Foreign Capital, Host Competition, and the American Data Center Equilibrium

The Authorization Price as an Exchange Rate: US · UAE · Saudi Arabia · Singapore · Japan · Canada · Ireland · France · India · Vietnam · Brazil

What Sovereign Money Bought in America, What Nine Foreign Jurisdictions Are Bidding Against It, and What Twenty-Five Dated Predictions Say Happens Next

In July 2026, a consortium of Abu Dhabi's MGX, BlackRock's Global Infrastructure Partners, and the AI Infrastructure Partnership closed the acquisition of Aligned Data Centers at a $40 billion valuation — 51 campuses and roughly 6.4 gigawatts of operational and planned capacity across the Americas — then committed a further $5 billion to expansion. In the same period, Abu Dhabi continued recruiting hyperscalers to build inside its own borders.

Two markets, one sovereign, opposite directions of travel. American authorization policy prices both, and almost no one setting that policy knows they are quoting into a global auction.

Publication:

https://www.mindcast-ai.com/p/ai-dc-global-investment-flows

Related works: AI Data Center Credit Risk — Permitting, Curtailment, and the Cost of Capital · The MindCast AI Data Center Record (August 2026) · The Authorization Market: Standardized Bargaining, Rationed Power, and the Competition to Build America’s AI Infrastructure (2026) · The Data Center Authorization Price: A 50-State Baseline (2026) · Capital Is the New Computing: Financing the Next Era of AI Infrastructure, 2025–2045 (November 2025) · Power Brokers & Digital Real Estate: How CRE Firms Are Building the AI Infrastructure Backbone (November 2025)

Critical references: The Model AI Infrastructure Authorization Code · The Data Center Authorization Market: A 50-State Regulatory Atlas

The Instrument

County councils, utility commissions, and legislatures fix the terms on which a data center may be built, powered, and operated. Sovereign wealth funds in Abu Dhabi, Riyadh, Singapore, Tokyo, and Toronto read those terms into underwriting models before capital commits. Hyperscalers read the same terms when deciding whether marginal capacity lands in Ohio or in Johor.

MindCast defines the traded quantity precisely. Authorization Spread (A→B) is the expected all-in authorization burden in one jurisdiction minus the burden in another, conditional on power delivery timing, technology access, and tenant credit quality. Burden comprises the four components the 50-State Atlas already grades — statutory conditions, tariff and cost-allocation exposure, commission-order risk, and local ordinance and veto risk. Adding the three conditioning variables converts a domestic grade into a cross-border comparison, and a jurisdiction with a light burden and a five-year energization queue does not offer a favorable spread whatever its grade suggests.

Four factor classes divide by who controls them. Land, water, and generation potential arrive from nature. Capital, deliverable power, and queue position accumulate across years. Advanced chips are allocated by another sovereign through export licensing. Authorization alone is manufactured by law inside a single session, which concentrates the entire competitive margin onto the one factor a legislature governs.

Classical trade theory supplies the ancestry and not the mechanics. Factor price equalization fails here, because capital moves under investment screening, the decisive technology moves under export license, electricity is not tradable across the relevant distances, and latency imposes hard geographic limits. The operative theory is narrower and testable: jurisdictional arbitrage under heterogeneous regulatory prices and constrained substitution.

Thirteen Jurisdictions, Two Entry Gates

Selection ran one test before any jurisdiction entered. Does the jurisdiction send capital into American assets, receive American buildout, or do both?

Capital-flow jurisdictions send material capital toward the United States: United Arab Emirates · Saudi Arabia · Singapore · Japan · Canada.

Mechanism and correction jurisdictions demonstrate how an authorization price moves and what follows when it moves badly: Ireland · France · India · Vietnam · Brazil, with Johor, Malaysia and Chile embedded.

A United States reference case completes the register, because a spread requires both quotes and the American side moves too.

China and Israel are excluded on stated grounds, along with Qatar, South Korea, the Netherlands, and Central Europe. Exclusions do more work than inclusions: six stated reasons demonstrate the selection test discriminating rather than decorating a convenient list.

Five Completed Corrections, Settled Against Public Record

Every foreign case grades against a named authority rather than a forecast.

Singapore paused new capacity in 2019, published allocation criteria, and reopened selectively — shedding commodity load to Johor while retaining the high-value workloads, the customers, and the cables.

