04Buildout

The gigawatt land-grab

Follow the money: what hyperscalers are spending, what NVIDIA is booking, and where the gigawatt-class campuses are actually rising.

Big-4 capex guidance · 2026

$710B

≈1.7× 2025 actuals · MSFT + GOOG + META + AMZN midpoints

NVIDIA DC revenue · FY27 Q1

$75.2B

~$300B/yr run-rate

Tracked frontier pipeline

~36 GW

Sum of planned capacity, clusters below

Largest single campus

Hyperion

Meta · up to 5 GW

Big-tech capital expenditure

US$B per calendar year (2026 = guidance midpoints)

Static chart frame · exact values in table view

Earnings releases & SEC filings, as of Jul 2026. Microsoft incl. finance leases; Oracle fiscal years mapped to prior calendar year.

NVIDIA data-center revenue

US$B by fiscal quarter (FY ends late January)

Static chart frame · exact values in table view

NVIDIA quarterly releases. Next-quarter guidance ~$91B total revenue, assuming zero China DC sales.

Frontier-scale clusters & campuses

26 sites · 9 operating · ~36 GW planned

Provenanced site ledger, surveyed Aug 2026. Power figures are IT load where trackers publish it and campus power where only the operator’s headline exists — open Provenance for each figure’s dated source.

