America’s compute supremacy, and the ground beneath it.

The United States runs the largest stock of data-centre capacity on earth and the AI labs driving global demand. The 500 billion dollar Stargate programme opened 2025. The question is no longer money or ambition but power, land and local consent – and a cold economic war with China over the chips the whole stack runs on.

US · Global share 45% Estimated share of world capacity
Stargate · 2025 500USD bn Announced AI build
Backlash · 2025 156USD bn Projects blocked or stalled
Analyst overview

The compute superpower, and the ground it has to land on.

The United States is not a next hotspot; it is the incumbent. It hosts the largest stock of data-centre capacity on earth and the AI labs driving global demand. What is new is not whether America builds, but whether it can land the build it has already financed – against a power grid, a planning system and a politics that were never designed for gigawatt campuses.

The flagship is Stargate: a programme of up to 500 billion dollars, targeting 10 gigawatts by 2029, announced in January 2025 by OpenAI, SoftBank and Oracle, already rising on real land at Abilene, Texas. Capital is the easy part. Private credit from Blue Owl, sovereign money via MGX, and hyperscaler balance sheets underwrite gigawatt campuses before they break ground.

What bends an American data-centre plan is not the money. It is power – nuclear restarts at Three Mile Island for Microsoft, Talen’s reactor deal for Amazon, Entergy building roughly 7.5 GW of new gas across ten plants for Meta’s Hyperion – and consent: by 2025, analysts counted on the order of 156 billion dollars of projects blocked or stalled by local opposition across some forty states. The build has met the neighbours.

Over all of it sits the cold economic war with China: export controls on the most advanced chips, and a contest to set the world’s AI rails. Four instruments read the United States: The Chess Game, the move-by-move US–China contest for compute; The Unrest Tracker on the build meeting its neighbours; The story of Stargate on the ground; and The Brief on the flagship campuses and the counties that decide them. Every modelled figure is tagged, and every projection is framed as such.

Entelligencia desk · North America · chapter 09 of 10
The Chess Game

The board is digital infrastructure. The players are two.

ENT · USA · CHESS / 2026.06
Moves logged · 16
Scenarios modelled · 4
Status · live, moving

The trade war that opened in 2018 with tariffs on steel and soybeans has hardened into something colder and more structural: a contest over who owns and operates the world’s digital infrastructure, the chips, clouds, cables, data centres and the rules that govern them. Read it as a game between two players. The United States moves with controls, capital and alliances; China moves with self-reliance, integrated infrastructure and upstream materials. Step through the board move by move, then model the endgame.

ENT · USA · THE BOARD
The board
Two players, sixteen moves.

From the 2018 tariffs to the 2026 compute race, step through the US–China contest for the world’s digital infrastructure, one move at a time.

The boardEighteen years of moves, one position.
United States China
The centre of the board is control of compute. Whoever commands it commands the AI era.
Even position0.0
China aheadUS ahead
US2018Trade
Opening gambit
Pawn · e2 to e4
Press play to walk the board, or pick a move from the ledger.
00 / 16
Chess positions are an editorial device, not a literal game tree. Move dates and actions are sourced from the timeline below and public reporting. The position evaluation is an Entelligencia directional read of who holds the initiative on digital infrastructure, not a market index. Re-verify at build.
The directory · five acts of the cold war Enter any act for the full story →
Severity ticks are an Entelligencia editorial read of how contested each act is, not a precise index. Act imagery is representative and stylised. Re-verify at build.
ENT · USA · ENDGAME
The endgame model
Model who wins the AI race.

An interactive model in four layers: what a chosen path does to the global compute race, where the capital and capacity concentrate, how stable the board and its supply lines are, and the move-by-move record behind it.

The endgame modelModel who wins the AI race.

A directional Entelligencia model, not a forecast. Pick a path and an escalation level, then read four layers: what it does to the global compute race, who captures the investment, how stable the board and its supply lines are, and the move record. Figures are ranges and directions, not precise outcomes.

