National Digital Infrastructure Capital Deployment 2026

Editorial and financing review: October 1, 2026. Operating figures retain their stated provider, geography and reporting period. A later review date does not make an earlier observation current.

Key takeaway

Digital-infrastructure investment should follow deliverable power, funded customer demand and a feasible operating asset. Existing capacity, contracted development, powered land and speculative queue positions carry different risks.

Recent state-policy changes and financing pressure make delivery timing more consequential. Require utility and approval milestones, equipment availability, customer credit and cash through energization; announced megawatts and AI funding do not establish income or a financeable site.

Financing assumptions

The Fed raised its target to 3.75%–4.00% on September 16. The September 30 Treasury close was 5.29% for the 10-year and 5.64% for the 30-year. Source: Federal Reserve FOMC Statement September 16 2026; Source: U.S. Treasury Daily Par Yield Curve September 2026.

Require tenant commitments, building function and utility delivery before borrowing against future income. Test longer lease-up and holding periods, higher debt cost and a wider exit cap rate; demand announcements do not establish funded cash flow. See U.S. Commercial Real Estate Outlook Q4 2026 for the common macro scenarios.

Evidence carried forward

The following source-period observations and earlier qualifications remain part of the research record. Current allocation and financing conclusions are stated above.

Hut 8's Beacon Point adds a contracted Gulf Coast campus to the development map: 1,000 MW of sponsor-reported utility capacity, 704 MW of contracted IT capacity, and a Q3 2027 initial data-hall target. The company estimates about $17 billion for the first two phases, including compute infrastructure. Keep those campus-level figures separate from Connect CRE's $399-million / 657,130-SF facility estimate; the captures do not reconcile their scopes, and neither figure is a metro market statistic. See Beacon Point Data Center Campus and Source: Hut 8 Beacon Point Data Center Campus 2026.

State Policy and Delivery Risk (September 2026)

The September 21 California and Texas actions add distinct policy gates to the national deployment map. California enacted seven bills described by the Governor's office as covering large-load electricity, grid costs, clean-energy supply, water information and supply, land-use review, and community information. Texas issued a gubernatorial directive pausing TCEQ data-center permits until ERCOT completes its review, with other state approvals tied to information sought by PUCT, ERCOT, and TWDB. Massachusetts adds permitting and community-benefits conditions with defined scope. The September 8 signed EO 658 qualifies the release and Connect CRE local-approval framing: Section 1 conditions permits and authorizations for projects exceeding 25 MW peak electricity demand on Framework conformance and an OEJE-aligned agreement with key stakeholders. Section 13 includes post-effective-date construction or expansion, demand and expansion language, and possible jointly approved threshold exemptions. Energy, water and reporting directions require implementation; the permitting-agency NDA restriction retains its legal exception. This is a statewide process requirement, not a project outcome or evidence of commercial capacity. These records establish rules and review conditions, not project outcomes: current evidence does not count affected permits, quantify schedule delay, identify canceled campuses, or show resulting capacity moving to other markets. Do not translate statewide policy into a local market or project forecast without later agency and site-specific records. See Source: Texas and California Data Center Policy Actions September 2026 and Source: Massachusetts Executive Order 658 Local Approval Requirements 2026, with clause scope in Source: Massachusetts Executive Order 658 Signed Order 2026.

Loudoun County's September 15 vote adds a distinct local mechanism: it advanced consideration of a resolution to pause the Board's final action on legislative data-center and substation applications, potentially for up to 12 months. The county said applications could still be filed and proceed through Planning Commission review, with administrative applications outside the proposal. As of September 24 the resolution was pending, with October 20 set for consideration. It is an entitlement-timing watch item in Northern Virginia, not evidence that approvals or project deliveries have stopped. See Source: Loudoun County Data Center and Substation Application Pause Proposal 2026.

Maryland's September 23 executive order adds state review for projects at least 25 MW when seeking state action; Governor Moore's separate commitment to seek repeal of the data-center sales-tax exemption still requires legislation. Virginia's September 18 Executive Order 22 immediately limits certain discretionary site-readiness and expedited-review assistance for new projects at least 25 MW, while the framework's proposed local-approval rule for projects above 25 MW is aimed at the 2027 session. In Prince William County, trade reports describe a September 22 vote shrinking the by-right overlay, but differ on its final acreage and the captured county page does not supply the adopted ordinance. Treat the state thresholds, tax proposal, and local boundary as separate diligence items; none establishes canceled or delayed capacity. See Source: Mid-Atlantic Data Center Rules September 2026.

Phoenix adds a property-level entitlement case to the deployment map. City Council adopted LKY's binding-waiver item under Ordinance S-52436 in November 2025; September 2026 reporting then described GI Partners' $14.72 million purchase of a 5.43-acre Phoenix data-center site from LKY. The reports appear related through seller and intended use, but do not identify the traded parcel by APN. The episode supports site-by-site review of zoning relief and utility delivery; it does not establish a generally permissive rule, completed power expansion, or delivered capacity. See Source: GI Partners South Phoenix Data Center Site 2026.

Federal Tax Policy and OZ 2.0 (September 2026)

Public Law 119-21 made Opportunity Zones ongoing through ten-year designation rounds and added rural investment incentives; the first new round is scheduled to take effect January 1, 2027. IRS guidance describes nominations, while final Treasury certification remains a separate step. Hawley's September 17 bill draft and Wyden's August 6 staff white paper propose different limits on data-center tax treatment. The Hawley draft directly changes the OZ business definition; the sponsor release describes a broader exclusion. Wyden's paper is a discussion proposal, not legislation. Neither proposal changes enacted eligibility in the captured record. Treat OZ benefits as cohort- and tract-specific diligence, not as a general data-center subsidy or a realized project benefit. See Source: Data Centers and Opportunity Zone Tax Policy 2026 and Opportunity Zones and Qualified Opportunity Funds.

Supporting detail

JLL's Source: JLL Data Centers in Space 2026 source keeps that frontier item in the correct lane. The capital-deployment implication is workload segmentation, not a new investable property type: asynchronous AI training, batch processing, simulation, and in-orbit data processing may eventually move toward orbital infrastructure, while latency-sensitive compute still needs terrestrial sites near users and networks. Track launch cost, hyperscaler pilots, debris management, and grid bottlenecks as strategic milestones; do not count orbital plans as current data-center supply.

JLL's Source: JLL The Future of Quantum Real Estate 2026 adds a second frontier lane. Quantum investment may eventually create specialized facility demand around quantum ecosystems, cloud access, cryogenic cooling, electromagnetic shielding, and hybrid quantum-classical rooms. For this capital map, keep it in watchlist status until quantum utility, pilot deployments, tenant commitments, and facility specifications are proven; it is not current colocation inventory or ordinary data-center absorption.

The Santa Barbara / Goleta case provides a concrete research-commercialization node without establishing quantum-specific commercial absorption. UCSB leased a 105,000-SF research and development facility at 71 S. Los Carneros Drive in Goleta for university, industry, and startup collaboration; the OASIS facility lists quantum systems among several application areas. This supports research-campus adjacency as a location to monitor, not a quantum-only facility or market-performance claim. Commercial Observer’s accompanying property transaction prices, quantum facility counts, and broader regional ranking remain outside this synthesis pending primary-record review. See Source: Commercial Observer Quantum Coast Santa Barbara 2026.

The June 15 policy / mega-campus batch adds a useful counterweight to capex headlines. Local and state data-center bans show that community acceptance can remove capacity from the pipeline, while the reported OpenAI federal-land discussion shows how AI users may seek public-land, power, and federal-site solutions when ordinary sites cannot scale. Keep both as watchlist and underwriting-context evidence until official policy texts, site-control documents, utility records, and executed campus commitments are preserved. See Source: Data Center Local Ban Wave 2026 and Source: OpenAI Federal Land Data Center Campus 2026.

Google's Michigan announcement adds an energy-procurement case to the hyperscaler lane without adding a financeable campus record. Google says 2.7 GW of new solar, advanced storage, and demand-flexibility resources will serve a proposed DTE-territory operation, and it pairs that commitment with ratepayer-protection language, a $10M Energy Impact Fund, and a water-assessment commitment. The Van Buren Township site remained under evaluation. Capital-deployment implication: utility structure and resource additions can precede final site disclosure, but the resource portfolio must not be counted as IT load, completed generation, or delivered data-center capacity. See Michigan Data Centers and Powered Land and Source: Google Michigan Clean Energy and Reliability Announcement 2026.

The subagent data-center article batch adds a cross-market capital-deployment trio: Prime's Metro Phoenix groundbreaking belongs in the growth-market campus lane, Newmark's $975M Northern Virginia financing belongs in the mature-core debt lane, and Google's New Florence, Missouri investment belongs in the noncore infrastructure-expansion lane. Together they reinforce that 2026 data-center capital is not one market; it is a set of power, utility, financing, and local-acceptance gates by geography. See Source: Prime Data Centers Metro Phoenix Campus Groundbreaking 2026, Source: Newmark 975M Northern Virginia Data Center Financing 2026, and Source: Google New Florence Missouri Infrastructure Investment 2026.