Frontier-scale AI datacenters
		Location		Accelerators
Hyperionest — estimated / secondary-source figure Provenance for Hyperion Planned figure is Meta’s 5 GW campus target; tracker models ≈1.7 GW IT through 2028, first power 2028. Planned power · est Expanded to 5 GW of compute capacity, >$50B investment — CNBC, 2026-07-13 Accelerators · est ≈481K NVIDIA Rubin-class chips projected (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Announced · est Announced and broke ground December 2024 — Louisiana Economic Development, 2026-07	Meta	Richland Parish, Louisiana	— → 5 GW	NVIDIA Rubin-era planned (~481K chips modeled)	Building
Anthropic–Google TPU fleetest — estimated / secondary-source figure Provenance for Anthropic–Google TPU fleet Capacity program, not a single campus: >1 GW in 2026 plus ~3.5 GW of next-gen TPU capacity from 2027. Planned power · est >1 GW of TPU capacity in 2026 (deal worth tens of billions) — CNBC, 2025-10-23 Planned power ≈3.5 GW additional TPU compute from 2027 (Google/Broadcom expansion) — Anthropic, 2026-04-06 Accelerators · est Up to 1M Google TPUs under the Oct 2025 agreement — CNBC, 2025-10-23 Announced · est Google Cloud TPU expansion announced Oct 23, 2025 — CNBC, 2025-10-23	Anthropic · Google Cloud	Multiple US sites	— → 4.5 GW	Up to 1M TPUs (v7 Ironwood era); next-gen TPUs via Broadcom from 2027	Building
Jamnagar AI campusest — estimated / secondary-source figure Provenance for Jamnagar AI campus Reported 3 GW target — would be the world’s largest by capacity; $20–30B reported cost. Planned power · est 3 GW data center planned in Jamnagar (report) — TechCrunch, 2025-01-23 Accelerators · est NVIDIA supplying chips under the Oct 2024 Reliance partnership — TechCrunch, 2025-01-23 Announced · est Gigawatt-scale Jamnagar plan reported Jan 2025 — Data Center Dynamics, 2025-01	Reliance Industries	Jamnagar, Gujarat, India	— → 3 GW	NVIDIA GB200-era (Reliance–NVIDIA partnership)	Announced
Fairwater Wisconsinest — estimated / secondary-source figure Provenance for Fairwater Wisconsin “World’s most powerful AI datacenter” claim; first facility complete Jun 2026, campus path to ~2 GW by 2028. Current power · est ≈369 MW IT live (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Planned power · est ≈2.26 GW IT projected by Q2 2028 (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Accelerators “Hundreds of thousands” of NVIDIA GB200 as a single cluster — Microsoft, 2026-06-23 Announced Fairwater Wisconsin unveiled Sep 2025 — Microsoft, 2025-09-18 First power · est First operational April 2026; construction complete Jun 23, 2026 — Epoch AI data-center directory, 2026-08-20	Microsoft	Mount Pleasant, Wisconsin	369 MW → 2.3 GW	Hundreds of thousands of GB200 in one cluster (~176K B200 live, tracker estimate)	Operating
Stargate Doña Ana County (Project Jupiter)est — estimated / secondary-source figure Provenance for Stargate Doña Ana County (Project Jupiter) Planned power · est ≈2.2 GW planned, targeting Q4 2028 (tracker estimate) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est ≈4.6M H100-equivalents of compute projected — Epoch AI — Stargate US sites, 2026-04-17 Announced Named in the five-new-Stargate-sites announcement — OpenAI, 2025-09-23	OpenAI · Oracle · STACK	New Mexico	— → 2.2 GW	GB300-era NVIDIA (≈4.6M H100-equivalents projected)	Building
Project Rainierest — estimated / secondary-source figure Provenance for Project Rainier Flagship of the Amazon–Anthropic expansion (up to 5 GW, >$100B AWS commitment over 10 years). Planned power · est Full site: 30 buildings drawing up to 2.2 GW — CNBC, 2025-10-29 Accelerators · est Nearly 500K Trainium2 across seven buildings at opening — CNBC, 2025-10-29 Accelerators Nearly 1 GW of Trainium2+3 capacity online across AWS by end-2026; up to 5 GW total — Anthropic, 2026-04-20 Announced · est Project Rainier disclosed with Anthropic as anchor, Dec 2024 — CNBC, 2025-10-29 First power · est Campus activated with ~$11B invested, Oct 2025 — CNBC, 2025-10-29	Anthropic · AWS	New Carlisle, Indiana	— → 2.2 GW	~500K Trainium2 live; ~1 GW Trainium2+3 across AWS by end-2026	Operating
