Four scenarios for 2030

Scenarios organize uncertainty without pretending to know a single outcome. The four below differ along two axes: the strength of realized AI productivity and the degree of platform concentration.

The future will likely contain elements of more than one scenario. Different sectors and regions can occupy different states. The purpose is to identify coherent combinations of capital markets, energy, enterprise adoption, labor, and geopolitics - and to specify the signs that one combination is becoming dominant.

Scenario typology, not a 2x2

The four scenarios are a coherent typology selected to expose distinct mechanisms, not the exhaustive cells of a two-axis matrix. Productivity realization, control-point concentration, geopolitical fragmentation, finance and regulation interact in more than two dimensions. The protocol therefore classifies each annual observation by the scenario whose distinctive signposts fit best and reports residual evidence rather than forcing every world into a nominal 2x2.

Scenario classification rule

For each annual review, compare realized indicators with the expected sign in Table 11.2, identify the scenario with the best overall fit, and publish the largest residuals. Do not declare all scenarios compatible with the evidence. A transition between scenarios is permissible only when the dated indicator record shows the change.

Table 11.1. Four internally coherent 2030 scenarios.

Scenario A: managed multi-model diffusion

Model capability, efficiency, and infrastructure improve, but the dominant enterprise control point sits with hyperscalers, systems of record, managed control-plane providers, and capable customers rather than one frontier lab. Models compete behind common endpoints, and capable open weights become a substantial production layer for specialized, cost-sensitive, or private workloads. Enterprises use private inference for sensitive tasks and managed frontier systems for bursty or difficult tasks.

Productivity diffuses because organizations can buy managed plurality rather than build it themselves. The market can be open at the model and application layers while remaining concentrated in chips, cloud, identity, and control planes. The scenario succeeds when secure task cost falls, open and proprietary systems remain substitutable, enterprises retain control rights, and gains spread through competition and organizational redesign rather than being absorbed entirely by the substrate.

Leading signs: Falling cost per successful task; high production conversion; rising model substitution behind managed endpoints; stable direct customer control over data and evaluation; successful exit tests; broad productivity; and cloud concentration that does not eliminate model or application competition.

Financial outcome: Large but differentiated winners; infrastructure resembles a competitive utility and technology stack rather than a single bubble.

Policy priority: Interoperability, skills, grid acceleration, competition, and targeted safeguards for high-risk uses.

Scenario B: capacity overbuild and capex reset

Enterprises continue purchasing pilots and capacity, but process redesign, data quality, reliability, and liability slow conversion to economic outcomes. Model prices fall faster than serving costs. Strategic financing and long-term commitments obscure weak independent demand. Data centers, generation, and semiconductor capacity arrive as revenue growth decelerates. Public listings broaden exposure near the top of the cycle.

The correction begins as a valuation event and migrates into project finance, suppliers, utilities, and regional development plans. Some firms fail or consolidate; equity holders absorb large losses; government intervention focuses on critical services and infrastructure rather than restoring old valuations. The technology survives, and distressed capacity later lowers prices, but the capital cycle imposes a recessionary drag in exposed regions and portfolios.

Leading signs: Falling utilization, repeated project delays, rising cancellations, weak cash conversion, increasing vendor financing, enterprise renewals below pilot growth, and negative free cash flow despite scale.

Financial outcome: Technology-bust dynamics with pockets of credit and utility stress; systemic severity depends on leverage and guarantees.

Policy priority: Exposure transparency, ratepayer protection, orderly resolution, continuity plans, and avoiding indiscriminate bailouts.

Scenario C: managed-digital-labor oligopoly

A small number of frontier platforms combine capital, compute, superior models, consumer distribution, enterprise applications, agents, identity, memory, payment, and transaction systems. They use high valuations to acquire application-layer control points and then route demand toward owned models and infrastructure. Customers increasingly purchase outcomes rather than software seats or labor hours. Smaller providers become complements, resellers, or acquisition targets. Enterprise systems of record remain, but the agent runtime and application interface become the primary economic control points.

Productivity is high, but rents and dependency are concentrated. A platform outage or policy change can interrupt many sectors. Workers experience rapid task displacement, and the platforms capture a large share of the surplus. Governments treat continuity as strategically essential, creating a form of soft systemic protection. Competition policy struggles because efficiency and concentration arise from the same integrated system.

Leading signs: Rising outcome-based revenue; repeated stock-financed acquisitions of application distribution; increasing owned-model share inside acquired products; declining direct use of enterprise software interfaces; reduced model neutrality; a rising share of critical workflows on one platform; and high margins despite base-model price competition.

Financial outcome: Sustained mega-cap concentration and strong cash flows, accompanied by operational and political systemic importance.

Policy priority: Portability, non-discrimination, critical-service continuity, data rights, liability, and limits on self-preferencing.

Scenario D: sovereign fragmentation

Export controls, cyber incidents, data-localization rules, industrial policy, and political distrust fragment the global stack. The United States, China, Europe, India, Gulf states, and other regions build or sponsor distinct compute and model ecosystems. Enterprises maintain separate deployments to serve different jurisdictions. Open-weight models spread, but licenses, standards, hardware, and safety requirements diverge.

Redundancy improves some forms of resilience while raising cost and reducing scale economies. Countries with cheap power, capital, and trusted political relationships attract infrastructure. Smaller countries depend on regional alliances or shared platforms. Global technology firms face duplicated compliance and capital expenditure. Productivity gains continue but diffuse unevenly, and national-security logic increasingly shapes commercial architecture.

Leading signs: Regional model mandates, incompatible standards, localized training and inference, restricted chip and model flows, duplicated corporate stacks, and politically conditioned cloud access.

Financial outcome: Lower global margins but protected regional champions; infrastructure is financed partly for strategic redundancy rather than utilization.

Policy priority: Allied interoperability, minimum cross-border standards, supply-chain resilience, and disciplined strategic subsidies.

Scenario signposts

Table 11.2. Pre-specified indicator signs for adjudicating the four scenarios.

What would make the worst outcomes systemic

The capex-reset scenario becomes a financial crisis only if leverage, guarantees, collateral, funding structures, and intermediary exposure transmit losses into core financial functions. The workflow-oligopoly scenario becomes an operational crisis only if tested substitutes and resolution cannot meet required recovery times. Sovereign fragmentation becomes a macroeconomic crisis if duplication and restrictions materially reduce productivity and investment. Scenario severity is adjudicated by channel-specific outcomes, not by an aggregate systemic score.