Escalation and fragmentationModerate
Global compute efficiency
Supply-chain fragmentation cost
Chip and equipment price pressure
Deployment-timeline slippage
Cross-bloc interoperability
Race read
Set a path to model the outcome.
The model weighs the US lead in leading-edge chips, capital and alliances against China’s lead in integrated build-out, grid and upstream materials.
US race index
China race index
Entelligencia scenario model (June 2026). Inputs: Chatham House on the AI race; CFR and PIIE on controls; WEF and Reuters timelines. Directional, re-verify at build.
of globally mobile compute investment is in play under this path, on Entelligencia’s directional model. The bars show where that capital and capacity are most likely to concentrate.
United Statesreshoring · chips · hyperscale
Allies and Europesovereign clouds · trusted rails
Indiascale · cost · neutral
Neutral hubsGulf · SE Asia · arbitrage
China and its spherestate demand · integrated stacks
Who captures the build
A contested board does not freeze investment; it redirects it. The model routes mobile commitments toward whichever bloc offers chip access, scale, safety or neutral positioning under the chosen path.
Entelligencia directional model (June 2026). A judgement of relative attractiveness, not tracked flows. Re-verify at build.
Supply-chain safety · unhedged → hedged
Taiwan-routed silicon
Wider compute supply
Board stability score
0of 100
Stability tracks · how the board is kept from tipping
Guardrails, not a truce
In 2026 Washington and Beijing began exploring limited transparency and crisis-communication channels around powerful AI and dual-use systems, mirroring Cold War arms-control logic: compete, but avoid catastrophic miscalculation.
Public position · Wall Street Journal, 2026
The silicon shield
Leading-edge logic still concentrates in Taiwan. US policy aims to keep China several generations behind while friend-shoring fabs and packaging to reduce single-point dependence.
Entelligencia analysis; TecEx timeline
Materials and the counter-move
China’s gallium, germanium and rare-earth controls are the mirror image of US chip controls. Stockpiles and alternative processing reduce the leverage of any single chokepoint.
WEF policy timeline, 2025
Power as the real constraint
Chips without megawatts are inventory. The side that integrates grid build-out with compute fastest converts its paper lead into deployed capability.
Entelligencia analysis; Levin (2026)
Toggle levers to see how hedges lift supply-chain safety under the current path. The model rewards layered hedges with diminishing returns. Directional, re-verify at build.
2018US
Section 301 tariffs open the war
The US imposes tariffs on a broad set of Chinese imports, citing IP theft and forced tech transfer. Technology and telecom are flagged as the real strategic front from the outset.
Source · PIIE; Wikipedia trade-war timeline, 2018
2019US
Huawei on the Entity List
Washington restricts Huawei’s access to US chips and software and begins lobbying allies to exclude it from 5G cores, turning network gear into a geopolitical choice.
Source · CFR; Telecoms.com, 2019
2019China
The “new infrastructure” strategy
Beijing formalises 5G, data centres, AI, industrial internet and ultra-high-voltage power as state priorities, the template for state-backed build-out at home and abroad.
Source · Warsaw Institute, 2019
2020US
The foreign-foundry rule
Global foundries using US technology, including TSMC, now need a licence to supply Huawei, hitting China’s leading telecom maker hard.
Source · Economic Times chip-actions timeline, 2020
2022US
CHIPS and Science Act
About USD 52 billion in subsidies and incentives to reshore semiconductor manufacturing and research, securing leading-edge fabs for AI, 5G and cloud.
Source · NDTV chip-war timeline, Aug 2022
2022US
Sweeping AI-chip export controls
Broad controls on advanced accelerators (A100 and H100 class) and high-end fab equipment for China, citing military-AI risk; US persons barred from supporting advanced Chinese fabs.
Source · Economic Times; NDTV, Oct 2022
2023China
Gallium and germanium controls
Beijing introduces export controls on critical minerals used in chips and EVs, the mirror image of US chip controls and a signal it can weaponise upstream materials.
Source · WEF policy timeline, 2023
2025China
DeepSeek and the efficiency move
A competitive Chinese model delivered at far lower compute requirements gains international traction, evidence that controls have not prevented world-class AI.
Source · TecEx timeline, 2025
2025US
The H20 reversal and the bright line
Nvidia’s China-tuned H20 is first blocked, then allowed to resume; Commerce reiterates that using Huawei’s top Ascend chips violates controls.
Source · NDTV; China-Briefing, 2025
2025US
Stargate and democratic AI rails
The US pairs a USD 500 billion domestic AI build with an export push: sovereign clouds, “OpenAI for Countries” and an executive order promoting the American AI stack abroad.
Source · Chatham House; OpenAI, 2025
2025–26China
Integrated stacks for the Global South
Beijing exports turnkey packages, 5G, cloud, smart-city and surveillance systems, digital IDs and payments, backed by state finance, as a route to digital modernisation.
Source · Chatham House; Warsaw Institute, 2025–2026
2026Both
Guardrails talks begin
US and Chinese officials start exploratory discussions on AI guardrails: limited transparency, crisis hotlines for cyber and AI incidents, and informal norms, to prevent escalation, not to end the contest.
Source · Wall Street Journal, 2026
The Unrest Tracker ENT-USA-UNREST / 2026.06 · reported and modelled

