The U.S. AI buildout. | Mingyang Fund

Project / owner / tenantStateCapacity scopeInformation available as of

U.S. LARGE AI BUILDOUT / LOCALITY CENTROIDS

Loading source-backed geography…

Marker size: estimated IT MWGrouped by Census locality; size = largest group

Epoch AI · AI Data Centers ↗ · U.S. Census · 2025 Gazetteer places ↗ · Natural Earth · 1:110m countries ↗
City markers do not identify exact campus boundaries. Unmapped groups remain in the directory. No physical grid connection is inferred from proximity.

Compare	Project / location	Owner / users	Current IT MW	Evidence basis	Source status

Current capacity is an Epoch estimate at the 11 Sep 2026 snapshot, not metered load. Program targets, planned capacity, actual power and operating capacity are kept separate. Same-address source groups share a campus ID; physical overlap across nearby addresses remains unresolved.

Regional power context

Load growth is a constraint on delivery.

EIA identifies ERCOT and PJM as major regions of data-center demand growth. Its high-demand case is a scenario, not a probability-weighted forecast.

EIA · Electricity demand scenarios ↗

No grid boundary or plant-to-project supply connection has been verified in this build. Geography alone does not establish a utility contract, delivered electricity or an energized campus.

Project diligence checklist

Track each gate separately.

Permit
Approved use and phase scope · not disclosed in current dataset

Interconnection
Queue position, study and upgrade obligations · not disclosed in current dataset

Power agreement
PPA / tariff terms and conditions precedent · not disclosed in current dataset

Energization
Verified date and delivered capacity · not disclosed in current dataset

Actual load
Metered usage, utilization and ramp · not disclosed in current dataset

Test utilization and delay →

Explore the capital flow.

Compare four companies and follow their spending into supplier exposure.

Open interactive CapEx →

Company financials provide context; these are not project budgets and are not AI-only CapEx. Cash P&E and reported CapEx differ in their lease treatment.

Company / period	Metric	USD bn	Basis	Evidence
MSFTQ4 FY2026	Cash purchases of P&E	35.8	Cash investing outflow; total company	Microsoft Q4 FY2026 ↗Public 2026-07-29
MSFTQ4 FY2026	Operating cash flow	55.4	Reported total company	Microsoft Q4 FY2026 ↗Public 2026-07-29
MSFTQ4 FY2026	Reported capital expenditure	41	Reported capital expenditures include finance leases; chart uses cash paid for property and equipment.	Microsoft Q4 FY2026 ↗Public 2026-07-29
METAQ2 2026	Cash purchases of P&E	30.116	Cash investing outflow; total company	Meta Q2 2026 ↗Public 2026-07-29
METAQ2 2026	Operating cash flow	31.862	Reported total company	Meta Q2 2026 ↗Public 2026-07-29
METAQ2 2026	Reported capital expenditure	31.08	Reported capital expenditures include principal payments on finance leases; chart uses purchases of property and equipment.	Meta Q2 2026 ↗Public 2026-07-29
NVDAQ2 FY2027 · ended July 26, 2026	Data Center revenue	89	Reported segment revenue	NVIDIA Q2 FY2027 results ↗Public 2026-08-26

Only selected verified reporting periods are loaded. No forecast consensus or comprehensive CapEx revision series is represented.