ARIA Convergence Intelligence · Trajectories v1.1

Trajectories

Hypotheses stated in advance, scored in public.

Trajectories is LG Strategy's instrumented approach to envisioning the futures the AI buildout could produce, and measuring the world's progression toward them. We state the futures as testable hypotheses, define the evidence that would confirm or break each one, and follow that evidence as it arrives.

Two futures, stated as testable hypotheses. Eight indicators, scored every quarter, with misses on the record.

One thing we hold constant across both futures: AI transforms business in each of them. The uncertainty is who banks it, and that question gets decided by actors with very different profit engines, responding to the same scarcity in very different ways.

Trajectories v1.1 · Q3 2026 reading, computed August 17, 2026, against the August 11 baseline · The Q4 edition publishes in November
Start with the Actors

Convergence forms through specific marriages.

Six hyperscaler realities

The hyperscalers make money differently, and power scarcity pushes them in different directions. Microsoft sells AI as software margin, so it contracts for power at almost any price rather than own it. Amazon sells compute by the meter, wants the model layer commoditized, and buys nuclear options. Google fights on intelligence-per-watt with its own silicon while AI complicates its search economics. Meta built capacity for itself and is now reported to be renting it out, the moment a cost center tries to become a business. Oracle became the arms dealer, selling raw capacity to frontier labs against a backlog concentrated on a few customers' credit. And the frontier complex itself, the labs and the vendor-financed neoclouds, is pre-profit, existentially short of cheap firm power, and therefore the class of actor most willing to relocate compute to wherever electrons are cheapest.

Energy splits three ways

The Western majors will sell power to compute, gas-fired and increasingly carbon-captured, behind the meter, but shareholder discipline stops them there: molecules to electrons, not electrons to intelligence. The utilities that hold interconnection position are being bought; this year's roughly $200 billion M&A wave is compute capital acquiring the queue. Only the energy-rich states are attempting the full conversion of energy position into compute sovereignty. Convergence, in other words, happens as specific marriages, each pair converging exactly as far as its economics allow.

The swing actor: the integrators

The global consultancies are fashioning themselves as the glue between AI and everything, and their incentives cut both ways. Their model monetizes complexity and change, hours times transformation projects, and genuine transformation, agents actually running processes, eventually cannibalizes those hours. The fork: become the operating layer, running AI-run processes on outcome pricing, or harvest the reengineering boom the way the industry once harvested ERP, where the consultant banked the surplus and the client got a system. Their own numbers will tell: revenue decoupling from headcount and outcome-based contracts growing means productization; headcount growing in lockstep with revenue means the hours model in an AI costume.

The Two Futures

Two trajectories, each grounded in who profits.

Neither future arrives whole. Control shifts away from any single actor. The instrument reads which direction the system is drifting, how fast, and on which evidence. The Q3 reading anticipates that a coalition of scarcity eventually leads to a world of abundance; the open question is when, and what happens on the way.

Trajectory One

The Coalition of Scarcity

The hypothesis

Compute demand keeps outrunning efficiency, and power stays the binding constraint for a decade. Each hyperscaler solves scarcity according to its own economics, contracting for power, owning it, or relocating to it, and durable coalitions form between compute capital and energy position. Energy-advantaged states and companies convert position into compute rents. New categories emerge, but they pay tolls to the stack.

What 2030 looks like if this is right

The data-center map follows the generation map. Energy-rich regions run meaningful shares of global inference. Utilities trade as strategic assets, marriages between compute capital and energy position define the industry's structure, and the phrase “AI infrastructure” means power first, silicon second.

What would prove it

Capital keeps flowing into generation built for compute, compute players keep integrating backward into energy, and the pre-profit frontier keeps relocating toward cheap electrons.

Trajectory Two

The World of Abundance

The hypothesis

Efficiency, open weights, and model commoditization outrun demand growth, and intelligence gets cheap the way bandwidth did. The binding constraint migrates from power to deployment and trust, and the surplus disperses to the domain owners who can put agents into production, and to categories that barely existed three years ago: agent-run firms that deliver services with ten people and a thousand agents; machine-to-machine commerce, where agents transact with agents; the verification economy, because when agents act, someone must prove they acted correctly; context infrastructure, the systems that tell agents what is true inside a company. The internet moved existing business online and created online business that had no offline form. Abundance does the same.

What 2030 looks like if this is right

AI spending shows up inside operating budgets, not as a separate industry. Overbuilt capacity rents cheap, and the actors that leased their power learned it first. The integrators that productized survived their own success; the ones that billed hours got disintermediated by their clients' agents. The largest companies born this decade sit in categories that had no name in 2023.

What would prove it

Inference prices keep falling faster than token demand grows, enterprise spend keeps shifting from platforms to domain applications and the new categories, and the scarce skill keeps proving to be deployment, not capacity.

The Scoreboard

Eight indicators, chosen because they move differently depending on which trajectory is arriving.

Each one tracks an actor's real behavior rather than a sentiment. Every indicator carries a metric, a baseline, and a reading rule set before the quarter begins. Select any indicator for its reading history, the signals that moved it, and the evidence that would settle it. Prior quarters stay on the record in the past-readings archive.

# Indicator What we watch Reading Call
The Q3 2026 read

Investors keep behaving as if power is the moat, while the economic evidence keeps drifting the other way.

The Product

The Trajectories Quarterly

Each quarter we score the eight indicators against the prior baseline: held, moved, or broke, with the evidence cited. The Misses section names where our prior read fell short before anything new is claimed. The full edition is $125 per edition. The tear sheet, including the scoreboard and the misses summary, is free.

How the Scoring Works

Every quarter, an edition: the misses on the record with the hits.

Each quarter we score the eight indicators against the prior baseline: each indicator's movement, which trajectory it favored, and a plain accounting of where the prior read held up and where it did not. We retire and replace indicators that stop discriminating, on the record. Trajectories carry version numbers; when the evidence forces a revision, we publish the revision and date it. This page is Trajectories v1.1.

The Q3 2026 edition is available now. It scores each of the eight indicators against the August 11 baseline and marks where the prior read held and where it broke. The Q4 2026 edition publishes in November 2026. Every prior reading is preserved unedited in the past-readings archive.

Trajectories are produced by the ARIA platform's convergence monitoring, the same pipeline behind the weekly signals and the ARIA Energy Index. Sources for every reading are cited in the linked signals.

Read this week's signals → Request a briefing →