What is the scale and geography of hyperscaler and private equity data center capital deployment in 2026, and what are the CRE investment implications?

Synthesized twelve source packages capturing hyperscaler capital commitments (Amazon, Google, Microsoft, OpenAI), PE platform acquisitions (SoftBank/DigitalBridge, BlackRock/Aligned), developer and operator activity (Crusoe, Flexential, Edged, QTS), and one aspirational developer announcement (MacroValor/Mount Hydrogen). Reviewed existing canonical wiki pages — Texas Digital Infrastructure Corridors, Digital Infrastructure Real Estate, Powered Land and Grid Advantage, and AI Office Demand Engine 2026 — to orient findings within the current knowledge graph and avoid duplicating prior analysis. Data extracted directly from source HTML and source notes; no figures fabricated.

The following table represents disclosed or reported capital commitments from the sources reviewed. Dollar figures are taken directly from source articles; capacity figures where available.

Estimated directly committed capital from named sources (excluding the aspirational MacroValor announcement and the PE platform acquisitions, which represent asset manager consolidation rather than new construction capital): approximately $545B+ in disclosed or reported investment commitments, with the Stargate program alone accounting for the majority. The PE acquisitions ($44B for Aligned and DigitalBridge combined) represent capital already deployed to acquire existing platforms, not new construction.

Interpretation: These are announced commitments across a roughly six-month reporting window (October 2025 – April 2026). The actual pace of draw on these commitments varies significantly — Stargate's $450B+ extends over three years; Google's $40B extends through 2027. The single-period disclosed amount understates the concentration of capital flowing into compute infrastructure; it does not represent a single-year construction budget.

Texas is the single most repeated geography across the sources reviewed.

• Google: ~$40B through 2027 in Armstrong County, Haskell County, and Ellis County (Waxahachie/Midlothian). The Haskell County campus is co-located with a new solar and battery storage plant — a rural, energy-infrastructure-first siting with no metro grid dependency. Ellis County is an existing campus expansion.
• Microsoft: $400M facility in Castroville, Medina County — roughly 25 miles west of San Antonio. Facility known as SAT 82 (~195,000 SF); construction begins August 2026. Microsoft is reported to have additional Castroville projects beyond SAT 82 in the pipeline.
• OpenAI (Stargate): Texas is among the earliest announced Stargate states. Precise Texas site details are not disclosed in the reviewed sources, but Texas appears in multiple Stargate references as a prior announced site.
• Edged U.S.: Edged Dallas 2 at Irving campus — 24 MW, Q2 2026 construction start, completion fall 2027. A smaller-scale enterprise/colocation build consistent with the Irving/Las Colinas corporate corridor market.
• Aligned Data Centers (now BlackRock-owned): Dallas-origin platform that grew from two original campuses (DFW and Phoenix) to 50+ global campuses before the BlackRock GIP acquisition.
• MacroValor/Favis (Mount Hydrogen): Austin, 3 GW aspirational announcement. Status: initial planning only. Filed as watch-list, not confirmed capital.

Texas receives representation in hyperscale buildout (Google, OpenAI Stargate), exurban data center development (Microsoft/Castroville), enterprise-tier operator expansion (Edged/Irving), and PE-level platform origination (Aligned). No other single state appears across this many distinct capital categories in the same review window.

The Texas concentration extends across tiers: metro-integrated campuses (Ellis County DFW extension, Edged Dallas), exurban/satellite builds (Castroville exurban ring around San Antonio), and rural energy-infrastructure-first sites (Haskell County, West Texas). See Texas Digital Infrastructure Corridors for the full corridor-by-corridor comparison.

Two major PE-level transactions from the review window define a consolidation pattern distinct from greenfield construction:

BlackRock GIP / MGX / AIIP acquiring Aligned Data Centers for ~$40B (October 2025)

• Aligned began in 2018 as a Macquarie Asset Management investment with two campuses (Dallas and Phoenix)
• Grew to 50+ campuses across the US, Mexico, Brazil, Chile, and Colombia; 5+ GW operational and planned
• Sellers: Macquarie Asset Management and co-investment partners (this was Macquarie's second major data center exit within 12 months, following the 2024 AirTrunk sale at $16B)
• Buyers: BlackRock Global Infrastructure Partners + Abu Dhabi's MGX + Artificial Intelligence Infrastructure Partnership
• Aligned retained innovative capital-markets structures: first-ever green data center securitization; first-ever sustainability-linked financing for a data center platform
• Expected close: H1 2026 (pending regulatory)

SoftBank acquiring DigitalBridge for ~$4B (January 2026)