Helix Digital Infrastructure launched with more than $10B committed by its founding investors, according to the June 15 announcement. KKR’s September 28 follow-up says Samsung’s separate $1B commitment builds on that launch capital, while Samsung says six affiliates are participating and identifies $500M from Samsung Electronics. Together, the issuer-reported commitments now exceed $11B; this is platform capital formation, not deployed capital, asset-level investment, or delivered MW. The Samsung release also describes potential collaboration in advanced technology, construction, storage, and cooling, but no project awards or allocations are disclosed. See Source: KKR NVIDIA Kuwait Helix AI Infrastructure Platform 2026 and Source: Samsung $1B Commitment to Helix Digital Infrastructure 2026. The same infrastructure capital map includes DigitalBridge / ArcLight platform formation, Switch’s nearly $10B credit facilities tied to a contracted pipeline, and local approval risk reflected in the NDA / public-opposition source; these remain separate from delivered inventory. See Source: DigitalBridge ArcLight Power Infrastructure Acquisition 2026, Source: Switch Credit Facilities Nearly 10B 2026, and Source: Data Center NDA Public Opposition 2026.

Broadcom / Apollo / Blackstone adds a sixth lane inside capital formation: compute-platform financing tied to hardware, networking, and frontier-lab capacity commitments. The reported $35B initial tranche and 20+ GW design target should be treated as source-scoped platform claims, but the direction is durable: private credit and insurance capital are moving directly into AI compute infrastructure, not only into real-estate shells. See Source: Broadcom Apollo Blackstone AI Infrastructure Platform 2026.

A September 30 Bisnow report citing the Financial Times describes preliminary Nvidia discussions with insurers about covering lender losses on GPU-backed loans to neoclouds. The proposed trigger is borrower default plus insufficient chip resale proceeds; the report says no agreements are in place. This extends the financing map from project and platform capital toward possible GPU residual-value risk transfer, but no insurance capacity, executed loan, or property financing can be counted from the report. Source: Nvidia Insurer Discussions for GPU-Backed Neocloud Loans 2026

The underwriting mistake is collapsing all five into one giant capex number. The better move is to ask which lane actually creates a real estate opening that outside capital can access.

The May 2026 podcast synthesis pass adds a softer but useful signal around the same distinction. TreppWire 396 pairs lending reset with AI and data-center growth, while TreppWire 392 frames data centers as a gigawatt-scale capital topic. Separately, AI-workflow episodes across family-office, BOMA, and Best Ever feeds show that enterprise AI adoption is broadening as an operating theme even where it does not yet create a direct office or data-center lease claim. Use these notes to track market attention and candidate source leads, not to size megawatts, power costs, or data-center rent.

The transcript / rich-episode pass gives the BOMA Suburban Chicago AI episode a clearer role: it points to AI in building operations, predictive maintenance, energy optimization, and smart-building technology. That belongs in the operating-infrastructure lane, not the hyperscaler-capex lane. It can support diligence questions about controls, sensors, vendor systems, and operating-data readiness, but it should not be used to infer data-center capacity, utility load, or AI-office absorption.

Source: Colliers Data Center Insights Series Q3 2026 adds an execution-and-acceptance layer beneath the data-center demand thesis. Colliers' public page frames power availability, supply-chain and labor constraints, larger capital commitments, and community acceptance as determinants of whether capacity can actually be delivered. Underwrite campus scale through power and schedule, procurement, labor, financing, phasing, sponsor liquidity, tenant concentration, and entitlement stress—not through AI demand or scarcity alone; the full interview was not captured.

JLL's quantum-computing real-estate discussion adds an emerging specialized-infrastructure lane: today, quantum space and supporting infrastructure are concentrated on university campuses, while private-sector demand is expected to expand over the next decade. The allocation implication is a watchlist for research-campus adjacency, specialized cooling and utility requirements, and future private facilities—not a current demand or valuation forecast.

An independent Carlstadt, New Jersey lease report adds a building-vintage and location cross-check: a confidential enterprise tenant reportedly leased the full 249,640 SF building, originally built in 2008 for a bank and subsequently upgraded. The example supports a narrow point in the digital-infrastructure screen—time-sensitive users may value immediately usable, upgraded capacity in an established fiber market over a brand-new shell—without proving a rent, valuation, or marketwide absorption benchmark.

Source: ULI/PwC Emerging Trends in Real Estate United States and Canada 2026 reinforces the power-first constraint: the public summary describes national data-center vacancy below 2%, broad preleasing before completion, and supply bottlenecks. Treat reliable energy access and delivery timing as gates alongside fiber, zoning, utility structure, customer commitment, and construction capital; the summary is not a capacity or valuation forecast.

Source: Green Street Data Centers and the AI Surge 2026 adds an early-2026 data-center demand framework: AI leasing, power-constrained primary markets, secondary-market expansion, and supply/development constraints are linked. Use it to stress power delivery, site selection, and schedule risk alongside demand; it does not provide a numeric capacity, rent, or valuation assumption.

Source: ENR Construction Economics Second Quarterly Report 2026 supplies a construction-cost and schedule control beneath the AI-demand narrative. Through April, utility starts were up 79% and office/data-center starts were up 61%, while ENR notes a median 17-month period from project planning to groundbreaking and continuing pressure from specialized labor, power availability, materials, and community resistance. Underwrite the data-center pipeline through deliverable power, procurement escalation, and entitlement timing; announced demand is not delivered capacity.

Source: Cushman & Wakefield Data Center Industrial Demand, Jobs and Local Revenue 2026 adds a quantified industrial spillover lens: data-center-related tenants in six analyzed markets leased 40 million square feet and represented 14.4% of new industrial leasing in 2025, while C&W estimates nearly 1,300 local jobs per 100 MW of new data-center development. Treat these as sample-specific economic-impact estimates, not a universal multiplier. For site underwriting, separate leased industrial demand from delivered compute capacity and verify power, tenant, timing, and local input-output assumptions.

Source: Green Street Infrastructure Signals What the Research Is Revealing 2026 adds a coverage-boundary check for data centers, towers, and transport. Green Street's public webinar page frames infrastructure research around supply/demand, fundamentals, valuation, macro, capital markets, and public/private-market perspectives, but provides no numeric capacity, rent, pricing, or return data. Use it to route diligence questions across infrastructure sectors, not to alter an asset-level underwriting case.

Green Street's H1 2026 Infrastructure and Project Finance League Table preview adds an aggregate capital-intensity check: global infrastructure deal value rose 18% year over year to roughly $1.3T while deal count fell 8.9% to 2,200, and project finance rose 53.6% to $682.7B as primary financings jumped 64.1%. The full rankings remain gated, so use this as concentration and capital-flow context—not as proof of project-level debt availability or data-center delivery. See Source: Green Street H1 2026 Infrastructure and Project Finance League Table.

Source: RCLCO Data Center Pipeline Boom or Bubble adds a capital-absorption stress case: its base scenario values the U.S. data-center pipeline at roughly $1.8T, with 80–134 GW of modeled new supply and an estimated $0.9T–$1.3T of equity at 25%–50% LTV. These are scenario assumptions, not entitled supply; use them to test power delivery, tenant commitments, construction timing, residual value, and exit-cap risk.

CBRE Pacific Capacity Scenario

CBRE's Pacific outlook forecasts Australian live data-center capacity rising from 1.3 GW to 1.8 GW by 2028 while demand could exceed supply by up to 1.2 GW. Use this as a regional capacity scenario, not site entitlement or utility evidence: interconnection, land, water, transmission, and delivery milestones remain required before capital deployment.

Question

What does the 2026 digital-infrastructure capital wave actually look like, and where does it create real estate opportunity versus narrative noise?

Method

Synthesized the current source set across hyperscaler announcements, Stargate program expansion, platform acquisitions, financed metro and exurban projects, and one watch-list mega-campus announcement. Read against Texas AI and Industrial Infrastructure Opportunity Map, Digital Infrastructure Real Estate, and Powered Land and Grid Advantage so this page could stay a national capital map rather than a corridor page.

2026 Reset

The useful 2026 point is not that AI infrastructure is large. It is that the money is arriving through different lanes that should not be underwritten as if they were the same thing. This is the first-order AI real estate map: compute, power, data centers, powered land, utility partnerships, and supply-chain spillover. Office demand is a narrower second-order read-through handled in AI Corporate Real Estate Footprint 2026 and AI Office Demand Engine 2026.

CBRE Data-Center Market Constraint Overlay

Source: RCLCO Data Center Pipeline Boom or Bubble 2026 adds a scenario-based capital-capacity warning beneath the broker operating data. Its 80-to-134-GW range, $1T-to-$3T stabilized-value range, and $0.9T-to-$1.3T base-case equity calculation flow from RCLCO's stated assumptions; they are not forecasts or underwriting benchmarks. The durable allocation rule is to distinguish pipeline headlines from projects that can demonstrate power, stabilization, tenant, construction-capital, and exit-sale paths under a delayed or more competitive delivery case.