Colossus 2est — estimated / secondary-source figure Provenance for Colossus 2 First single ~gigawatt-class AI datacenter; third building (“MACROHARDRR”) takes the site toward 2 GW. Current power · est ≈1 GW of AI compute across the Memphis complex, mid-2026 — Wikipedia — Colossus (data center), 2026-08 Planned power · est Site capacity near 2 GW after the Dec 2025 third-building purchase — Introl, 2026-01-03 Accelerators · est 555K+ NVIDIA GB200/GB300-class GPUs, ≈$18B — Introl, 2026-01-03 First power · est First ~110K-GB200 (~210 MW) cluster online ~Jan 2026 — Wikipedia — Colossus (data center), 2026-08	xAI	Memphis, Tennessee	1 GW → 2 GW	~555K GB200/GB300 delivered for the site (~$18B)	Operating
Stargate Shackelford Countyest — estimated / secondary-source figure Provenance for Stargate Shackelford County Vantage “Frontier” campus; first building targeted late 2026. Planned power · est ≈2.0 GW planned (tracker estimate) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est ≈4.2M H100-equivalents of compute projected — Epoch AI — Stargate US sites, 2026-04-17 Announced Named in the five-new-Stargate-sites announcement — OpenAI, 2025-09-23	OpenAI · Oracle · Vantage	Texas	— → 2 GW	GB300-era NVIDIA (≈4.2M H100-equivalents projected)	Building
Stargate Saline Townshipest — estimated / secondary-source figure Provenance for Stargate Saline Township Related Digital build southwest of Ann Arbor; formal Michigan expansion event Jun 2026. Planned power · est ≈1.4 GW planned, targeting Q4 2028 (tracker estimate) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est ≈2.9M H100-equivalents of compute projected — Epoch AI — Stargate US sites, 2026-04-17 Announced · est Michigan Stargate expansion event with OpenAI and Oracle — CNBC, 2026-06-01	OpenAI · Oracle · Related Digital	Michigan	— → 1.4 GW	GB300-era NVIDIA (≈2.9M H100-equivalents projected)	Building
Stargate Port Washington (Lighthouse)est — estimated / secondary-source figure Provenance for Stargate Port Washington (Lighthouse) Planned power · est ≈1.3 GW planned, targeting Q4 2028 (tracker estimate) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est ≈2.6M H100-equivalents of compute projected — Epoch AI — Stargate US sites, 2026-04-17 Announced OpenAI, Oracle and Vantage announce the Wisconsin Stargate site — Vantage Data Centers, 2025-10-22	OpenAI · Oracle · Vantage	Port Washington, Wisconsin	— → 1.3 GW	GB300-era NVIDIA (≈2.6M H100-equivalents projected)	Building
Stargate Abileneest — estimated / secondary-source figure Provenance for Stargate Abilene Flagship of the Stargate program — 8 buildings; planned figure is campus facility power (tracker-modeled IT ≈ 0.84 GW). Current power · est ≈421 MW IT live across buildings 1–4 (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Planned power · est 1.2 GW campus at completion (Q4 2026 target) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est ~100.8K B200 + up to ~302K B300 projected (≈450K GB200-class headline) — Epoch AI data-center directory, 2026-08-20 Announced Stargate program unveiled Jan 2025; Abilene the flagship site — OpenAI, 2025-01-21 First power First two buildings energized within a year of June 2024 groundbreaking — Crusoe, 2025-09-30	OpenAI · Oracle · Crusoe	Abilene, Texas	421 MW → 1.2 GW	~450K GB200/GB300-class at full buildout	Operating
Stargate Milam Countyest — estimated / secondary-source figure Provenance for Stargate Milam County SB Energy (SoftBank) build; first building targeted Oct 2026. Planned power · est ≈1.2 GW planned; first building Oct 2026 (tracker estimate) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est ≈2.5M H100-equivalents of compute projected — Epoch AI — Stargate US sites, 2026-04-17 Announced Named in the five-new-Stargate-sites announcement — OpenAI, 2025-09-23	OpenAI · SB Energy	Texas	— → 1.2 GW	GB300-era NVIDIA (≈2.5M H100-equivalents projected)	Building