The build has met the neighbours.

In the United States the bottleneck on compute is no longer capital or even power. It is consent. Across more than forty states, residents, county boards and courts have turned data centres into a live political fight over land, water, noise and who pays for the grid. This tracker maps where the resistance is, the instruments it uses, and the politicians lining up for and against. Figures are reported or modelled and contested where noted; representative visuals.

Where resistance has landed

Six fronts in the fight over the build

Each card is a place where a data-centre project ran into organised opposition, a moratorium, a water fight or a courtroom. Open any case for the detail.

Cases are representative and reported; details and figures should be re-verified before publication. Tap a card for sources.
The toolkit of resistance

Five instruments communities use to slow, shape or stop a build

Opposition rarely beats a hyperscaler on capital. It wins on procedure: the levers below are how a county turns a render into a negotiation.

A temporary halt on accepting or approving new data-centre applications while a jurisdiction writes rules. It buys time and shifts the default from yes to wait.
What it does
Freezes the pipeline for months to a year, often pending a zoning study or comprehensive-plan update.
Where it appears
Township and county boards in Pennsylvania, Virginia, Indiana and Georgia, among others.
Operator counter
Lock in entitlements before a moratorium lands; lobby for state pre-emption of local pauses.
Minimum distances from homes and schools, noise caps, height and lot rules, and dedicated overlay districts. These do not forbid data centres; they make the cheap sites unbuildable.
What it does
Pushes campuses away from residents and raises land and engineering cost; noise limits target cooling-fan and generator sound.
Where it appears
Suburban councils in California and fast-growing metros writing distance and decibel rules.
Operator counter
Site on industrial land far from housing; invest early in sound mitigation and community benefit agreements.
Caps or disclosure requirements on cooling-water draw, or mandates for closed-loop and air cooling. In water-stressed states this is the objection that resonates fastest.
What it does
Forces water-use disclosure and can require lower-water cooling, raising capital cost and, sometimes, energy use.
Where it appears
Arid and drought-prone states including Nevada, Arizona and parts of Texas.
Operator counter
Commit to closed-loop cooling and reclaimed water; publish water-usage effectiveness figures.
Discretionary approvals that let a board attach conditions, demand studies, or simply refuse. They convert a by-right build into a negotiation the community can lose only by being out-lawyered.
What it does
Gives elected boards leverage to extract conditions on jobs, power, water, traffic and bonds, or to deny.
Where it appears
Rural counties such as Box Elder, Utah, where discretionary permitting is the main control.
Operator counter
Negotiate host agreements early; offer tax revenue and grid upgrades to win the discretionary vote.
Tariff structures and contracts that make large loads pay their own grid and generation costs, so that households do not subsidise hyperscale demand. This is the instrument that turns a NIMBY fight into a mainstream one.
What it does
Creates large-load tariff classes, minimum-take contracts and cost-allocation rules so data centres carry their own infrastructure.
Where it appears
Utility commissions and legislatures in Ohio, Virginia, Georgia and others debating who pays for new capacity.
Operator counter
Agree to pay-your-own-way tariffs in exchange for speed and certainty; co-fund generation directly.
For, against, and in between

Three stances, and a coalition that scrambles the usual lines

Data-centre politics do not map cleanly onto left and right. A pro-build administration, a centre that wants investment with protection, and a populist left-right bloc that wants pauses now share the same arena. Positions are paraphrased from public statements and actions; no quotations are invented.