Ireland closed the Dublin region to new large energy users from 2021 through grid-operator and regulator action, with no statute and no vote, while elected officials still marketed the country as open for business. The market reopened in December 2025 when the regulator published explicit terms requiring applicants to supply generation covering full demand and source at least 80% of annual demand from new Irish renewables. Four years, and the physical constraint never moved.

Alberta received requests for 19,565 megawatts of data centre load against a 1,200 megawatt interim connection limit, fully allocated to two projects, with 37 proposals deferred.

Quebec restricted utility allocation below requested load. Chile lost a granted permit in court over groundwater — the same water objection Washington counties are raising, already settled abroad against a developer.

Ireland and Alberta converged on the same instrument without coordinating. Both now prioritize projects that bring their own generation, which is authorization architecture converging internationally rather than only across American states.

The Certainty Premium

Authorization cost and authorization certainty are separable variables, and the market pays far more for the second. A high but fixed, published, non-reversible price beats a low price subject to reversal, because a queue position that clears on schedule is worth more to an underwriting committee than a discount a commission order or a court can withdraw after capital commitment.

Four distinct failure modes support the reading. Ireland priced low and ended in refusal. Vietnam offers generous terms on labor and tax against an industry employing in the low hundreds per campus. Chile granted what its courts withdrew. Johor is absorbing displaced demand without pricing the constraint it approaches.

Twenty-Five Predictions, Each With a Settlement Source

MindCast generated the register by running a Cognitive Digital Twin Foresight Simulation across twenty-three behavioral twins — thirteen jurisdictional and ten firm-side, covering hyperscaler classes, developer classes, a sovereign-backed operator, a utility counterparty, and a county authority. Two independent executions ran the same register against the same observation record, and divergence between them is logged rather than averaged away.

Every entry carries a deadline, a falsifier, a named settlement source, and a confidence band with a class label. Event-class entries resolve against an observable occurrence. Interpretive-class entries express confidence in a structural read and never pool with event probabilities.

Eleven Primary Predictions — the Theory Tests

• P1 · 75–85% · event — Interstate substitution dominates international substitution. Of the next three material American authorization-price increases in markets with a viable adjacent-state corridor, at least two produce more identifiable displaced capacity next door than in any foreign jurisdiction. 12 months after each trigger.

• P2 · 80–90% · event — Authorization price moves administratively before it moves legislatively. A majority of qualifying American price movements arrive through commission order, approved tariff, or interconnection queue policy rather than enacted statute. Non-statutory share P10 55% · P50 72% · P90 85%. December 31, 2027.

• P3 · 70–80% · interpretive — Repricing propagates as a cycle, not a one-way transfer. A jurisdiction absorbing displaced large-load demand subsequently raises its own authorization burden rather than remaining durably permissive. December 31, 2028.

• P4 · 70–80% · interpretive — The certainty premium appears in energized capacity before announced capacity. Realized energization correlates more strongly with published, bounded authorization regimes than announced commitment does. December 31, 2028.

• P5 · 85–95% · event — Latency-bound American serving capacity stays domestic. No study jurisdiction captures latency-sensitive American-serving workloads at material scale. Migrating share P10 0% · P50 1% · P90 3%. December 31, 2027.

• P6 · 70–80% · event — The next foreign correction begins below the legislature. The first new constraint-driven correction originates in a utility, grid operator, regulator, allocation authority, or permitting body before any national legislature deliberately resets data center policy. December 31, 2028.

• P7 · 65–75% · event — Large-load tariff architecture converges rather than merely proliferating. A majority of qualifying tariffs carry at least three of five design elements: minimum-take obligation, ten-year-or-longer term, credit support, exit charge, dedicated infrastructure cost allocation. December 31, 2027.

• P8 · 70–80% · event — Cell III jurisdictions pivot to deliverability before incentives produce a step-change. At least two of Brazil, India, and Vietnam materially reform interconnection, energization, or approval certainty before achieving a major improvement in realized deployment. December 31, 2028.

• P9 · 60–70% · event — Gulf capital adapts structurally before retreating volumetrically. Following a qualifying American technology-access or investment-screening tightening, at least one major Gulf allocator alters transaction structure, partner architecture, or access-preserving exposure before broadly reducing American compute investment. Four quarters after trigger.

• P10 · 70–80% · event — Gulf announced capacity converts to energized load materially below schedule, systematically rather than in a single project. Conversion ratio P10 25% · P50 42% · P90 60%. December 31, 2027.

• P11 · 60–75% · event — Foreign capital into American data center infrastructure stays institutionally concentrated. Twenty or fewer foreign institutions account for the material direct institutional set. December 31, 2027.

