresearch.org/research/economic-effects-of-ai
• LEAP Wave 10 (2026-07-28): https://leap.forecastingresearch.org/reports/wave10
• Anthropic, "The AI expectations gap" (Oct 2026): https://www.anthropic.com/research/ai-expectations-gap
• Anthropic economic scenarios: https://www.anthropic.com/institute/econ-scenarios
• Amodei, "We Must Pace the Frontier": https://darioamodei.com/post/we-must-pace-the-frontier
• AI Futures Project Q1 2026 update: https://blog.aifutures.org/p/q1-2026-timelines-update
• AI 2040: https://ai-2040.com/
• Manifold API market "will-the-world-economy-grow-by-more-k3vtdwa6r8"
• Metaculus questions 19356 and 5121 (via r.jina.ai)

Question. By the last day of each quarter, has the IMF reported world real GDP growth of 10% or more for a completed calendar year?

Definitions.

• World real GDP growth: the IMF World Economic Outlook (WEO) headline world aggregate annual percentage change in real GDP, using purchasing-power-parity weights: WEO subject code NGDP_RPCH, aggregate WEOWORLD in IMF DataMapper. The corresponding WEO and WEO Update tables label this aggregate “World Output.” Use annual-average growth over the preceding calendar year, not fourth-quarter-over-fourth-quarter growth.
• Completed calendar year: a January–December year that has ended before the qualifying publication. Estimates for an ended year count even if provisional or still labelled estimates or projections; forecasts published before that year ends do not become qualifying merely through the passage of time.
• Apply the Observation period convention: the measured year must end after the opening date, so 2026 is the earliest eligible year.

Resolves YES if an eligible IMF publication reports a value of at least 10.0% for this series and an eligible completed year. Initial estimates and later revisions both count. Use the most precise value published for that observation in the same release, preferring the underlying database over a rounded table or summary; if only a rounded value is published, use it as printed. A later downward revision does not reverse a YES under the Resolution window convention. No attribution to AI is required.

Resolves NO if no qualifying report has been published by the quarter-end. Exclude forecasts for unfinished years, hypothetical scenarios, nominal GDP growth, GDP-per-capita growth, market-exchange-rate-weighted alternatives, regional or national growth, quarterly annualized rates, and independently calculated aggregates. Revisions or repetitions concerning 2025 or earlier do not count.

Resolving source. IMF publications only: the World Economic Outlook database, including its IMF Data portal and DataMapper series; WEO reports and WEO Updates, particularly “Overview of the World Economic Outlook Projections,” World Output row; or another dated IMF-authored publication explicitly reporting this same WEO statistic. A renamed or relocated direct successor series counts if it retains the defined measure. Third-party reporting alone does not qualify. Apply Public confirmation (A2), narrowed to IMF-authored sources, the Resolution window convention, and Resolution date (A3); use the publication date of the qualifying release, not the measured year's end.