Brookings' September 2026 conference draft adds a broader financing scenario: $10.3T of projected AI infrastructure investment during 2025–2032, averaging 3.63% of GDP. Its scope includes data-center buildings, power, networking, and specialized IT equipment; the paper allocates roughly two-thirds of a representative campus's cost to IT. Do not compare the total directly with RCLCO's real-estate pipeline valuation or treat it as CRE construction spend. Brookings models only 182.8 GW operating by 2032 from a 509 GW announced pipeline and says it is premature to call the exposure systemic; the practical diligence implication is better tracking of leases, JVs, private credit, project debt, tenant concentration, and asset-level leverage. Source: Brookings Financing the AI Buildout 2026

Source: CBRE North America Data Center Trends H1 2025 adds the first-half baseline for the market-wide CBRE constraint layer. The applied rows (market_observations.id=44075-44083) show 8,155 MW of primary-market supply, 1.6% vacancy, 5,242.5 MW under construction, 74.3% of under-construction capacity already committed, 2.5% pricing growth for 250-to-500-kW requirements, up to 19% pricing growth for larger 10 MW-plus deployments, and 538.6 MW of H1 net absorption in Northern Virginia. That makes H1 a power-first scarcity checkpoint rather than a loose demand headline.

Source: CBRE North America Data Center Trends H2 2025 adds a market-wide operating constraint layer underneath the project-announcement map. The dedicated overview import now applies the previously review-only H2 2025 rows (market_observations.id=43890-43908): 1.4% primary-market vacancy, 9,432 MW of supply, 2,497.6 MW of 2025 absorption, 5,994.4 MW under construction, $3B of operational data-center investment volume, $11.2B of data-center SASB CMBS issuance, and CoreWeave / Stargate financing context.

The underwriting read-through is narrower than "data centers are hot." In CBRE's H2 2025 framing, rent growth, site pricing, and development timing are being driven by limited powered land, booked grid capacity, and the need for power access within roughly 18 to 36 months. That reinforces this page's existing lane discipline: power queue position, utility structure, fiber, zoning / community acceptance, and customer commitment decide whether a data-center or powered-land thesis is underwritable.

Source: CBRE Hillsboro OR Data Center Market H2 2025 adds a smaller-market version of the same deployment gate. Hillsboro has no 5 MW-plus contiguous facility availability in CBRE's H2 2025 text, while PGE's planned $6.5 billion grid-modernization investment and Oregon's 20 MW POWER Act data-center customer class show how utility capex and policy design can become underwriting inputs. Treat Hillsboro as a scarcity / grid-policy node, not just another generic industrial market.

Source: CBRE Hillsboro Data Center Market H1 2025 adds the prior planned-capacity base for that same Oregon node. NTT's additional 216 MW expansion toward 354 MW of total planned capacity (market_observations.id=44064-44065) shows why Hillsboro remains a real capacity-growth market, while CBRE's sub-500 kW accommodation / larger-requirement scarcity language keeps the underwriting test focused on deliverable operator vacancy and utility-cost allocation.

Source: CBRE New York Tri-State Data Center Market H2 2025 adds the low-latency mature-metro version. New York Tri-State still has financial-services demand and New Jersey powered-land interest, but CBRE says 5 MW-plus contiguous-space requirements remain hard to satisfy, record-high pricing reflects limited supply, and PJM pricing / high power costs / limited tax incentives push some occupiers toward adjacent markets. Treat this as evidence that latency demand does not eliminate the power-cost and incentive gate.

Source: CBRE New York Tri-State Data Center Market H1 2025 adds the prior-period scarcity checkpoint. In H1, CBRE said no capacity above 5 MW was currently available, AI-related occupiers were preleasing in 5-to-10-MW increments where available, and Equinix's NY3 project was coming to market as the firm's 10th Tri-State data center (market_observations.id=44074). Treat New York Tri-State as a legacy-connectivity and operator-campus market where demand exists, but local capacity converts only when power, contiguous blocks, and facility-level operator inventory clear.

Source: CBRE Northern Virginia Data Center Market H2 2025 adds the mature-core benchmark. Northern Virginia delivered more than 1 GW of capacity in 2025 and still had 0.5% colocation vacancy in H2, with 15-to-20-MW contiguous requirements hard to secure and most 2026 capacity already committed. Treat it as proof that the highest-conviction market still has to be underwritten through power-delivery timing, preleasing, and utility process rather than demand alone.

Source: CBRE Northern Virginia Data Center Market H1 2025 adds the earlier checkpoint for the same corridor. H1 vacancy was already 0.72%, EdgeCore had begun a 114 MW Loudoun facility, and AWS / BlackChamber land-control rows added 97 and 65 acres in Leesburg and Prince William (market_observations.id=44066-44069). The H1 lesson is that even the strongest U.S. demand node was already a preleasing, entitlement, Dominion-batching, and expansion-corridor problem before H2 tightened further.

Source: CBRE Silicon Valley Data Center Market H2 2025 adds the premium innovation-cluster counterpoint. Silicon Valley can still command pricing above other North American markets for 10 MW-plus contiguous colocation requirements, but CBRE says new construction remains limited by power challenges and Silicon Valley Power's 4% rate increase took effect in January 2026. Treat it as a reminder that proximity to AI labs does not remove the power-cost and capacity-delivery gate.

Source: CBRE Silicon Valley Data Center Market H1 2025 adds the prior-period operating base for the same Bay Area node. H1 vacancy dropped to 4.5%, CBRE said over 50 MW was immediately available or under construction, and the clean activity rows include a 9 MW AI-company prelease plus 16 MW of Vantage commissioned preleased capacity (market_observations.id=44070-44073). Treat this as Silicon Valley's innovation-cluster scarcity lane: valuable, but capped by power availability and by sites reverting to industrial use when data-center power cannot clear.

Source: CBRE Minneapolis Data Center Market H1 2025 adds a secondary-market / enterprise-driven version of the gate. Minneapolis has limited large multitenant colocation depth and a project pipeline that should be treated as proposed context, not delivered capacity, while the clean Eagan build-to-suit datapoint is only 61,000 SF / 5 MW. Treat it as evidence that noncore data-center markets can matter where power, renewable access, and site execution line up, but only as project-specific powered-land underwriting.

Source: CBRE Houston Data Center Market H1 2025 adds the energy-market counterpoint. Houston's H1 2025 data-center vacancy was 19.7% after recent single-operator delivery, even as CBRE cited oil-and-gas demand, behind-the-meter power studies, and 15 MW of scheduled capacity by year-end 2026. Treat Houston as a power / energy-user / legacy-space underwriting lane, not as automatic evidence of tight colocation scarcity.

Source: CBRE Denver Data Center Market H1 2025 adds the small selective-market version. Denver had 20.1 MW across 27 online facilities in H1 2025, and CBRE says only two colocation facilities offered large amounts of power and space. Treat Denver as a suburban powered-land and retrofit / utility-process market, not as a mature-scale hyperscale node.

Source: CBRE Austin-San Antonio Data Center Market H1 2025 adds the Central Texas tight-market version. Austin-San Antonio had 2.3% vacancy, 97% of under-construction capacity preleased, and 462.0 MW under construction in H1 2025, with CBRE pointing to northeast Austin suburbs served by Oncor and increased activity south of Austin in Lower Colorado River Authority territory. Treat the market as a service-territory and power-delivery lane, not as a conventional CBSA comparison.

Source: CBRE Atlanta Data Center Market H1 2025 adds the utility-screened hyperscale-demand version. Georgia Power had secured over 8 GW of new generation capacity scheduled after 2030, while a major hyperscale user had preleased over 300 MW and planned another 600 MW by year-end. Treat Atlanta as a power-procurement and customer-commitment market where utility collateral, zoning confirmation, transmission, and permitting decide whether demand converts into deliverable capacity.

Source: CBRE Chicago Data Center Market H1 2025 adds the mature-market / utility-credit version before the H2 update. Prime Data Centers leased 72 MW to an AI-related tenant, while two Yorkville projects were seeking approval for 38 buildings and 3+ GW of power. Treat Chicago's H1 signal as evidence of AI leasing and approval-stage power demand, but keep the underwriting tied to ComEd timing, letters of credit, and exurban power-site execution.

Source: CBRE Dallas-Ft. Worth Data Center Market H1 2025 adds the Texas scale-market setup before DFW's H2 tightening. CBRE's preserved H1 local profile reports 869.5 MW of existing inventory, 425.1 MW under construction, and 78% of that pipeline preleased, with Oncor transmission planning, rising development-site pricing, unprecedented utility requests, and behind-the-meter power studies already visible. Source: CBRE North America Data Center Trends H1 2025 Market Profiles separately says DFW was expected to double in size by 2026 and that 89% of under-construction space was preleased, but that landing-page snippet should stay source-scoped until reconciled to the dedicated profile denominator. Treat DFW as a scale-and-power-queue market, not as ordinary industrial land demand.

Source: CBRE Phoenix Data Center Market H1 2025 adds the early Phoenix utility-study case before its H2 absorption step-up. CBRE reported just over 75 MW of H1 2025 space-and-power absorption, several colocation projects scheduled for completion over the next 12 months, Phoenix / Mesa development-restriction pressure, and SRP study work for 26 citywide projects. Treat Phoenix as a hyperscale-demand market where municipal restrictions and utility-study queues are part of underwriting, not as generic Sun Belt land demand.

Source: CBRE Seattle Data Center Market H1 2025 adds the dense-power / fast-occupancy urban-network version. CBRE reports AI, enterprise, and hyperscaler demand, a Washington state sales and use tax exemption tailwind, liquid-cooling adoption, generator options for fast-track solutions, Digital Realty SEA10 demand of up to 1 MW, and Sabey's scheduled 6 MW SDC Building 5 expansion for Q1 2026. Treat Seattle as a network and power-density market, while keeping Central Washington separate as the larger powered-land basin.