Fourteen Secondary Predictions — the Jurisdiction Calls

• S1 · 80–90% · event — Ireland's reopened framework holds its generation obligation. No Dublin-region connection offer issues to an applicant not supplying dedicated generation or storage covering full demand.

• S2 · 85–95% · event — Singapore retains selective efficiency and value criteria rather than reopening unrestricted general-purpose capacity.

• S3 · 85–95% · event — Alberta's large-load queue remains oversubscribed, with requested data centre load staying above 10 GW against allocated capacity.

• S4 · 75–85% · event — At least one of ten named Washington counties — Grant, Douglas, Chelan, Kittitas, Benton, Franklin, Yakima, Klickitat, Okanogan, Stevens — adopts or extends a moratorium, interim ordinance, or equivalent restriction.

• S5 · 75–85% · event — A Canadian province does not materially expand data center allocation without accompanying generation, transmission, or explicit capacity reallocation.

• S6 · 75–85% · interpretive — The first visible break in an overextended sovereign hosting commitment appears in delivery timing or project phasing before it appears as explicit cancellation.

• S7 · 70–85% · event — No Indian institution discloses a direct American data center asset position exceeding $250 million.

• S8 · 65–80% · event — Vietnam secures no major hyperscale commitment primarily on labor and tax terms absent an antecedent improvement in grid delivery or data-transfer policy.

• S9 · 60–75% · event — Johor adopts a formal grid, water, capacity-allocation, or equivalent large-load constraint.

• S10 · 70–80% · interpretive — Private-equity-owned developers show greater sensitivity to unresolved queue duration than integrated hyperscalers, producing more resequencing, sale delay, or site substitution as timelines approach fund horizons.

• S11 · 55–70% · interpretive — Tariff adoption clusters along substitution corridors, with states adjacent to a repriced jurisdiction acting sooner than non-adjacent states.

• S12 · 70–80% · event — Displaced capacity from an American or Canadian repricing event lands predominantly in an adjacent jurisdiction sharing the latency envelope. Adjacent-absorption share P10 60% · P50 78% · P90 90%.

• S13 · 55–65% · event — A tightening in American export licensing is followed within four quarters by an increase, not a decrease, in aggregate disclosed value of new Gulf commitments into American compute assets.

• S14 · 70–80% · event — Quebec and Alberta allocation postures continue to diverge, with no federal Canadian harmonization instrument adopted.

Secondary entries carry a December 31, 2027 deadline unless the publication states otherwise. Quantiles above are model-implied scenario values rather than empirically calibrated statistical estimates, and the publication states the generation rule.

What Each Reader Takes

Government affairs and site selection teams get an enumerable map of substitution corridors and the four conditions that must hold simultaneously before capacity can actually move.

State lawmakers and staff get the finding that passivity sets a price, and that the threat from foreign competitors is smaller than either side of the debate assumes.

County councils and local authorities get the timing structure of leverage, and the observation that adjacent counties bid against each other on terms neither would choose alone.

Capital allocators and investment committees get four unpriced exposures — cost-allocation reversal, local veto latency, queue-position value decay, and regulatory exposure correlated by capital origin. Gulf allocators face one American apparatus governing both chip access and investment treatment; Canadian and Japanese allocators do not.

Host-government officials abroad get the diagnostic that each factor position permits only certain instruments, and the rule that runs across all of them: spend on the factor you lack. Nearly every jurisdiction in the study currently spends on the factor it already holds.

Where the Analysis Goes to Work

Money travels toward the United States while compute capacity travels away from it, and the two flows can move in opposite directions at the same moment. Capacity leaves a state without leaving the country, which reconciles real state-level capital flight with small national-level foreign substitution. Announcement megawatts and authorized megawatts are different assets, and only one of them earns.

The competitive question is not whether America should charge an authorization price. Every jurisdiction already charges one, including the jurisdictions that believe they have not set it. The question is whether America knows the spread it is quoting before capital and compute trade against it.

Alongside the publication: the Model Data Center Authorization Code, which grades any data center law on obligation, consequence, and reach; the 50-State Authorization Price Atlas; and the 50-State Baseline, scoring every jurisdiction on one instrument.

For an authorization-spread assessment of a cross-border pipeline, a sovereign counterparty exposure review, a jurisdictional authorization profile, or a multi-country comparison, contact MindCast AI — a predictive law and behavioral economics firm in Bellevue, WA — at [email protected].