Stance 01 · Accelerate
Build, and clear the path
Treat compute as strategic national infrastructure. Cut permitting friction, lean on federal land and power, and frame the build as an AI race the US must win.
Donald Trump
President
Public position: has promoted accelerating AI data-centre and energy build-out, including executive actions to speed permitting and power for compute.
Source · White House actions; press reporting, 2025
David Sacks
White House AI & crypto advisor
Public position: has publicly argued the US must build compute aggressively to stay ahead, and against measures he views as slowing the build.
Source · public statements, 2025
Stance 02 · Build, but pay
Welcome it, protect the ratepayer
Accept the investment and jobs, but insist that data centres carry their own power costs, disclose water use, and give communities a real say. The centre-ground position in contested states.
Abigail Spanberger
Virginia
Public position: has emphasised protecting households from data-centre-driven power costs while supporting the industry’s economic role.
Source · campaign and public statements, 2025
Mikie Sherrill
New Jersey
Public position: has highlighted electricity affordability and grid impact as data-centre demand grows.
Source · public statements, 2025
JB Pritzker
Illinois
Public position: has courted data-centre investment while facing pressure on power costs and siting.
Source · public statements, 2025
Stance 03 · Pause
Slow down, scrutinise, limit
Want moratoria, hard limits or close scrutiny, citing power bills, water, land and corporate power. A left-right populist bloc that agrees on the brakes if not the reasons.
Bernie Sanders
US Senate · Vermont
Public position: has criticised the power and resource demands of AI data centres and their costs to ordinary consumers.
Source · public statements, 2025
Josh Hawley
US Senate · Missouri
Public position: has raised concerns about data centres raising electricity costs for households.
Source · public statements, 2025
Abdul El-Sayed
Michigan
Public position: has voiced scepticism about data-centre deals and their costs to residents and the grid.
Source · public statements, 2025
First page of the Artificial Intelligence Data Center Moratorium Act
Exhibit · federal · the hardest instrument

A bill to halt the build outright

The Artificial Intelligence Data Center Moratorium Act, introduced in the US Senate by Bernie Sanders, would bar the construction or upgrading of large AI data centres until federal law guarantees product safety review, worker protection, and community power to approve or reject a campus, with reporting on water, energy and emissions. It is the maximal form of the pause stance: where a county moratorium buys a year, a federal moratorium would stop the build nationwide.

US Senate bill text, 119th Congress · public legislative record; status and prospects to be verified before publication.
States by instrument

Who is reaching for which lever

A representative read of where each instrument is adopted, active, proposed or absent across the six tracked fronts. Below, the same states plotted by opposition intensity against how pro-build state policy is.

State / front Moratoria Setbacks Water limits Special-use Ratepayer
Virginia
Data Center Alley
Active Adopted Proposed Adopted Active
Pennsylvania
PJM grid
Adopted Active None Active Proposed
Wisconsin
Ballot front
Proposed Active None Adopted Proposed
California
Suburban front
Active Adopted Proposed Adopted None
Nevada
Water front
Proposed Active Adopted Active None
Utah
Box Elder
None Proposed Proposed Adopted None
Adopted · in force   |   Active · in use or in litigation   |   Proposed · under debate   |   None · not a primary lever here. Representative reading; re-verify locally.
Opposition intensity vs pro-build state policy
Hover a point
By the numbers Reported or modelled, across roughly forty states. Figures should be re-verified before publication.
$156bn
in US data-centre projects blocked or stalled by local opposition through 2025
188
organised opposition groups active across roughly 40 states
20+
projects canceled or paused in early 2026 amid local pushback
+$37
projected monthly rise in a Virginia household power bill by 2040
70%
polled expressing unease about a large data centre near their home
The read

The next phase is fought one county at a time.