Source: CBRE Central Washington Data Center Market H1 2025 adds the rural powered-land / private-power version of the same Washington market. CBRE reports limited multi-tenant colocation, hyperscaler-owned and single-tenant leased facilities as the development drivers, PUD transmission-line delays, developer- and end-user-funded substations, alternative power, microgrids, and interim generator solutions. The clean planned-campus rows are a Tri-Cities hyperscaler plan for $4.8B, 16 data centers, and 215,000 SF each; treat them as planned project scale, not delivered capacity.

Source: CBRE Charlotte-Raleigh Data Center Market H1 2025 adds the Carolinas power-commitment and service-territory version. The clean rows are 1 MW delivered at a Raleigh cell tower site, a 1,000,000 SF Madison conversion building, a 1,600,000 SF Gaffney announced campus, and a $10B Hamlet / AWS announced buildout (market_observations.id=44052-44055). Treat the source as evidence that Duke Energy commitments, capital contributions, mid-2027 delivery timing, and perimeter service-territory opportunities decide which Carolinas sites become financeable capacity.

Source: CBRE Southern California Data Center Market H1 2025 adds the connectivity-rich infill colocation version. The clean structured row is One Wilshire marketing over 10 MW of available shell capacity (market_observations.id=44056), while CBRE's pricing-increase, AI net-absorption, and CoreSite / Goodman Vernon-development claims stay in source-note context. Treat Southern California as a carrier-hotel and infill power-cost market, not a generic hyperscale land-banking market.

Source: CBRE Montreal Data Center Market H1 2025 adds the hydropower-rich but allocation-constrained Canada example. The clean rows are QScale's 14 MW Levis delivery and Enovum's 3.4 MW expansions (market_observations.id=44057-44058), while Hydro-Quebec service-agreement restrictions, sub-5 MW threshold timing, AI/HPC demand, liquid cooling, and scarcity of contiguous blocks above 1 MW stay in source-note context. Treat Montreal as a power-allocation and high-density deployment market, not as unconstrained cheap-power capacity.

Source: CBRE Toronto Data Center Market H1 2025 adds the GTA power-and-security version of the Canada AI-demand screen. The clean rows are CoreWeave's approximate 52 MW January preleasing, Microsoft's 48 MW campus-expansion critical load, Hive's 7.2 MW conditional Etobicoke acquisition capacity, and the 50-400 MW planned greenfield range bounds for 2027-2028 (market_observations.id=44059-44063). Treat Toronto as a market where AI workloads are moving beyond strict latency toward power availability, site security, utility-study execution, and built-out colocation scarcity.

Source: JLL Liquid Cooling Enters the Mainstream in Data Centers adds a design-conversion checkpoint to the AI facility map. JLL's updated article reports operator examples above 70 kW per rack and says immersion cooling may reduce floor-to-floor height needs while increasing structural loading to at least 20 kPa. This can change the usable fit-out envelope, but the benefit is inseparable from electrical, plumbing, slab, and live-service constraints. No universal operating-cost savings or conversion capacity follows from the article.

JLL Global Capacity and Cost Overlay

Source: JLL 2026 Global Data Center Outlook adds a global sizing and cost layer beneath the project map. JLL forecasts 97 GW of new global data-center capacity between 2026 and 2030, potentially taking global capacity to 200 GW by 2030, with a 14% base-case global supply CAGR. The source also frames the Americas as the fastest-growing region at 17% supply CAGR, with 2030 average 1-20 MW lease-rate forecasts of $222/kW/month in the Americas, $207/kW/month in EMEA, and $179/kW/month in APAC.

That does not turn every powered-land headline into underwritable capacity. JLL's delivery-risk rows are the useful discipline: 57% of projects experienced construction delays of three months or more in 2025, global equipment lead times averaged 33 weeks, and 2026 shell-and-core construction costs are forecast at $11.3M per MW before land and active IT equipment. The right read-through is that AI infrastructure is large enough to be durable, but power, equipment, liquid cooling, construction execution, local approvals, and tenant/capital-stack proof still decide which sites convert into investable real estate.

Source: Cushman & Wakefield Asia Pacific Data Centre Construction Cost Guide 2026 adds an APAC-specific cost-inflation check to that global frame. C&W reports average APAC data-centre construction costs rose 10% year-over-year in 2025 and identifies Japan, Singapore, South Korea, Australia, and Hong Kong as the five most expensive mid-spec build markets. For the capital-deployment map, this is a reminder to separate demand durability from project feasibility: high-cost APAC markets need stronger sponsor basis, contingency, power-cost assumptions, and lending reserves before growth headlines become investable delivery.

Source: Cushman & Wakefield Asia Pacific Data Centre Investment Landscape 2025 adds the investor-screen complement. The page frames APAC data centres as an institutional asset class with yield-stability and inflation-hedging appeal, but it also says the report evaluates 14 key markets through population per megawatt, rent revenue, CapEx requirements, and yield-on-cost. For this capital map, that means APAC capital should be sorted by demand density, revenue capture, capex intensity, and development spread rather than by country growth narrative alone.

Source: Cushman & Wakefield APAC Data Centre Update H2 2025 adds the current APAC delivery scale beneath that investor screen. The public page reports 13,763 MW of operational capacity, a 19,371 MW development pipeline, 3,677 MW under construction, and 15,694 MW in planning as of H2 2025, with Malaysia and India accounting for 58% of new operational capacity. For capital deployment, that shifts APAC from only a growth narrative to an execution market, while still requiring power, regulation, geopolitical, operator, and project-level diligence before planned capacity is capitalized.

Source: Cushman & Wakefield APAC Data Centre Update H1 2026 supplies the H1 2026 refresh: the visible public summary reports 26,455 MW of pipeline, 4,764 MW under construction, 21,691 MW planned, and 1,372 MW delivered in the first half, while regional vacancy declined to 10.3%. The capital-deployment implication is that demand is absorbing new capacity even as scale increases, but the investable subset is still constrained by power delivery, infrastructure readiness, regulatory certainty, market maturity, preleasing, and execution.

Source: Cushman & Wakefield Modular Data Centres APAC July 2026 adds a delivery-model control to the APAC capacity map. C&W's public preview describes prefabricated power, cooling, networking, and IT components as a way to deploy faster, scale incrementally, and replicate facilities across locations for AI, edge, 5G, and cloud workloads. The capital implication is conditional: modular delivery may improve schedule and phasing, but project scale, logistics, site conditions, operational requirements, power, and cooling still determine whether a proposed facility is real and financeable.

Source: Cushman & Wakefield Turning Land into Value Data Centres June 2026 adds the land-value gate beneath that delivery model. C&W's APAC preview ties infrastructure-ready land value to utility access, power capacity, fiber, water, zoning, and grid-delivery timing. Treat “powered land premium” as an execution-dependent conclusion requiring technical diligence and an executable energization path, not as a generic land valuation uplift.

Source: Cushman & Wakefield H2 2025 EMEA Data Centre Market Update adds an EMEA delivery-capacity screen. C&W reports EMEA pipeline capacity approaching 15 GW and a 33-market maturity index, but its more important capital-allocation point is that power availability, grid access, regulatory complexity, sustainability requirements, scalable land, and faster grid connection timelines determine where that pipeline can become real capacity. That keeps EMEA capital deployment in the same discipline as North America and APAC: demand is necessary, but executable power and approvals decide investability.

Source: Cushman & Wakefield Artificial Intelligence and the Next Chapter for Australian Commercial Real Estate adds an economy-wide AI scenario overlay to the infrastructure map. C&W models four paths across 19 industries and Australian cities: logistics and industrial retain the highest floor, while office outcomes vary sharply with labor substitution, productivity, development freezes, and quality. Use this to widen city-, sector-, and asset-grade sensitivities around AI-linked capital deployment; the modeled scenarios do not replace observed leasing, employment, power, or capacity evidence.

Source: JLL North America Data Center Report Year-end 2025 narrows that global frame to the North American delivery map. JLL reports 1% vacancy for a second consecutive year, more than 35 GW under construction, 92% of under-construction capacity precommitted, and 64% of capacity under construction in frontier markets. The capital-deployment implication is that the opportunity set is moving outward from mature hubs, but only precommitted, power-secured, utility-executable projects deserve underwritable treatment. Texas may overtake Northern Virginia by 2030 if the state is viewed as one market, but that is a power / land / utility-execution thesis, not a generic Texas land thesis.

Soluna's August 2026 Project Dorothy 3 land purchase supplies a site-specific frontier-market example: 397 acres in Briscoe County next to a company-owned 150 MW wind farm and an existing data-center campus. The sponsor describes up to 300 MW of future AI capacity, but the captured record does not establish construction financing, a tenant lease, delivered capacity, or utility approval. Count the acreage as company-reported site control and the MW figure as a development target; do not translate either into operating supply or a general West Texas valuation premium. See Source: Soluna Project Dorothy 3 397-Acre Acquisition 2026 and Project Dorothy 3 AI Data Center Campus.

Source: JLL Power Progress in Your Global Data Center Expansion adds the occupier-portfolio version of the same constraint. The source is useful because it separates AI training near power from inference closer to users, keeps traditional cloud / storage workloads in the demand mix, and ties capacity reservation to multiyear utility lead times. It also shows why power partnerships and tenant-credit economics belong in the capital-deployment map: the same demand headline can price differently depending on energization timing, rack density, counterparty credit, escalation floors, and pass-through exposure.