• DigitalBridge (NYSE: DBRG; Boca Raton, FL): global digital infrastructure asset manager investing in data centers, cell towers, and fiber networks
• Acquisition price: $16/share — a 50% premium to the unaffected 52-week average as of December 4, 2025
• Marc Ganzi continues to lead DigitalBridge post-close as a separately managed platform
• Expected close: H2 2026
• SoftBank rationale: grow AI and digital infrastructure platforms and capabilities (SoftBank is also a Stargate co-sponsor via the OpenAI/Oracle/SoftBank consortium)

What these two transactions mean for data center real estate pricing:

The $40B Aligned acquisition is the most significant data point. It establishes that a scaled, diversified data center platform — one that originated in Dallas and Phoenix before expanding nationally and internationally — now commands infrastructure-fund pricing from sovereign wealth capital (MGX), institutional infrastructure managers (BlackRock GIP), and AI-mandate vehicles (AIIP). This is no longer a real estate trade; it is a utility-class infrastructure allocation.

The SoftBank/DigitalBridge deal validates a second dynamic: the specialist GP platform itself has M&A value. SoftBank did not acquire underlying data center assets directly — it acquired DigitalBridge's management platform, relationships, and deployment capability. This signals that the scarcest resource in digital infrastructure may not be capital or even land but operational expertise and existing platform relationships with hyperscale tenants.

Pricing implication (interpretation): PE consolidation at $40B and $4B premiums compresses the future cap rate window for standalone data center assets. Any new entrant paying current construction costs must model exit valuations against a transaction set where institutional buyers have demonstrated willingness to pay infrastructure-fund multiples rather than traditional real estate cap rates.

Across all twelve sources, grid capacity and power access are the explicit or implicit site-selection determinant — not labor markets, permitting timelines, or tax incentives alone.

Evidence from the sources:

• Google / Haskell County: Campus built alongside a purpose-built solar and battery storage plant. Google is not connecting to an existing metro grid; it is co-developing the energy infrastructure. Rural West Texas was selected because of grid capacity and land, not workforce proximity.
• OpenAI / Saline Township, MI: DTE Energy will serve the campus "using existing excess transmission capacity." The siting rationale is stated in plain text — the grid headroom already existed in Washtenaw County. OpenAI's six prior Stargate sites all follow the same pattern: routing to wherever grid headroom exists, not wherever the tech workforce is deepest.
• Amazon / Northern Indiana: NIPSCO Generation LLC — Amazon created a dedicated utility subsidiary to manage power access, pay for new plants and lines, and isolate costs from local ratepayers. This is vertical integration into utility infrastructure.
• Amazon / AWS Government: 1.3 GW of purpose-built government AI infrastructure, with site-selection criteria that add security clearance access (SCIF proximity, federal campus adjacency) as a third axis alongside power and fiber.
• Flexential / Parker, CO: Denver proper has a moratorium on new data centers due to water and power demand concerns. Parker (Douglas County) is the regulatory escape valve. The project uses dry cooling to eliminate water consumption for computing equipment — a direct response to the political opposition that data centers face in power- and water-constrained urban markets.
• QTS / Phoenix: An 80-acre, 210 MW campus near Sky Harbor confirms Phoenix's positioning as a Tier 1 hyperscale market. Phoenix's advantages: land cost, transmission access, tax incentives, and fiber density — the same formula as Tier 2 markets that have captured overflow from Northern Virginia and Silicon Valley.

The consistent pattern: At hyperscale, the selection process works backwards from grid capacity. Hyperscalers identify available transmission headroom first, then negotiate land and permits. The market that can offer 100+ MW of available capacity at an acceptable interconnection timeline wins the campus, not the market that can offer the best labor pool or lowest property tax rate.

This extends and validates the Powered Land and Grid Advantage concept: the binding real estate input is power access and transmission capacity, not square footage or traditional location factors.

The reviewed sources collectively document a secondary-market emergence pattern beyond the traditional hyperscale Tier 1s (Northern Virginia, Silicon Valley, Dallas, Chicago):

Denver Metro (Colorado)

Two distinct data center investments in the Denver metro from a single review window:

• Crusoe Spark Factory, Brighton, CO (March 2026): $200M+ investment in a 352,000 SF manufacturing plant for modular AI factory units. Brighton is northeast of Denver on the I-76/E-470 freight corridor. This is not a conventional data center — it is a facility for producing modular, prefabricated AI compute units (Crusoe Spark) at scale. It adds a new category to Denver's industrial demand base: AI hardware manufacturing.
• Flexential Parker, CO (March 2026): $192M, 249,000 SF, 22.5 MW data center in Parker (Douglas County). Fifth Denver-area facility for Flexential. Denver proper has a moratorium on new data center construction; Parker offers the regulatory alternative. Dry cooling system eliminates water consumption.