The United States can finance and power almost any amount of compute. What it cannot do, at least not quickly, is force a county to accept it. The opposition is now organised, bipartisan in its anger about power bills, and armed with instruments that work: a moratorium buys a year, a special-use permit buys conditions, a courtroom can void an entitlement outright. For operators and investors the lesson of 2025 and 2026 is that the binding diligence question has moved from can it be financed to will the community, the utility and the court allow it.

Sources: industry and press tallies of blocked and stalled projects (2025–2026); state utility-commission and legislative records; local reporting on township, county and ballot actions; public statements and actions by the named officials. Cases, figures and instrument readings are reported or modelled and should be re-verified before publication. Photography is representative editorial art, to be replaced with sourced imagery.

The story

Stargate, on the ground.

The number that opened 2025 was up to 500 billion dollars. The question this report keeps asking is the same one it asks everywhere: how much of that is announced, and how much is rising on real land, drawing real power, beside real communities. Here is where Stargate actually is – and the questions that travel with it.

Scroll to follow the build
STAGE01/ 09
The one that is real
Abilene, Texas
The Lancium Clean Campus, built out by Crusoe and anchored by Oracle for OpenAI. First buildings were energised through 2025 and reported partially live by September. This is the site that turns the headline into concrete and steel.
approx 1.2gigawatts planned across eight buildings
The contested one
Saline Township, Michigan
Related Digital’s campus, nicknamed “The Barn”: three buildings of roughly 550,000 square feet, targeted for end of 2027. The board approved it 4 to 1 only after Related Digital sued over exclusionary zoning and the township settled; OpenAI and Oracle broke ground in June 2026.
16bndollars in planned investment
The widening footprint
Five more sites named
Through 2025 the venture named further sites: Shackelford County and Milam County in Texas, Doña Ana County in New Mexico, Lordstown in Ohio, and a location in Wisconsin – a pipeline aimed at multiple gigawatts.
7+gigawatts in the stated pipeline
The White House · the announcement
01 · The announcement

A number, before a building

On 21 January 2025, a day after the inauguration, OpenAI, SoftBank and Oracle stood beside the president to announce Stargate: a joint venture to build AI infrastructure in the United States, with SoftBank carrying financial responsibility and OpenAI running operations.

The framing was deliberately vast: up to 500 billion dollars over four years, with 100 billion said to be deploying immediately. MGX, Arm, Microsoft and NVIDIA were named as partners. Almost none of it existed yet on the ground.

500bnUSDHeadline, four-year
100bnUSDSaid to deploy immediately
Announced Source: White House announcement; OpenAI; SoftBank; press reporting, Jan 2025. Photo: pool / wire, to be licensed.
02 · Where it really is

It starts in Abilene

The first real Stargate is in Abilene, Texas – the Lancium Clean Campus, built out by Crusoe and leased through Oracle to OpenAI. Buildings were energised across 2025 and reported partially live by September. A planned expansion of the flagship campus was shelved in March 2026, with Meta reported to be weighing the freed-up site.

The map is the gap between a press release and a permit. Tap the markers: one site is pouring concrete, one is in court, and five more are still names on a pipeline.

Tap a marker to inspect a site
VerifiedAnnounced Source: Crusoe; Oracle; OpenAI; DCD; Reuters, 2025. Representative national map; site positions indicative.
03 · Concrete and steel

What the headline looks like as buildings

On the ground the first campus is real hardware: the Lancium Clean Campus, built out by Crusoe and leased through Oracle to OpenAI. Eight buildings are planned, with first halls energised across 2025 and reported partially live by September.

A press release is a sentence; this is a substation, a water line and a construction crew. It is also the exception – the one site where the money has already turned into steel.

1.2GWPlanned across the campus
8Buildings in the build-out
VerifiedAnnounced Source: Crusoe; Oracle; OpenAI; DCD, 2025. Representative aerial; final site imagery to be sourced.
04 · The Barn, Michigan

A 4-to-1 vote, then a lawsuit

In Saline Township, Related Digital’s 16 billion dollar campus – three buildings of roughly 550,000 square feet each, targeted for end of 2027 – won a 4 to 1 township board vote only after Related Digital sued over exclusionary zoning and the township settled. OpenAI and Oracle broke ground in June 2026.

