Forecasting the Future 2026

Bridgewater Associates and Global Citizen’s 2026 Forecasting the Future challenge asks entrants to make at least ten binary forecasts about artificial intelligence, modern mercantilism or the interaction between them. The forecast list is only one part of the task. Entrants must also explain the cause-and-effect framework connecting those forecasts and provide an analytical appendix showing the evidence behind them.

I built the project around a specific claim: the next phase of the AI race will be constrained by the ability to deliver power and strategic inputs, not just by progress in models or chip design. A data centre becomes productive capacity only when chips, electrical equipment, cooling, land, grid connections and finance arrive together. The same distinction applies at national scale. Owning resources is useful; converting them into reliable output at the required location and within the commercial window is more important.

The competition

The submission is organised into three parts. Part 1 contains twelve one-sentence binary forecasts, each with an assigned probability, a measurable threshold, a resolution date and an authoritative source. Part 2 develops the framework connecting those forecasts. Part 3 tests the framework against public data on electricity demand, interconnection queues, capacity pricing, European industrial power costs, critical-mineral processing, export controls and strategic manufacturing investment.

The forecasts cover four connected areas:

• AI and electricity: US and global data-centre demand, grid-connection delays and the price of reliable capacity.
• Strategic inputs: critical-mineral processing, Chinese export controls and US restrictions on advanced accelerators.
• Industrial competition: European electricity costs, manufacturing construction and nuclear restarts.
• State bargaining: trade-policy intervention and bilateral agreements linking compute access to investment or security commitments.

Research question

The project asks which countries can turn energy, strategic inputs, capital and policy into productive capacity quickly enough to shape the next decade of AI and industrial competition.

That required separating several ideas that are often treated as interchangeable. Resource ownership is not refining capacity. Proposed generation is not connected power. A government subsidy is not an operating factory. A futures price is not a literal forecast. An export licence can matter even when trade continues. The research therefore follows the physical and legal chain from resource to deployment rather than treating AI investment as a single spending number.

What I built

I created a forecast ledger containing the statement, probability, threshold, date, primary resolver and revision rule for each forecast. I then linked every forecast to a claim-level evidence register so the supporting data could be checked independently of the prose.

The analysis uses frozen public extracts from DOE and Lawrence Berkeley National Laboratory, the International Energy Agency, PJM, Eurostat, the WTO–IMF Trade Policy Activity Index, the US Bureau of Industry and Security, the Nuclear Regulatory Commission and the US Census Bureau. The accompanying charts distinguish observations, estimates, projections and scenario ranges rather than drawing them as one continuous historical series.

I also built a dependency map across the forecast set. The data-centre demand forecasts share a common adoption driver; the interconnection and capacity-price forecasts share a grid-scarcity mechanism; the mineral concentration and export-control forecasts share a processing chokepoint; and the chip-control and compute-diplomacy forecasts both depend on the strategic value of advanced accelerators. Treating those links explicitly avoids presenting twelve correlated claims as twelve independent bets.

Framework

The framework centres on deployable energy sovereignty: the ability to convert strategic resources, processing capacity, equipment, generation, storage, grids, finance and legal access into dependable productive output.

The causal chain runs from resources and processing through generation, grids and logistics into delivered price, reliability and time-to-power. Those conditions shape compute capacity, industrial output, profits, inflation and trade. Governments then respond through subsidies, tariffs, export controls, procurement, stockpiling and bilateral agreements, feeding policy back into the supply chain.

The United States, China and Europe occupy different positions in that system. The United States combines energy resources, capital markets and leading AI firms with slower infrastructure deployment. China has greater import exposure but stronger manufacturing and build-out capacity across grids, batteries, solar equipment and mineral refining. Europe retains capable firms and a large market but faces higher delivered energy costs and more fragmented fiscal support.

Evidence and exhibits

The research note includes portfolio-native charts for US and global data-centre electricity demand, US interconnection-queue capacity and EU non-household electricity prices. It also includes causal diagrams showing the deployable-energy-sovereignty framework and the distinction between the broad pandemic market shock and the contract-specific mechanics that drove the May 2020 WTI contract below zero.

Those exhibits are not decorative. Each one supports a forecast or tests a link in the causal framework. Scenario ranges remain ranges, queue capacity is not presented as operational supply, and the WTI diagram avoids implying that lower energy use caused the entire equity-market crash.

Result

The finished research note presents twelve forecasts in the format required by the challenge, followed by the framework and analytical appendix that support them. The project also produced reusable forecasting infrastructure: structured resolution criteria, source manifests, public-data transformations, chart metadata, dependency checks and a written update framework for changing probabilities when new evidence arrives.