Together these two investments suggest Denver is becoming a secondary data center and AI infrastructure manufacturing hub — with the Denver urban core moratorium pushing development into suburban Douglas and Adams counties. The dual pattern (traditional compute facility + modular AI factory manufacturing) suggests the Denver metro is capturing both the operator side and the supply chain side of data center infrastructure.

Ann Arbor / Southeast Michigan

• OpenAI Stargate, Saline Township, MI (October 2025): 1.6M SF campus across three buildings in Saline Township (~25 miles from Ann Arbor). Developed by Related Digital. 2,500+ union construction jobs. DTE Energy excess transmission capacity siting.
• Michigan represents Stargate's seventh-state footprint. Saline Township has no obvious tech-cluster or talent-density rationale comparable to the Texas, Silicon Valley, or Northern Virginia siting logic. The DTE excess-capacity siting logic is the explanation: hyperscale campuses follow available grid headroom, not existing tech geography.
• Implication: Ann Arbor/Southeast Michigan is a secondary hyperscale market created by grid availability rather than organic tech demand. Adjacent industrial, logistics, and workforce housing within the 25-mile DTE service territory may see collateral demand.

Phoenix, Arizona

• QTS Phoenix (March 2026): Blackstone's QTS is pursuing a $510M refinancing backed by PHX2DC2 (36 MW) and PHX2DC3 (45 MW) at the Phoenix 2 campus — 80 acres near Sky Harbor Airport, scaling to 210 MW at full build-out. Fifth building broke ground in 2023. Two hyperscale tenants lease the three-asset collateral pool.
• Phoenix is a confirmed Tier 1 hyperscale destination. The QTS refinancing benchmarks data center debt capacity at approximately $6.3M per MW of leased capacity for hyperscale-tenanted facilities in an established secondary market. Phoenix's formula: lower land and power costs than Northern Virginia/Bay Area, fiber density via Sky Harbor, Arizona tax incentives, and increasingly stable power delivery.
• Aligned Data Centers (the BlackRock acquisition) also originated with Phoenix as one of its two original campus locations — reinforcing Phoenix's role in the hyperscale market hierarchy.

The following is agent interpretation, not source-quoted analysis.

Powered land valuations are re-anchoring at infrastructure-fund multiples

The BlackRock/Aligned and SoftBank/DigitalBridge transactions confirm that capital competing for data center real estate is not using traditional real estate cap-rate logic. When $40B of infrastructure-mandate capital (BlackRock GIP + sovereign wealth) acquires a scaled platform, the implied per-asset valuation floors are set by infrastructure yield curves — typically 100–200 bps tighter than traditional industrial real estate. CRE investors buying data-center-adjacent industrial land (large site, transmission access, fiber proximity) are competing against this capital stack if the land is genuinely hyperscale-ready.

The REIT premium for scale and tenant quality is widening

QTS's $510M refinancing at hyperscale-tenanted occupancy demonstrates that institutional debt markets price data center cash flows differently from commodity industrial cash flows: the indenture trustee / arranger structure, two investment-grade hyperscale tenants, and 80-acre campus scale enable debt terms unavailable to smaller operators. Equinix and Digital Realty remain the public-market benchmarks; private platforms that can assemble 100+ MW of hyperscale-tenanted capacity are the acquisition targets for the next BlackRock/Aligned-style consolidation.

Secondary markets require a power-first underwriting lens

Denver, Ann Arbor, and Phoenix are not capturing data center demand because of tech employment density; they are capturing it because of grid headroom plus land availability plus regulatory openness. Any secondary-market data center underwriting should begin with a utility capacity assessment, not a demographic or employment analysis. The Denver moratorium-and-suburban-spillover pattern is a template: where urban core municipalities restrict data center development due to power/water concerns, suburban markets with utility clarity become the growth vehicle.

Industrial adjacent land near planned hyperscale campuses carries an embedded option premium

The Crusoe/Brighton model — a 352,000 SF AI hardware manufacturing plant near Denver — signals that the data center supply chain is becoming as physically large as the end-user campuses. Sites within the freight and utility corridor of a confirmed or planned hyperscale campus (within 25–50 miles of major campus announcements) now carry an embedded option on supply-chain industrial demand that does not appear in traditional market absorption forecasts.

The government AI infrastructure layer creates a distinct real estate demand segment

Amazon's $50B AWS government AI commitment signals that federal AI infrastructure spending is structurally large and distinct from commercial hyperscaler buildouts. The site-selection criteria differ: SCIF proximity, federal campus adjacency, security clearance access are all inputs alongside power and fiber. Northern Virginia retains its dominant position in this tier (Amazon's $700M Bristow, VA land acquisition was cited in the source). Secondary federal data center markets — Huntsville, Colorado Springs, San Antonio (with Joint Base San Antonio) — carry an incremental demand signal from the government AI infrastructure wave.