Source: Avison Young Q1 2026 U.S. Data Center Market Overview adds a current U.S. operating check beneath the capital map. Avison Young reports 29.0 GW of U.S. colocation inventory at Q1 2026, up 22% quarter over quarter and 48% year over year, while net absorption reached 5.29 GW and vacancy declined to 1.2%. That strengthens the capital-deployment read but also narrows it: capital is chasing a market where available capacity is scarce and new supply is largely pre-leased, so the underwritable opportunity is still tied to power delivery, preleasing, construction execution, skilled labor, and local approvals.

Source: CBRE Data Centers - U.S. Real Estate Market Outlook 2026 adds the 2026 CBRE forward-looking layer. The source verifies the previously review-only outlook rows: 300 MW-plus delivery inside 36 months is now a site-selection threshold, 500 MW-plus AI campuses push schedules into multi-year territory, interconnection can stretch to 24, 36, or 48-plus months, and primary-market preleasing is expected to remain in the mid-70% range versus a 40% to 50% historical norm. Capital deployment should therefore favor proved power paths and preleased / customer-backed capacity over generic land-control stories.

Source: Colliers 2026 Data Center Marketplace Report adds the Colliers capital-intensity counterpart. The applied rows (market_observations.id=33900-33914) show the scale of the lane: $580B-plus global data-center investment in 2025, $120B-plus hyperscaler debt issued for AI infrastructure, 47% year-over-year build-cost growth, 15.6 GW of North America absorption, and 90%-plus preleasing of new capacity. The capital-deployment read is selective, not euphoric: private credit and utility deposits are moving earlier in the cycle, but $64B-plus of delayed or canceled U.S. projects since 2023 shows that local process and power deliverability can still kill nominal demand.

Source: Trepp AI Buildout Data Centers CMBS Capital July 2026 adds the securitized-credit boundary to that capital map. Trepp describes data-center issuance as remaining active through the 2023 issuance trough and retaining a larger CMBS share as traditional-property issuance recovered. This is evidence of investor attention to cloud / compute collateral, not a project-screening shortcut: debt and equity deployment still require power delivery, tenant credit, DSCR, loan maturity, and refinancing proof.

Funding-market watch (September 2026): The Real Deal's account of Barclays commentary reports $11.9B of data-center ABS issuance through September 30, versus $15.5B for full-year 2025, and says data-center ABS spreads diverged from broader ABS spread compression. It also describes hyperscaler corporate debt spreads widening year over year. These figures lack the underlying security-level Barclays data in the preserved source set; the full-year and partial-year issuance periods differ, and the ABS and corporate spreads use different benchmarks. This adds a financing-competition watch item to the delivery-chain screen, not evidence that data-center collateral quality or project demand has weakened. See Source: The Real Deal on Data Center Securities and AI Debt Supply and CMBS and Commercial Mortgage Securitization.

The Five-Lane Capital Map

1. Hyperscaler and program commitments

This is the largest headline bucket and the hardest for ordinary real estate capital to access directly.

Representative signals include:

• Amazon's $15B northern Indiana program with 2.4 GW
• Google's roughly $40B Texas buildout through 2027
• Stargate's $450B+ program-scale commitment and 8+ GW plan
• Amazon's government AI infrastructure lane — a November 2025 announcement of up to $50B and 1.3 GW of planned capacity, with intended groundbreaking in 2026; not observed investment, delivered supply or parcel-level government demand. See Source: Amazon to Invest Up to $50B to Expand AI Infrastructure for U.S. Government.

The core lesson is not the raw dollar amount. It is that hyperscalers route wherever power, transmission, land, and utility structure clear. Prestige geography matters far less than infrastructure readiness.

2. Platform consolidation

This lane is different from new campus development. It is infrastructure capital buying operating platforms.

This is the valuation lane, not the land-acquisition lane. It affects what institutional scale looks like and what private portfolios are measured against.

2A. Power-system capital

This lane sits upstream from real estate but increasingly governs which sites become real. DOE nuclear-financing comments point to public-sector intervention in long-lead generation procurement, while the reported NextEra / Dominion transaction shows regulated utility platforms being reframed around data-center load growth. This is not directly acquirable real estate, but it changes the underwriting context for powered land, utility partnership, and ratepayer-risk exposure.

3. Financed metro and exurban delivery

This is the most actionable middle layer because it contains actual projects, debt, and delivery signals.

This is the lane where outside investors can often act indirectly through powered land, utility relationships, adjacent industrial, or operator and lender exposure. It should not be converted into a broad office-demand claim unless a separate office leasing source proves that specific read-through.

The Fisk University Quantum Leap source belongs outside this delivery lane even though the headline uses data-center language. The preserved article body supports an institutional Innovation Center / campus-technology read, not commercial colocation or hyperscale inventory. That distinction is useful because AI-infrastructure RSS items often blend real data-center capacity, campus technology projects, and marketing language. See Source: $1B Fisk Upgrade Includes On-Campus Data Center and Source: Prime Data Centers Breaks Ground on Second Sacramento Facility.

4. Supply-chain and manufacturing spillover

Source: Crusoe Building $200M Denver-Area AI Factory supplies a March 2026 manufacturing-plan example. The $200M-plus commitment spans leasing, build-out and an initial module fleet, separate from property price or delivered compute supply. Q3 first modules were expected, not verified here.

Use Industrial Location and Tenant Demand Underwriting for manufacturing tenancy, space and local demand, and Energy Infrastructure and CRE for factory-versus-module service. Supply-chain exposure is a strategy to test against actual customer, cash-flow and execution evidence, not automatically cleaner or more understandable than campus exposure for every investor.

5. Watch-list announcements

Source: MacroValor, Favis Will Build Hydrogen-Powered AI Mega Campus reports an April 1, 2026 Austin mixed-use proposal with a 3,000 MW headline. Keep it in this watch-list lane. The announced scale does not establish allocated IT load, delivered power, financing, permits or current project progress. Continuous, zero-emission operation remains an attributed project claim. Use Energy Infrastructure and CRE to test the supply-to-operations evidence before comparing it with operating or funded projects.

This distinction matters because the digital-infrastructure branch now produces a lot of spectacular announcements. Some deserve underwriting attention. Some only deserve monitoring.

What Actually Creates Real Estate Opportunity

Source: Connect CRE IIF Data Center Geography 2026 sharpens the location screen by separating facility density from campus scale. Its IIF-attributed summary places dense facility clusters in major population hubs such as Data Center Alley and Silicon Valley, while larger hyperscale campuses are more common in less-populated regions. It also cautions that power/land correlation does not identify whether supply attracts development or demand drives generation expansion. Treat these as qualitative, secondary-reporting signals; no capacity or county metrics are available.

Powered land beats general market branding

Northern Indiana, Haskell County, Castroville, Parker, Phoenix, and Northern Virginia Digital Infrastructure Corridor provide cases for site-specific power, transmission, and delivery diligence. Keep reported proposals, financing events, and delivered operations separate; these examples do not prove that utility constraints have cleared at every site or rank them above all other underwriting factors.

The best adjacent exposure may not be the campus itself

The cleaner tradable edges are often:

• powered land before full campus announcement
• utility-enabled industrial parcels
• supply-chain manufacturing and modular assembly sites
• mature campuses with proven tenants and financing liquidity

Texas is a stack, not a single node

Texas keeps recurring in this source set because it wins across multiple forms at once:

• metro-integrated enterprise campuses
• outer-ring hyperscale buildout
• exurban and rural power-first siting
• platform scale and operating history

That does not mean every Texas announcement deserves the same weight. It means Texas has the broadest state-level stack in the current material.

Best For

• Investors with real utility, powered-land, or infrastructure relationships
• Institutional capital targeting proven operators, mature campuses, or platform-scale exposure
• Industrial investors near confirmed compute and equipment clusters who understand spillover demand

Wrong Fit

• Treating every AI infrastructure headline as equivalent to financed or delivered demand
• Assuming metro prestige matters more than grid headroom and entitlement clarity
• Mixing platform M&A, campus announcements, and supply-chain manufacturing into one blended underwriting story

What To Track Next

• Site-level Texas detail for Stargate rather than state-level mention only
• More Castroville and Medina County disclosure
• Whether Phoenix, Parker, and Indiana-type secondary wins keep compounding
• Which watch-list projects move into entitlement, financing, and actual construction

Gaps

• Stargate remains much clearer at the program level than at the site-by-site level.
• The federal AI lane now has an initial geography node in Northern Virginia and Washington DC, but it is still thinner than the Texas and Phoenix branches.
• Some of the biggest headlines are commitments rather than fully disclosed project pipelines.
• MacroValor remains a watch-list signal, not a de-risked project.

Source: BGO 2025-2026 Sustainable Investing Report adds an infrastructure-design and operating-resilience control. BGO's data-center case study links low-carbon power, PUE, closed-loop cooling, water, modularity, grid flexibility, waste-heat reuse, and circular construction, while explicitly labeling the features indicative rather than universal. For underwriting, require utility capacity, cooling and water design, heat-rejection, construction budget, operating target, tenant workload, environmental permits, and resilience evidence; do not capitalize the report's example features as a generic data-center premium.

