No probability history is available for this market yet.
Top 10% share of U.S. household wealth (2030-2050) under AI progress scenarios
This market is conditional on selecting exactly one AI progress scenario (Slow, Moderate, or Rapid) based on public evidence by 2030-12-31. Moderators should use best judgment based on public benchmark results, deployment evidence, credible reporting, and observed capability demonstrations. If no scenario can be selected with sufficient confidence by the deadline, the market should close unresolved.
Slow Progress
By the end of 2030 in this slower-progress future, AI is a capable assisting technology for humans; it can automate basic research tasks, generate mediocre creative content, assist in vacation planning, and conduct relatively standard tasks currently performed by humans in homes and factories.
Researchers can get literature reviews on almost any topic at the level of a capable PhD student, but AI systems rarely produce novel and feasible solutions to difficult problems. Scientific breakthroughs remain mostly the result of human-run labs and grant cycles, though AI supports copy editing, data cleaning, and analysis.
AI can handle roughly half of freelance software-engineering jobs that would take an experienced human around 8 hours in 2025. AI-supported customer service can resolve most complaints. Writers get modest productivity boosts; short stories can be respectable, but long-form work still requires heavy human rewriting.
At home, an AI system can draft emails, top up grocery orders, collate news, and perform well-scoped tasks that would take a human about an hour or less. It can produce workable weeklong family-vacation itineraries and bookings. Self-driving improves but true level-5 systems do not yet exist. Home robots can make coffee and load/unload a dishwasher in some modern homes, but slower than most humans and with occasional guidance.
In advanced factories, autonomous systems can perform specific repetitive tasks requiring precision but little adaptability (for example wafer handling in semiconductor fabrication facilities).
Moderate Progress
By the end of 2030 in this middle-of-the-road future, AI is an effective collaborator across creative, corporate, and technical domains.
R&D teams can use autonomous lab systems with human assistants for rapid improvements in areas like solar-cell chemistries or fusion-reactor components, though advances are mostly incremental rather than revolutionary. Most freelance software-engineering jobs requiring up to 5 days of experienced-human effort can be handled by AI. Firms can replace most customer-service workflows with AI if customers accept AI service channels.
AI can draft 100,000-word novels good enough for mainstream 2025-style publication with normal editorial work, and can create 5-minute songs with breakout potential and quality comparable to top 2025 label releases. AI agents can manage weeklong travel plans end-to-end, including last-minute rebooking for weather or delays. Executive teams still lead companies, but can delegate multiday operational projects to agentic software.
Level-5 robo-taxis that can drive anywhere a human can finally exist. Home robots can navigate U.S. homes, make coffee, and load/unload dishwashers as fast and reliably as most humans, without guidance. In advanced factories, robots can adapt across tasks without extensive reprogramming and make real-time decisions with greater mobility.
Rapid Progress
By the end of 2030, AI systems are capable of competing with the best human minds and workers, and can surpass them.
Human leadership remains useful for high-level vision, but day-to-day execution can be delegated to AI systems. Autonomous researchers can compress years-long research timelines into days, weeks, or months, yielding game-changing technologies. Human freelancers are outperformed in software engineering and many service roles.
Creative systems can produce top-tier albums and prize-caliber novels, adapt those works into films, and execute commercialization workflows with minimal human oversight. Level-5 robo-taxis are widespread and materially safer than human drivers. General-purpose robots can robustly perform complex household and factory tasks at or above typical human speed and reliability.
Each yearly projection group resolves to the United States value for that calendar year (2030 through 2050) in the OWID series for share of wealth owned by the richest 10% (sourced from the World Inequality Database).
If OWID is unavailable, use the underlying U.S. WID value for the same concept. If later revisions change the archived value, use the latest public value available on the market resolution date.
This market's starting probabilities were generated from aggregate scenario tables in Forecasting the Economic Effects of AI.
This is a conditional multicount projection market for 2030-2050. A moderator selects one of the three AI-progress scenarios (Slow, Moderate, Rapid). Only the selected scenario is used for final resolution.
Calibration method: all three scenarios are fit with an asymmetric Laplace distribution using three quantiles (10th, 50th, 90th). For rapid, these come directly from the paper's reported aggregate percentiles. For slow and moderate, the 50th percentile is the scenario-specific median and the 10th/90th are taken from the unconditional forecast to capture the full uncertainty spread. Yearly quantiles are linearly interpolated from the paper's total-sample 2030 endpoint to each scenario's 2050 endpoint, then converted to threshold ladders using an asymmetric Laplace fit.
Limitation: The paper reports full 10th/50th/90th percentiles only for the rapid scenario and the unconditional (all-scenario) forecast. For slow and moderate scenarios, only the median and a 95% confidence interval around the median are provided — the CI is too narrow to represent the true forecaster uncertainty spread. As a workaround, the unconditional 10th and 90th percentiles are used as the uncertainty envelope for slow and moderate, with the scenario-specific median shifting the center of the distribution. This preserves the full width of forecaster disagreement while reflecting each scenario's directional effect on the median, but it assumes the tail spread is similar across scenarios — which may not hold if, for example, rapid AI progress introduces substantially more downside or upside variance than slower progress.
The paper's inequality prompt is interpreted here as U.S. household wealth concentration. This market resolves against the U.S. top-10% wealth-share series as published via OWID (backed by the World Inequality Database). Yearly ladders are seeded by linearly interpolating scenario quantiles from the paper's total-sample 2030 endpoint to each scenario's 2050 endpoint.
The live resolution source is Our World in Data / World Inequality Database.
Scenario anchors used for seeding
| Scenario | Method | q10 | q50 | q90 | Respondents | Paper table |
|---|---|---|---|---|---|---|
| Slow Progress | Unconditional p10/p90 + scenario median | 67.50 | 73.90 | 80.30 | 402 | Table 38 (Wealth 2050, Total) unconditional p10/p90 + slow median |
| Moderate Progress | Unconditional p10/p90 + scenario median | 67.50 | 75.00 | 80.30 | 402 | Table 38 (Wealth 2050, Total) unconditional p10/p90 + moderate median |
| Rapid Progress | Reported 10th/50th/90th | 69.80 | 79.00 | 85.60 | 402 | Table 38 (Wealth 2050, Total) 10th/50th/90th |
Liquidity over time
i Shows the market’s available liquidity over time. Liquidity can decay continuously and may change discretely if liquidity adjustments are applied.Open this panel to load liquidity history.