Prometheusest — estimated / secondary-source figure Provenance for Prometheus Meta’s ~1 GW cluster target; nuclear supply deals (TerraPower, Oklo, Vistra) signed Jan 2026. Current power · est ≈562 MW IT live (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Planned power · est Announced as a ~1 GW multi-building cluster (tracker models 854 MW IT by Q1 2027) — NBC4 Columbus, 2025-07 Accelerators · est ≈269K NVIDIA B200 installed (tracker estimate) — Epoch AI data-center directory, 2026-08-20 First power · est Building 1 operational Oct 14, 2025 (tracker estimate) — Epoch AI data-center directory, 2026-08-20	Meta	New Albany, Ohio	560 MW → 1 GW	~269K B200 (→ ~409K by Q1 2027, tracker estimate)	Operating
El Paso campusest — estimated / secondary-source figure Provenance for El Paso campus $14B venture with BlackRock (80/20); Meta sole initial occupant, operations expected 2028. Planned power 1 GW of compute capacity, operating from 2028 — Meta, 2026-07-28 Accelerators AI-optimized data center; chip platform not yet disclosed — Meta, 2025-10 Announced Groundbreaking announced October 2025 — Meta, 2025-10	Meta · BlackRock	El Paso, Texas	— → 1 GW	AI-optimized campus (chips TBD)	Building
Stargate UAEest — estimated / secondary-source figure Provenance for Stargate UAE 1 GW cluster anchoring the 5 GW UAE–US AI campus; first 200 MW due Q3 2026. Planned power 1 GW Stargate UAE cluster within a 5 GW campus — OpenAI, 2025-05-22 Accelerators NVIDIA Grace Blackwell GB300 systems — OpenAI, 2025-05-22 Announced Stargate UAE announced May 22, 2025 — OpenAI, 2025-05-22 Status First 200 MW phase on track for Q3 2026 delivery — G42, 2025-10	G42 · OpenAI · Oracle	Abu Dhabi, UAE	— → 1 GW	NVIDIA GB300 systems	Building
Visakhapatnam AI hubest — estimated / secondary-source figure Provenance for Visakhapatnam AI hub $15B, Google’s largest AI hub outside the US; ground broken Apr 28, 2026. Planned power · est Gigawatt-scale data center campus (US$15B, 2026–2030) — Adani Group, 2025-10-14 Accelerators Google AI hub infrastructure; accelerator mix not disclosed — Google Cloud, 2026-04-28 Announced Adani–Google partnership announced Oct 14, 2025 — Adani Group, 2025-10-14 Status Google broke ground on the Visakhapatnam AI hub Apr 28, 2026 — Google Cloud, 2026-04-28	Google · AdaniConneX	Visakhapatnam, Andhra Pradesh, India	— → 1 GW	Google AI infrastructure (TPU-era, details undisclosed)	Building
Fairwater Atlantaest — estimated / secondary-source figure Provenance for Fairwater Atlanta QTS-built, grid-only design (no UPS/gen-sets); linked with Wisconsin into one “AI superfactory” WAN. Current power · est ≈636 MW IT live across buildings 1–4 (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Planned power · est ≈636 MW IT at current buildout (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Accelerators · est ≈304K NVIDIA B200 (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Announced · est Fairwater Atlanta launched Nov 2025 — Data Center Dynamics, 2025-11 First power · est Buildings 1–2 operational Oct 14, 2025 (tracker estimate) — Epoch AI data-center directory, 2026-08-20	Microsoft	Fayetteville, Georgia	636 MW	~304K B200 (tracker estimate)	Operating
HUMAIN AI factoriesest — estimated / secondary-source figure Provenance for HUMAIN AI factories PIF-backed sovereign AI buildout; capacity target “up to 500 MW” across factories. Planned power AI factories with a projected capacity of up to 500 MW — NVIDIA, 2025-05-13 Accelerators First phase: an 18,000-GPU NVIDIA GB300 supercomputer — NVIDIA, 2025-05-13 Announced HUMAIN–NVIDIA partnership announced May 13, 2025 — NVIDIA, 2025-05-13	HUMAIN · NVIDIA	Riyadh, Riyadh Province, Saudi Arabia	— → 500 MW	Several hundred thousand NVIDIA GPUs over 5 years; first phase 18K GB300	Building
Stargate Argentinaest — estimated / secondary-source figure Provenance for Stargate Argentina Letter of intent only (up to $25B); local reporting found no binding contracts, site, or permits as of mid-2026. Planned power · est Up to 500 MW capacity under the letter of intent — The Real Deal, 2025-10-21 Accelerators · est Sized to host “the next generation of AI computing”; chips unspecified — The Real Deal, 2025-10-21 Announced · est LOI unveiled at Casa Rosada Oct 10, 2025 — The Real Deal, 2025-10-21 Status · est No binding contracts, financing, site, or permits disclosed as of Jun 2026 — Buenos Aires Times, 2026-06	Sur Energy · OpenAI	Patagonia, Argentina	— → 500 MW	Next-generation AI computing (chips TBD)	Announced