Source: BGO 3Q 2026 Global Outlook and House Views identifies powered data centers as a selective opportunity while emphasizing current financing conditions. That reinforces the delivery-chain gate: utility capacity, interconnection, fiber, cooling, water, land, permits, construction capital, tenant workload, and exit liquidity must be evidenced before a powered-land or data-center demand thesis receives value.

Source: BlackRock 2026 Private Markets Outlook adds an institutional-capital and execution lens to the data-center thesis. BlackRock describes data centers as moving from niche to core real estate, reports that 39% of real-estate investors in a cited source planned to buy data-center properties, and emphasizes grid connectivity, behind-the-meter power, and fully outfitted suites as real-estate participation points. Use the signal to prioritize power, interconnection, cooling, delivery, counterparty, and infrastructure-exit diligence; the cited investor-intention figure is not a current market-clearing forecast.

Power Ownership, Conversion Finance, and Mature-Core Liquidity

The backlog adds several mechanisms that sharpen the national digital-infrastructure screen. Digital Realty's reported $3.5B purchase of Blackstone-affiliated funds’ 80% interests in two Manassas facilities and 50% interest in one Sterling facility (blended 64%), each reported at 96 MW of IT capacity, is a development-to-stabilization portfolio-liquidity datapoint; it supports portfolio-scale capital formation but is not a metro-wide valuation benchmark. Source: Digital Realty Agrees to Buy Blackstone’s Stake in Northern Virginia Data Center Portfolio for $3.5B

Power is also moving upstream into project design and ownership. A proposed Tonopah, Arizona data-center plan pairs development with a dedicated energy system to reduce reliance on constrained delivery systems, while the Nvidia/Cloverleaf report describes compute capital taking a stake in a data-center power company. These are qualitative signals that behind-the-meter generation and power-infrastructure ownership are becoming real-estate-adjacent underwriting lanes; both remain subject to utility, environmental, site, tenant, and execution diligence. Source: W. Valley Data Center Developer Building Massive Energy Project Source: Nvidia Buys Stake In Data Center Power Company

Irving supplies a conversion-finance example: an $11.6M C-PACE loan reportedly supports conversion of a vacant 113,406 SF flex building into a 3.0 MW data center with a 45,000 SF data hall. Adaptive reuse can bridge obsolete industrial inventory and digital infrastructure when shell, power, cooling, and tenant execution align; the reported completion target remains project-specific. Source: Lone Star PACE Provides $11.6M C-PACE Loan for Data Center Conversion Project in Irving

Chicago-area evidence adds the public-cost and entitlement side of the screen. Reported tax breaks for suburban data centers and opposition to a possible Hawthorne Race Course data-center redevelopment show that incentives, community acceptance, and fiscal externalities can affect project value alongside power and tenant demand. Source: As Data centers cash in on tax breaks, homeowners pay up Source: 108-Acre Hawthorne Race Course Buyer Could Put Data Center On Site

Regional AI-Growth-Zone Deployment

Vantage Data Centers and Nebius announced a plan to deploy high-density, NVIDIA-powered AI infrastructure at Vantage's CWL1 campus in Newport, Wales, described as the first announced commercial capacity commitment in the South Wales AI Growth Zone. The item supports a global mechanism in which a campus, committed AI operator, and regional policy/power capacity convert into commercial demand; it remains an announcement requiring delivery and tenant verification. Source: Vantage Data Centers, Nebius Partner on AI Infrastructure at South Wales Campus

Data-Center Governance and Documentation Risk

The reported subpoenas and document requests involving Fermi's planned 17-gigawatt Project Matador add a governance and diligence risk to frontier data-center development. The source says agencies sought records concerning the project and former management and did not allege wrongdoing; Fermi said it was cooperating. The reusable underwriting implication is procedural: political ties, former-management records, land/power claims, and agency inquiries can affect execution, disclosure, and financing confidence even before a finding of misconduct. See Source: Fermi Subpoenaed For Project Matador, Former Management Records.

Government Lease and Life-Sciences Execution Risk

Alexandria Real Estate Equities' reported dispute with New York City over the Alexandria Center for Life Science adds a public-sector ground-lease and project-execution risk to life-sciences real estate. The $1.5B campus and proposed 550,000-SF expansion sit within a legal fight over lease rights and government actions; the article says courts had sided with the city in multiple rulings and Alexandria warned of a potential $183M loss. This is not evidence against life sciences as a sector, but it shows why land tenure, public counterparties, approvals, and litigation must be underwritten alongside lab demand. See Source: Alexandria Warns $183M Loss On Table In Legal Battle Against NYC.

Colocation Capital Is Broadening the AI Stack

Saragon is reported to have begun with 10 data centers and up to $1 billion in committed capital from I Squared Capital. Source: Saragon Launches to Serve Growing Demand for Colocation This supports a colocation and AI-inference capital layer, but platform announcements do not establish delivered capacity or local lease-up.

Conversions and Anchor Leases Expand the Digital-Infrastructure Map

The backlog adds WhiteFiber’s planned conversions of two North Carolina industrial facilities, a reported 20-year OpenAI lease at a Pike County, Ohio development, and a proposed $100 billion campus in Western Kentucky. Source: WhiteFiber Agrees to Buy Two Industrial Facilities in North Carolina, Plans Data Center Conversions Source: OpenAI Signs 20-Year Lease at Nvidia-Backed Data Center Development in Pike County, Ohio Source: Brookfield, Partner Coalition Plan $100B Data Center Campus in Western Kentucky These are differentiated watchlist signals: conversion feasibility, anchor tenancy, power, and delivery evidence must be screened separately.

Commercial Observer's reported $1.3T data-center spending discussion adds a capital-stack scale marker to the digital-infrastructure thesis. Treat the figure as a source-scoped pipeline/funding narrative rather than committed construction: underwriting still requires contracted power, interconnection, permits, tenant credit, financing terms, and delivery evidence. See Source: Sunday Summary: Who’s Funding $1.3 Trillion in Data Center Spending?.

May 19 2026 RSS Watchlist

• Treats Blackstone's reported $5B AI-cloud venture with Google as digital-infrastructure capital-stack evidence, not a site-level data-center supply metric. See source-blackstone-google-ai-cloud-venture-2026. Caveat: Preserve as strategic capital-allocation context; verify venture terms and site-level exposure before structured import.

Source: The Real Deal Data Center Brokers and Power Real Estate February 2026 adds an execution-first site-screening layer: data-center deals combine land control, power and fiber access, substation proximity, water/cooling, floodplain, zoning, tenant prelease, financing, construction, and community acceptance. The move toward hundreds of megawatts and secondary locations can expand the opportunity set, but it also increases contingency and delivery risk. Treat preleasing or broker enthusiasm as insufficient until utility, interconnection, permits, budget, financing, tenant-credit, and local-opposition evidence clears.

Source: Cushman & Wakefield EMEA and France Data Center Market Study 2026 adds a Europe-wide routing layer: AI and digital demand are expanding across EMEA, but energy and regulatory constraints determine whether hub-level demand becomes deliverable supply. Use Paris, Frankfurt, London, Dublin, and Milan as comparison nodes for power, grid, regulation, land, cooling, tenant, financing, and exit diligence; the public synopsis supplies no capacity or valuation series.

Source: Cushman & Wakefield France AI Data Centers From Algorithm to Infrastructure 2026 turns the France demand thesis into a design-and-capex screen: AI workloads may require high-density compute, increased power capacity, liquid cooling, and different building standards. A low-carbon energy mix and public AI strategy can improve the macro case, but do not substitute for utility commitments, cooling/heat-rejection engineering, construction budget, tenant workload, permit status, and financing. Source: Deloitte 2026 Commercial Real Estate M&A Outlook adds a transaction-model lens to digital-infrastructure underwriting: Deloitte expects data-center deal structures to shift toward joint ventures, development partnerships, and campus expansions as competition for stabilized assets increases. Treat partnership optionality as a financing and execution hypothesis, not as committed capital; require power capacity, interconnection, permits, construction budget, tenant credit, partner governance, and exit evidence before assigning value.

Source: Hines 2026 Midyear Global Investment Outlook adds the capacity-scarcity transmission layer: AI, energy, and infrastructure demand are competing for constrained power, land, labor, and delivery capacity, while Hines identifies powered land as an opportunity area. Underwrite the physical conversion chain—utility capacity, interconnection, permitting, cooling, construction, tenant workload, partner governance, and exit liquidity—rather than capitalizing a data-center demand headline or powered-land optionality alone.

Related Pages

• Texas AI and Industrial Infrastructure Opportunity Map
• Digital Infrastructure Real Estate
• Powered Land and Grid Advantage
• Northern Virginia and Washington DC
• AI Corporate Real Estate Footprint 2026
• Industrial Hub
• Analyses Hub
• United States

California Fallback Demand Versus Deliverable Power

September 2026 Bisnow reporting relays JLL commentary that developers and tenants are reconsidering California as power and community opposition constrain other western markets. The same report says Vernon’s larger projects in development would use currently available capacity, with the city queueing future large users while expanding infrastructure. This is a qualitative interview-based signal, not a quantified pipeline or evidence of energized supply. California therefore remains a selective fallback hypothesis: validate utility capacity, queue position, permits, land, water, tenant commitment, and local policy for each site before underwriting. See Source: Bisnow California Data Center Fallback Market 2026 and LA Data Center Market.