The developer points to a closed-loop cooling design to limit water draw. The dispute is the pattern this report tracks everywhere: the deal clears the local body, and the community contests it after.

4–1Township board vote
2027Stated completion target
AnnouncedContested Source: Related Digital filings; township board records; local reporting, 2025. Render: developer / Related.
05 · Who pays for the power

The fight is about the bill

After the vote, residents took it to the street and to court. The placards are blunt – “Who’s paying for the power?” and “Keep Michigan pure” – and they name the real dispute. It is less about the buildings than about who carries the cost of the electricity and water the load will draw.

A gigawatt campus needs new generation and new transmission. Whether the bill for that lands on the hyperscaler or on every other ratepayer in the state is the question a township vote cannot settle.

Contested Source: WXYZ Detroit; local reporting, 2025. Photo: WXYZ, to be licensed.
06 · Where the bill lands

It ends at the utility commission

The cost question does not stay local. It moves to the state utility regulator – in this case the Michigan Public Service Commission – which rules on how the cost of new generation and transmission is allocated.

Bodies like it, across many states, are quietly becoming the decisive venue of the build-out. Whether data-centre load pays its own way or is socialised across ratepayers is settled here, in hearings most residents never see.

Verified Source: Michigan Public Service Commission records; reporting, 2025. Photo: MPSC, to be licensed.
07 · Announced vs committed

How much of 500 billion is real?

The honest answer is that the headline is a ceiling, not a balance sheet. Some of Stargate is contracted and under construction. Some is announced intent. Some is contested. Sort it for yourself.

08 · The footprint expands

From one site to a pipeline

By late 2025 the venture had named additional sites – Shackelford and Milam counties in Texas, Doña Ana County in New Mexico, Lordstown in Ohio, and a Wisconsin location – with a stated ambition past 7 gigawatts of capacity.

Each new dot is a rural county most of the country could not place on a map – a new grid interconnection, a new water question, and a new commission. The build is no longer a single campus; it is a national programme negotiating itself one jurisdiction at a time.

7+GWStated capacity ambition
6States with named sites
Announced Source: OpenAI / Stargate site announcements; regional reporting, 2025.
09 · The questions that travel with it

Speed, meet scrutiny

Stargate compresses a decade of infrastructure into a few years. That speed is the point – and the problem. Closed-loop cooling reduces but does not erase water demand. Gigawatt loads lean on gas turbines and strained grids. And every site asks a community to accept a national bet on local terms.

The numbers above are the announcement. The film below is the ground.

Source: Entelligencia analysis. Figures are status-banded above; treat the headline as a ceiling.
Featured reportNews
A report by
Sky News

Inside Data Center Alley.

Stargate’s gigawatts have to land somewhere. In this Sky News report, correspondents visit the heart of America’s data-centre boom – Northern Virginia’s “Data Center Alley”, the largest concentration of data centres on earth – where the build is colliding with the communities and the power grid around it. We feature it as a real-world counterpoint to the announcement above.

Featured report · broadcast news
Sky News@SkyNews
Sky News. A broadcast report from inside Northern Virginia’s “Data Center Alley”, the largest data-centre cluster on earth, featured here as a counterpoint to the build-out this chapter documents.
News By · Sky News Subject · Data Center Alley

Featured third-party report. This Sky News broadcast is embedded under YouTube’s standard player and credited to its makers; it is not commissioned by or affiliated with Entelligencia, and is shown as editorial counterpoint to the chapter. Poster image is the video’s YouTube thumbnail. Confirm embedding permissions and credit before publication.

The Brief

No campus survives contact with the county.

ENT-USA-BRIEF / 2026.06
Tracks · 4
View · summary, then the full brief

In the Gulf the render outruns the budget. In the United States the money is the easy part. Capital here is abundant and patient; what bends a data-centre render is the ground beneath it – the land, the megawatts, the water table and the county commission that has to vote yes. Scroll each track left to right: the projection resolves into what stands today, what private capital has financed, or what a court erased. Figures are reported or modelled and contested where noted.