Agent synthesis — not source-quoted.

1. Power access is the price-setter, not the location premium. Sites with confirmed transmission capacity and reasonable interconnection timelines command hyperscale consideration regardless of metro tier. Sites without grid headroom are effectively excluded from the market regardless of market quality. The binding constraint question is "how many MW can this substation support?" not "what are the demographics?"
2. Consolidation at the platform level sets the floor for individual-asset pricing. When infrastructure-mandate capital acquires scaled platforms at $40B, individual assets within those platforms are effectively priced at infrastructure yield curves. Data center cap rates in proven hyperscale markets are unlikely to expand toward traditional industrial norms as long as this capital is actively allocating.
3. Regulatory risk is a real secondary-market differentiator. Denver's urban moratorium, water-use politics, and suburban displacement to Parker and Brighton is a pattern that will repeat in other metros. Suburban markets with grid capacity, water availability, and planning-approval clarity will consistently outperform their demographic profile for data center demand. The inverse: urban core data center land in water-constrained or NIMBYed municipalities carries headline-cap-rate risk from regulatory reversal.
4. The government AI layer adds a non-cyclical demand floor. Federal AI infrastructure investment is policy-backed (Trump Administration AI Action Plan, June 2025) and cross-agency in scope (defense, healthcare, energy). This reduces the correlation between data center demand and private-sector AI funding cycles — even if commercial AI capex moderates, government AI infrastructure buildout carries institutional inertia that supports the demand base.
5. The supply chain is as large as the campus. Crusoe's $200M+ AI factory in Brighton is not a data center — it is a factory for making modular data centers. As the industry matures, the full real estate footprint includes manufacturing, logistics, and assembly facilities alongside the operating campuses. Industrial land within freight corridors adjacent to hyperscale clusters carries collateral demand that grows with the cluster.

• OpenAI Stargate Texas site specifics: The sources confirm Texas as an early Stargate state but do not identify specific Texas site locations, MW commitments, or construction timelines. A dedicated Texas Stargate source would substantially improve the Texas-specific capital accounting.
• Amazon AWS government AI facility locations: The $50B commitment and 1.3 GW of capacity addition are confirmed, but facility locations beyond Bristow, VA are not disclosed in the reviewed sources. The geographic distribution of government AI infrastructure buildout is unclear.
• Microsoft Castroville beyond SAT 82: Data Center Dynamics reported additional Microsoft Castroville projects beyond SAT 82 but details are not in the reviewed source. The full scale of Microsoft's San Antonio exurban footprint is not confirmable from current sources.
• Aligned Data Centers post-acquisition operating strategy: The BlackRock/Aligned deal closed (expected H1 2026) but the post-acquisition capital deployment and campus expansion plan are not yet published. The 5+ GW "operational and planned" figure includes both — split is not disclosed.
• MacroValor / Mount Hydrogen credibility: The 3 GW Austin hydrogen-powered campus announcement is filed as aspirational. MacroValor's natural hydrogen network is unverified; no engineering or site permitting documentation is available. Status: watch-list only.
• OpenAI Stargate Ann Arbor capacity split: The 4.5 GW partnership with Oracle referenced in the source combines Michigan with prior announced sites; the specific capacity allocation to the Saline Township campus is not disclosed.

• Source: Amazon Plans to Invest $15B to Build New Data Center Campuses in Northern Indiana
• Source: Amazon to Invest Up to $50B to Expand AI Infrastructure for U.S. Government
• Source: Google to Invest $40B in Texas Data Centers Through 2027
• Source: Reports: Microsoft Plans $400M Data Center in Castroville, Texas
• Source: OpenAI Plans to Build Data Center Near Ann Arbor, Invest $450B Over Next Three Years
• Source: SoftBank Agrees to Acquire DigitalBridge in $4B Deal
• Source: BlackRock-Led Consortium Agrees to Acquire Aligned Data Centers for $40B
• Source: Crusoe Building $200M Denver-Area AI Factory
• Source: Flexential Building 249K SF Denver-Area Data Center in Parker, CO — $192M, 22.5 MW
• Source: Edged U.S. to Open New Data Center Facility at Irving Campus
• Source: QTS Tapping Two Phoenix Data Centers for $510M Refinancing
• Source: MacroValor, Favis Will Build Hydrogen-Powered AI Mega Campus

• Texas Digital Infrastructure Corridors
• Digital Infrastructure Real Estate
• AI Office Demand Engine 2026
• Powered Land and Grid Advantage
• Alliance and North Fort Worth
• Plano Richardson Telecom Corridor
• East Austin Tesla and Airport Corridor
• Equinix
• REIT Landscape
• Conviction Theme Investing
• Analyses Hub
• United States