Data-Center Credit Surveillance: CyrusOne DFW1

A $687.1M loan secured by a CyrusOne Carrollton data center appeared in CRED iQ’s August industrial-alert discussion, but CRED iQ says the initial performing-matured classification was later revised to current after the borrower exercised an extension. Bisnow reported an August closing and an August 2027 extended maturity. This is a loan-status and data-classification check, not evidence that data-center cash flows failed: separate operating, debt-service, extension-condition, and collateral facts before using it as a sector credit signal. See DFW Data Center Market, Source: CRED iQ Property Types Feeling the August Heat 2026, and Source: Bisnow August CMBS Distress and CyrusOne Extension 2026.

Bain's 2026 scenario adds a demand-funding stress lens: it estimates $1.5T in annual global AI infrastructure spending by 2031, with a $6T annual AI market needed under its approximate 25% capex-to-revenue assumption. Bain estimates existing consumer and enterprise uses could account for $1.2T–$1.8T, leaving about $4.2T in revenue from new value categories. The infrastructure-spending definition includes data centers, compute, and continuing chip/network upgrades; this is not a real-estate-only capex forecast. Treat the figures as scenario assumptions, not committed tenant demand, and keep site underwriting tied to contracted workloads, credit, utility deliverability, financing, and construction evidence. No property or market observation was added. See Source: Bain AI Infrastructure Funding and Revenue Scenario 2031.

Infrastructure Services and the Property Cash Boundary

Issuer evidence, as of 2026-10-20. CBRE's original Onnec release describes a conditional corporate acquisition and explicit 2027 infrastructure-services forecasts. It adds evidence that a service provider can seek fees from network design, installation and maintenance across the asset cycle. It does not add owned facilities, energized megawatts or new local leases. The nearly $7 billion combined services outlook has its own business-line scope and excludes periodic development land sales; it cannot become current property NOI or a historical growth series. Source: CBRE Onnec Acquisition Agreement and 2027 Guidance.

Hypothesis and counterevidence. Service capabilities may capture spending across customers and locations without owning each building. Test integration costs, earnout conditions, customer concentration, technical delivery, staffing and actual collections against weaker customer capex and project delays. For a property investor, separately test site control, available power, funded tenant demand, costs and lease cash. A completed transaction, comparable segment results and verified customer/project delivery could change the view.

Sources

• Source: Federal Reserve FOMC Statement September 16 2026

• Source: U.S. Treasury Daily Par Yield Curve September 2026

• Source: Samsung $1B Commitment to Helix Digital Infrastructure 2026

• Source: Bain AI Infrastructure Funding and Revenue Scenario 2031

• Source: CRED iQ Property Types Feeling the August Heat 2026

• Source: Bisnow August CMBS Distress and CyrusOne Extension 2026

• Source: Commercial Observer Quantum Coast Santa Barbara 2026 — September 28, 2026; OASIS size and lease cross-checked to UCSB.

Source: Equinix Sets More Records with Big Q1, Raises Outlook on AI Momentum supplies an issuer-level operating case: Equinix reported 82% stabilized utilization, 6% year-over-year revenue growth, 3 GW of developable capacity, and $4.1B of expected 2026 capex, with roughly 60% of its largest Q1 deals tied to AI. Use these as platform-specific demand and capital-intensity evidence, not a sector-wide forecast.

• 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: Google Michigan Clean Energy and Reliability Announcement 2026
• 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
• Source: Related Digital Secures Financing for $16B Oracle Data Center Project in Michigan
• Source: Vantage Data Centers to Invest $2B for New Campus in Northern Virginia
• Source: Prime Data Centers Breaks Ground on Second Sacramento Facility
• Source: $1B Fisk Upgrade Includes On-Campus Data Center
• Source: DOE To Finance Parts for 10 Nuclear Reactors to Power Data Centers
• Source: NextEra Dominion Utility Merger and Data Center Power Demand 2026
• Source: The Whale Hunting Era Of Data Center Development May Be Ending
• Source Collection: CBRE Insights Market Reports Public Crawl 2026
• Source: CBRE New York Tri-State Data Center Market H2 2025
• Source: CBRE Northern Virginia Data Center Market H2 2025
• Source: CBRE Northern Virginia Data Center Market H1 2025
• Source: CBRE Silicon Valley Data Center Market H2 2025
• Source: CBRE Silicon Valley Data Center Market H1 2025
• Source: CBRE Minneapolis Data Center Market H1 2025
• Source: CBRE Houston Data Center Market H1 2025
• Source: CBRE Denver Data Center Market H1 2025
• Source: CBRE Seattle Data Center Market H1 2025
• Source: CBRE Central Washington Data Center Market H1 2025
• Source: CBRE Charlotte-Raleigh Data Center Market H1 2025
• Source: CBRE Southern California Data Center Market H1 2025
• Source: CBRE Montreal Data Center Market H1 2025
• Source: CBRE Toronto Data Center Market H1 2025
• Source: CBRE Hillsboro Data Center Market H1 2025
• Source: JLL 2026 Global Data Center Outlook
• Source: JLL North America Data Center Report Year-end 2025
• Source: Soluna Project Dorothy 3 397-Acre Acquisition 2026

October 8 — CRED iQ corrected alert coverage

Four source-defined August alert disclosures and one explicitly disputed CyrusOne classification/chronology comparison; reported balances remain JSON disclosures, not normalized current defaults or losses. Existing six April/July rate observations are reused and exactly verified. All four newly distressed alert categories, corrected CyrusOne extension and remaining article paragraphs were examined. Different loan amounts/extension dates and the August-versus-September Ritz chronology are expressly unresolved. No arithmetic adjustment, current-default status, loss or confirmed hotel sale is inferred.

Evidence: Source: Bisnow August CMBS Distress and CyrusOne Extension 2026. Receipts: 2026-10-08-rss-cred-iq-august-alert-disclosures, 2026-10-08-cred-iq-prior-import-verification.

Loudoun Proposal — September 24 Evidence Boundary

Full retained county and reporting-source review confirms that September 15 advanced a proposed final-Board-action pause. The October 20 resolution was pending as of September 24. This record does not establish an enacted moratorium, an October 6 decision, project cancellations or loss of powered capacity. Source: Source: Loudoun County Data Center and Substation Application Pause Proposal 2026.

WhiteFiber Conditional Conversion Update — August 17 Evidence

The primary agreement reports $60M combined consideration for two Yadkin County industrial properties, with planned NC-2/NC-3 conversions. Conditional gross electrical supply is at least 30MW each and 60MW combined; possible additional 69MW each within seven years after closing yields the reported potential combined maximum 198MW. The furnished exhibit describes approximately 200MW, expected Q4 2026 closing and targeted Q3 2027 initial service. Preserve the source-specific figures. Utility confirmation, an energy study, property-separation approvals and agreed occupancy/leaseback terms are material closing conditions; nonbinding customer LOIs do not prove tenancy. The evidence does not establish closing, title, energized IT load or project investment returns. Source: WhiteFiber Yadkin Purchase Agreement SEC 2026

Named conditional project records: WhiteFiber NC-2 Yadkin Conversion Project, WhiteFiber NC-3 Yadkin Conversion Project. These are announced future identities with unknown parcels and no verified completion or energized load.

Ohio PORTS-Pike Disclosure Review — August 17–18, 2026

Source: NVIDIA PORTS-Pike Guaranties and Furnished Announcement 2026 and Source: OpenAI PORTS-Pike Project Announcement 2026 establish dated conditional project announcements. The $105B is the initial cumulative residual-value guarantee payment cap with ready-for-service and tenant-default conditions, recovery offsets and termination rights; it is not loan proceeds or construction cost. Initial 4.25IT-GW, future approximately 8IT-GW campus and at-least-10GW new generation have different universes. The parent rounds optional support to approximately 3.8GW; its furnished release states 3.75IT-GW. None measures energized current supply.

The first 800MW expected in 2028 and forecast construction/operating employment depend on financing, power, permits and environmental review. Lease payments begin as capacity becomes available. Corporate investment, community grants and in-kind credits are separate from campus property price. An anchor agreement can support phased development, but an investor should stress delayed delivery, financing cost, reduced demand, replacement-tenant recovery and water/regulatory costs against acquisition price, horizon and liquidity. Signed utility delivery, final permits/design and commenced rent would change the conclusion. Do not count issuer forecasts as realized regional growth or move company capital to one site.

Named prospective identity: PORTS-Pike Technology Campus. The native project record preserves unknown parcel/title, size, price and current operation; it does not establish power delivery.

Paducah Original Scope Review — July 2026 Announcements

The full secondary article describes proposed capacity and investment, not current energized inventory: utility up to 1.8 GW, compute above 1.2 GW, and a generation/storage program up to 4.6 GW. Battery power is not energy duration or guaranteed firm supply. The legacy DOE photograph does not identify future-campus acreage or remediation completion. DOE and coalition discovery have 2031 versus 2032 targets; definitive documents and regulatory conditions require separate primary review. At the investor’s price/horizon/risk tolerance, stress schedule, environmental, grid/gas/storage, tenant-contract and capital-financing gates before converting the announcement into an investment conclusion. Source: Brookfield, Partner Coalition Plan $100B Data Center Campus in Western Kentucky

DOE Primary Scope

Source: DOE Paducah Energy Hub Primary July 2026 independently reports the 2031 target, continued cleanup, selected development/energy roles and pending Kentucky PSC approval. Proposed gas/storage/campus power and job outcomes are forecasts, not commissioned capacity or realized benefits. Exact land/title/contract and site geometry remain unknown.