Meta Hyperion, Louisiana
Stage 01 / 03
Drag, scroll or use the arrows · projection fades to today
The full brief · longer reads The tracks above are the summary. Open any brief for the full arc, slide by slide, with a build-reality node behind each claim.
All four briefs carry final imagery; the build-curve brief flashes a photo montage of several images per beat. Figures are reported or modelled and contested where noted. Representative visuals; final imagery to be sourced. Re-verify at build.
How the pattern resolves

Read every US campus as land plus power plus consent, not just capital.

In the United States the announced number is rarely the fiction; the financed build usually follows, and fast. The variable that decides whether a render becomes a building is local: an interconnection queue, a water table, a county commission, a courtroom. For an investor or operator the discipline is to separate the financed project from the permitted one. The capital is real and the demand is real, but the megawatt only becomes an asset once the grid, the county and the court all agree. The render is the marketing; consent is the gate.

At a glance · the index

America’s preconditions, scored.

Each spoke is one precondition, scored 0 to 10 for like-for-like reading – indicative editorial weighting, not a published index. Select any spoke or row to read the factor.

The people

Who actually decides America’s build-out.

Five named figures across the AI-compute builders, the silicon chokepoint, federal policy and the state politics of cost and consent. Each card carries a public-record position, never an invented quote; named roles and figures should be re-verified against current company and government records before publication. Portraits are reference images and remain subject to licensing clearance before publication.

Portrait of Sam Altman
Flip
OpenAI · demand
Sam Altman
Chief Executive · OpenAI
Profile · 01 of 05
Sam Altman
Chief Executive, OpenAI
Influence
Drives the largest single source of AI-compute demand in the US. OpenAI is the lead sponsor of the Stargate buildout announced in January 2025.
Working on
Securing hundreds of billions of dollars of compute through Stargate with SoftBank, Oracle and partners, and multi-gigawatt site development across several states.
Public position: Altman has argued that abundant compute and energy are prerequisites for advancing AI, and has championed very large-scale US infrastructure.
Paraphrase of public statements and the Stargate announcement · 2025
Portrait of Larry Ellison
Flip
Oracle · infrastructure
Larry Ellison
Chair & CTO · Oracle
Profile · 02 of 05
Larry Ellison
Chairman and Chief Technology Officer, Oracle
Influence
Oracle is a core Stargate partner and a fast-growing AI-cloud landlord. Ellison directs its data-centre and cloud strategy and its push into renting large GPU clusters.
Working on
Building and leasing GPU capacity, including the Abilene, Texas campus tied to Stargate, and positioning Oracle Cloud as an AI-infrastructure provider.
Public position: Ellison has emphasised Oracle’s role in building the physical infrastructure that large AI models run on.
Paraphrase of Oracle earnings remarks and the Stargate announcement · 2025
Portrait of Jensen Huang
Flip
NVIDIA · silicon
Jensen Huang
Founder & CEO · NVIDIA
Profile · 03 of 05
Jensen Huang
Founder and Chief Executive, NVIDIA
Influence
NVIDIA’s accelerators are the chokepoint of the AI build. Its allocation decisions shape who can scale, and US export controls on its China-bound chips sit at the centre of the cold war.
Working on
Supplying the GPUs behind nearly every US hyperscale campus, while navigating export rules that restrict its most advanced chips from China.
Public position: Huang has argued that broad access to US compute strengthens the American platform, and has cautioned that overly tight export controls can push customers toward rivals.
Paraphrase of public statements · 2024 to 2025
Portrait of David Sacks
Flip
White House · policy
David Sacks
AI & Crypto Advisor
Profile · 04 of 05
David Sacks
White House AI and Crypto Advisor
Influence
The administration’s point person on AI policy. Shapes the federal posture on permitting, energy and compute that frames the national build.
Working on
Pushing to accelerate US compute and energy build-out and to keep the US ahead of China on AI infrastructure.
Public position: Sacks has framed AI as a race the US must win, arguing for building compute aggressively while easing constraints he views as slowing it.
Paraphrase of public statements · 2025
Portrait of Abigail Spanberger
Flip
Virginia · politics
Abigail Spanberger
Governor · Virginia
Profile · 05 of 05
Abigail Spanberger
Governor of Virginia
Influence
Leads the state at the centre of the data-centre economy and its backlash. Virginia’s approach to ratepayer protection is the template other states watch.
Working on
Balancing data-centre investment with protecting households from rising power costs and giving localities more say over siting.
Public position: Spanberger has emphasised protecting Virginia ratepayers from data-centre-driven electricity costs while supporting the industry’s economic role.
Paraphrase of campaign and public statements · 2025. Re-verify current office before publication.
The supply chain

From capital to compute, who is active in the United States.