Polaris Forge 1 Provenance for Polaris Forge 1 Applied Digital campus fully leased to CoreWeave across three buildings (~$11B contracted revenue). Current power Building 1 at its full 100 MW critical IT load — Applied Digital, 2025-11-24 Planned power 400 MW critical IT capacity contracted to CoreWeave — Applied Digital (via News Dakota), 2025-08-30 Accelerators · est Purpose-built for CoreWeave AI/HPC workloads — Applied Digital (via News Dakota), 2025-08-30 First power Building 1 fully energized Nov 24, 2025 — Applied Digital, 2025-11-24	CoreWeave · Applied Digital	Ellendale, North Dakota	100 MW → 400 MW	CoreWeave GPU cloud capacity (NVIDIA, models undisclosed)	Operating
Omaha TPU campusest — estimated / secondary-source figure Provenance for Omaha TPU campus Google’s flagship disclosed-scale TPU training campus; expansion projected through Q3 2028. Current power · est ≈237 MW IT live (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Planned power · est ≈395 MW IT projected by Q3 2028 (tracker estimate) — Epoch AI data-center directory, 2026-08-20 Accelerators · est ≈181K TPU v5e + 59K v5p + 48K v6e (tracker estimate) — Epoch AI data-center directory, 2026-08-20 First power · est Building 1 operational Apr 2024 (tracker estimate) — Epoch AI data-center directory, 2026-08-20	Google	Omaha, Nebraska	237 MW → 395 MW	~288K TPUs (v5e/v5p/v6e, tracker estimate)	Operating
Colossus 1est — estimated / secondary-source figure Provenance for Colossus 1 First 100K H100 built in 122 days in 2024; ~150 MW grid plus on-site turbines and Megapacks. Current power · est ≈300 MW incl. on-site generation (~150 MW grid connection) — Wikipedia — Colossus (data center), 2026-08 Accelerators · est 230K GPUs: 150K H100, 50K H200, 30K GB200 — Introl, 2026-01-03 First power · est First 100K-GPU cluster operational after a 122-day build, 2024 — Data Center Dynamics, 2024-12	xAI	Memphis, Tennessee	300 MW	~230K GPUs (150K H100, 50K H200, 30K GB200)	Operating
Stargate Lordstownest — estimated / secondary-source figure Provenance for Stargate Lordstown SoftBank/Foxconn advanced data-center design; ground broken 2025, targeted operational 2026. Planned power · est <0.3 GW modeled (tracker estimate) — Epoch AI — Stargate US sites, 2026-04-17 Accelerators · est <0.3M H100-equivalents modeled — Epoch AI — Stargate US sites, 2026-04-17 Announced Named in the five-new-Stargate-sites announcement; ground already broken — OpenAI, 2025-09-23	OpenAI · SoftBank · Foxconn	Lordstown, Ohio	— → 300 MW	NVIDIA (early-stage; <0.3M H100-equivalents projected)	Building
Stargate Norwayest — estimated / secondary-source figure Provenance for Stargate Norway Hydropower-fed AI gigafactory; ambition to add a further 290 MW beyond the initial 230 MW. Planned power 230 MW initial capacity (ambition for +290 MW) — Nscale, 2025-07-31 Accelerators 100,000 NVIDIA GPUs targeted by end of 2026 — Nscale, 2025-07-31 Announced Nscale/Aker/OpenAI announce Stargate Norway Jul 31, 2025 — Nscale, 2025-07-31	Nscale · Aker · OpenAI	Narvik, Nordland, Norway	— → 230 MW	100K NVIDIA GPUs targeted by end-2026	Building
Fluidstack build (Texas)est — estimated / secondary-source figure Provenance for Fluidstack build (Texas) Half of Anthropic’s $50B US datacenter program with Fluidstack; no per-site MW published. Announced $50B US datacenter investment with Fluidstack; Texas and New York first, online through 2026 — Anthropic, 2025-11-12 Accelerators · est Custom-built facilities; Fluidstack picked for ability to deliver “gigawatts of power” — CNBC, 2025-11-12	Anthropic · Fluidstack	Texas	—	Custom Anthropic build (chips undisclosed)	Building
Fluidstack build (New York)est — estimated / secondary-source figure Provenance for Fluidstack build (New York) Half of Anthropic’s $50B US datacenter program with Fluidstack; no per-site MW published. Announced $50B US datacenter investment with Fluidstack; Texas and New York first, online through 2026 — Anthropic, 2025-11-12 Accelerators · est Custom-built facilities; Fluidstack picked for ability to deliver “gigawatts of power” — CNBC, 2025-11-12	Anthropic · Fluidstack	New York	—	Custom Anthropic build (chips undisclosed)	Building