Bounded Paducah identity: Planned Paducah Data Center Campus Kentucky. Keep the proposed campus separate from the whole photographed legacy federal site and from actual operating inventory.

Northern Virginia Portfolio Measurement Scope — June 2026

Source: Digital Realty Agrees to Buy Blackstone’s Stake in Northern Virginia Data Center Portfolio for $3.5B separates $3.5 billion equity consideration from $7.8 billion gross value including assumed debt and remaining capex. The above 6.5 percent cap rate is expected on stabilization; fully leased does not prove occupied or rent-producing. Two facilities were forecast to stabilize in H1 2027 and the third in 2028. Review closing, debt/capex, power and rent commencement before current-income valuation. Full source/graphic review passes; primary and native integration pass with 86 verified creates across five unique packets.

Independent Issuer Release Scope

Source: Digital Realty Blackstone Northern Virginia Purchase Primary June 2026 reports 288 MW total IT, 80 percent interests in two Manassas facilities and 50 percent in one Sterling facility on Digital Dulles campus. Preserve 100 percent gross value versus blended 64 percent equity purchase. Third-facility H1 2028 stabilization and expected June 30 close are forecasts; 100 percent leasing is not current physical occupancy. Complete Safe Harbor reviewed. Source-specific native and independently reviewed completed-close evidence pass.

Filed Tenant Credit and Completion Risks

Source: Digital Realty Blackstone Agreement and Development Risks SEC June 29 2026 reports that tenants are subsidiaries and investment-grade parents are not lease parties or assured guarantors. Estimated stabilized cap rate depends on completion and future first-year NOI; parent credit, expected leverage neutrality and Core FFO accretion do not establish current income or achieved return. Dated native integration passes.

Filed Completed Close and Seller Proceeds

Source: Digital Realty Blackstone Completed Acquisition and Seller Offering SEC July 1 2026 reports June 30 completed acquisition. July 1 seller secondary offering produced no proceeds to Digital Realty; separate securities disposal from campus acquisition funding or operational completion. Dated native integration passes.

Physical scope: Digital Dulles Campus Sterling is separate from a specific acquired facility and the existing proposed PhaseIII record; Astra Enterprise Park Land Development Near Kansas City is separately dated land development, with exact jurisdiction still unknown.

Beacon Point dated finance and delivery boundaries

Source: Hut 8 Q2 Results Primary August 2026 reports $4.25 billion Beacon Point Phase 1 notes, separately from $3.25 billion River Bend financing. Its 1,330 MW utility-construction presentation includes 500 MW added after June 30; it is not a pure quarter-end market-inventory measure. Source: Hut 8 Beacon Point Second Lease Primary July 2026 separates base contract value, optional renewal value and expected non-GAAP NOI. Full commercialization does not establish current occupancy or rent-producing capacity.

For an investor’s price, objectives, horizon and risk tolerance, test delayed power and rent commencement, construction carry/cost overruns, counterparty/backstop terms, refinancing and exit-cap scenarios. Executed terms, verified commissioning and observed operating cash flow can change an investment conclusion. Independent primary source processing and applicable native integration pass; the original secondary article’s unavailable lead image remains a scoped full-processing hold.

Complete liquid-cooling evidence review, October 11, 2026

The complete September 11 Connect article reports JLL liquid-cooling commentary and operator examples. Equinix deployment across more than 100 data centers is an attributed operator claim, not measured industry adoption or a verified inventory. Digital Realty's up-to-70-kW-per-rack offering is a maximum capability, not typical occupied density or building demand. Nvidia server designs are qualitative product context, not a facility contract. Floor-space and height benefits, reduced cooling energy and substantial savings are qualitative; no floor-area gain, PUE target, tariff, cost, capex, percentage saving or payback is supplied. Retrofit overhaul, power/infrastructure constraints, live-hall service risk and premature obsolescence are explicit counterevidence to a universal upgrade premium. Three complete faithful panels and the linked stock image were inspected; latter two panels are clamped duplicates. Visible piping and racks illustrate the subject; no named facility, operating performance, rack count, coolant specification or measured capacity is inferred. Attribution on.u/Shutterstock is kept; raw redistribution is not permitted. Primary engineering values are independently reviewed in the JLL record and are not additional printed numbers in this recap.

The complete retained JLL insight has displayed update September 1, 2026 and says original publication March 18, 2024. Author Shaun Lee and expert Andrew Green, Regional Data Center Practice Lead, Work Dynamics, are identified. JLL reports Equinix deployment beginning in 2023 across more than 100 data centers in 45 cities, Digital Realty's up-to-70-kW-per-rack colocation offering and Nvidia liquid-cooling server designs. These are attributed operator examples, not independent inventory, typical density or property contracts. Green describes rack densities growing beyond 70 kW and direct-to-chip/immersion approaches for high-performance workloads; the inevitability/only-viable claims are dated expert assessments, not universal engineering standards. An immersion design can use four-meter floor-to-floor height versus six meters described for most existing facilities. For the same compute in the same floor area, cited structural loading rises from 12–15 kPa to at least 20 kPa. These conditional design examples require site-specific load paths and slab, plumbing, electrical and resiliency review; neither figure is measured for a named asset. Removing air-cooling equipment can free space, but no quantified area gain, capacity uplift, investment premium or PUE value is supplied. Cooling energy/cost savings are qualitative with no tariff, percentage, capex or payback. Direct-to-chip requires coolant plumbing; retrofit may disrupt other live halls. Partial upgrades with spare capacity may be possible, while full 100% AI conversion is described as unlikely because of power/infrastructure constraints, not a legal or universal physical prohibition. Premature technology obsolescence is an explicit risk. All five contiguous faithful panels and the original worker/rack image were inspected; no actual service, capacity, identities or performance follows from that illustration. Existing candidate PUE 1.15, rejection above 1.4, Tier claims and universal water requirements remain outside this source's support.

• Source: JLL Liquid Cooling Enters the Mainstream in Data Centers.
• Source: Connect Liquid Cooling Complete Secondary Review.

Complete reproduced source review, October 11, 2026

Complete September 30 reporting preserves one combined $1 billion Samsung commitment alongside more than $10 billion of founding platform commitments. These are announced capital commitments, with funding/deployment, project allocations and realized returns unknown. Worldwide data-center/power/connectivity strategy and expected Samsung benefits are participant plans, not delivered capacity or signed contracts. Existing separate issuer affiliate allocations retain attribution. The unidentified hero image establishes no bounded property, workplace or Helix ownership; original historical image identity remains unresolved.

• Source: Connect CRE Samsung Helix Complete Commitment Review September 30 2026.

Complete official packet review, October 11, 2026

The full29-page September9 packet contains a blank resolution template, the adopted June9 initiation, separate board/staff drafts and an August13preliminary estimate. Page28 reports approximately237parcels/3654acres versus9698existingacres; the criterion counts1/0/78/3/155 sum237. Its pie labels use inconsistent rounding. Page29 depicts the existing overlay only and disclaims precise construction/property-boundary use. These figures and redlines do not certify the final ordinance, final post-deadline map, parcel eligibility, current supply or a scarcity premium. Keep the later county summary’s deadline and historical recommendations in their own dated sources.

• Source: Prince William County Overlay Complete Staff Packet Review September 9 2026.

Complete reproduced source review, October 11, 2026

As captured October11, the county project page states September22 adoption as Ordinance26-67, describes five eligibility paths and explicitly gives December21 as the fifth-criterion close. The county says the updated qualifying-parcel map will follow that deadline. These are county summary statements; the complete final ordinance and parcel-specific rights remain unheld. Older September9 recommendations for120days and500-foot setbacks, and a retained September22 hearing notice, are historical material rather than adopted deadline proof. Keep the listed SUP pathway distinct from a blanket development ban and from guaranteed approvals or power service.

• Source: Prince William County Overlay Complete Project Page Review October 11 2026.

Complete reproduced source review, October 11, 2026

Complete September22reporting distinguishes the8–0overlay vote from5–3rejection of zero-day grace. The3654-acre figure derives from an August13preliminary staff estimate; the packet map shows existing overlay only, and the county’s later summary says the updated qualifying map follows December21. Keep the3641-acre secondary figure unresolved. USDataCenterMap’s36operational/40pipeline sites include11under construction;4.5GW mixes reported and operational power and is not available energized capacity. Loudoun’s contemplated legacy change and up-to12monthpause belong to a different jurisdiction. QTS Digital Gateway100B/2100acres and DullesSouth1940acres are separate dated proposals/outcomes, not delivered supply or quantified losses. No final parcel rights or investor premium is certified.

• Source: Bisnow Prince William Overlay Complete Vote and Pipeline Review September 22 2026.

Complete reproduced source review, October 11, 2026

The April 24 developer release names the $16 billion Saline project and its equity, debt-anchor and advisor roles, but expressly retains customary closing conditions. It does not disclose debt principal, funded equity or completed closing. Linked photos and sampled drone views show partial construction; the rendering is proposed, and more-than-one-GW wording is not defined commissioned IT capacity. Jobs, savings, benefits, preserved land and corporate pipeline figures retain distinct forecast and entity scopes.

• Source: Related Saline Financing Primary Release Review April 24 2026.