US compute stack · interconnected
Hover a node to trace its links · click for the brief
Capital
Blue Owlprivate credit
SoftBankStargate lead
MGXUAE · AI fund
BlackstoneDC platforms
KKRinfra · PE
Brookfieldinfra · power
Suppliers
NVIDIAGPUs
AMDGPUs · CPUs
Vertivpower · cooling
Schneider Electricpower · cooling
Supermicroservers
Dellservers · integration
Power & energy
Constellationnuclear · PPAs
Talen Energynuclear · PJM
Entergyutility · gas
Vistrageneration
NextErarenewables
Operators & REITs
Equinixinterconnection
Digital RealtyREIT · wholesale
CoreWeaveAI-native cloud
Vantagehyperscale builder
CrusoeAbilene · Stargate
QTSBlackstone platform
Switchcampus operator
Hyperscalers & demand
Microsoftcloud · OpenAI
Googlecloud · AI
Amazon (AWS)cloud · AI
MetaHyperion · AI
Oraclecloud · Stargate
OpenAIAI · Stargate
Flow

Capital sets the pace and is abundant: private credit and sovereign money underwrite gigawatt campuses before they are built. Suppliers and the power layer gate what can actually energise; operators and REITs raise and lease the halls; and the hyperscalers and AI labs anchor the demand the whole stack is sized against. The US signature is that the chain is deep and contested: many independent actors, a binding power constraint, and a local-consent gate the Gulf does not face.

05 tiers · 30 entities

Logos identify the parties for evaluation only, not as an endorsement, and remain subject to each owner’s brand-usage rights before any sponsor or commercial use; names shown without a logo are indicative of the categories active across the US market. Connections are Entelligencia’s reading of public deal and partnership announcements, to be confirmed before publication.

Voices of the industry

[Placeholder] What the people building it say.

Three industry voices on the supply chain, each opening a pop-out that can carry a voice clip, an interview, an opinion with exhibits, or a simple quote. Placeholder cards for now; real names, photos and content to follow.

Voice clip
[Name to come]
[Role] · [Organisation]

[Placeholder] A short standfirst on what this voice adds to the supply-chain picture.

Listen
Interview
[Name to come]
[Role] · [Organisation]

[Placeholder] A short standfirst on what this voice adds to the supply-chain picture.

Read
Opinion
[Name to come]
[Role] · [Organisation]

[Placeholder] A short standfirst on what this voice adds to the supply-chain picture.

Open

Placeholder module · voices, photos and content to be added chapter by chapter.

The wider programme

More in Enable Americas.

This chapter is the briefing. Enable Americas is the full intelligence product: the synthesis edition of the global programme, with the United States in focus.
Edition 01 Brazil The scale chapter
Edition 02 Chile The Pacific gateway
Edition 03 Colombia The Andean chapter
Edition 04 Enable Americas The synthesis · US in focus
Edition 04 Enable Americas flagship cover
The flagship · publishing 2028

The definitive map of who is delivering the Americas’ compute, with the United States in focus.

Operator by operator, county by county: what is financed, what is permitted, and what a court or community is disputing. Site visits, named interviews and the people-map of the market, held to one standard, that nothing is published until it is verified.

The editions so far
Three chapters into the Americas
01
Enable Brazil
02
Enable Chile
03
Enable Colombia

Each prior edition mapped one market end to end. Enable Americas draws them together, with the United States at the centre of the synthesis.

By invitation
One convening
Q3 2028
The Deployment
The Enable Americas launch, in the room where the Americas’ compute is financed and built. The convening is what they paid to be in.