Power-site underwriting OS for AI campus developers to find, option, and de-risk power-adjacent land before rivals do.

1. Capital is moving upstream from compute into land acquisition, so a real buyer now exists before a campus is designed.
2. Anthropic being cited as a beneficiary means dedicated site pipelines can be built around named AI demand rather than generic colocation speculation.
3. Parcels near large power sources are becoming strategic inventory, making power diligence the first gating workflow instead of a later engineering task.
4. A reported $50B Anthropic-linked buildout raises the cost of site mistakes and rewards tools that standardize investment-grade underwriting.

Catalyst. Coatue's launch of a land-buying vehicle and the Anthropic-linked buildout signal that site control has become an urgent, funded workflow at the start of AI infrastructure development.

The product ingests candidate parcels and structures every diligence step around one question: can this site realistically support a large AI load on the timeline an anchor tenant needs. It combines parcel-level evidence capture, standardized kill-go checklists, and confidence scoring across power proximity, water, zoning, environmental constraints, and campus fit. Teams get a shared memo they can use internally for investment committee decisions and externally with consultants, utilities, lenders, and prospective tenants. Over time, the platform becomes a proprietary dataset of what actually makes AI campus sites transact and get built.

What's different. Generic commercial real estate software is not built for AI-load siting, and consultant-led studies are too slow and inconsistent for a land rush. This product is power-first and memo-first: it helps teams decide which parcels deserve exclusivity before they pay for full engineering. Each screened site adds structured evidence about what utilities, tenants, and capital providers actually accept, creating a feedback loop consultants and brokers do not own.

Jobs to be done

flowchart LR
  Buyer[AI campus developer] --> Pain[Too many parcels, unclear power certainty]
  Pain --> Product[Power-site underwriting OS]
  Product --> Outcome[Faster site options and fewer dead-end parcels]

Executive takeaways

• Capital is moving upstream from GPUs into powered land; Coatue's launch makes pre-option site control a funded workflow, not just a brokering exercise [1][19][21][27].
• The bottleneck is power certainty and utility/process timing, not simply parcel discovery; load-queue, transmission, and large-load intake evidence shows why spreadsheets are too brittle [8][9][15][16].
• This is a narrow but valuable buyer set: dozens of developer, operator, and land-vehicle teams can justify high ACVs because one bad parcel can waste months of diligence and option spend [6][10][15][19][22].
• Adjacent tools cover parcel data, power overlays, or advisory services, but the market still lacks a memo-first system of record that unifies kill/go evidence for AI-campus land [3][5][7][10][13][14][28].
• Northern Virginia is the best proving ground because county standards and Dominion's explicit data-center workflow create repeatable, observable pain; Texas and Midwest corridors are the next logical expansion [15][16][17][18][19][20][27].
• Biggest risk is not demand; it is adoption friction from utility opacity, consultant channel resistance, and the fact that this sells like infrastructure diligence, not lightweight proptech SaaS [12][15][16][17][28].

Market definition

This market is pre-development software for large-load AI/data-center site underwriting: parcel intake, power/load feasibility, zoning and environmental gating, and investment-memo creation for 100MW+ campuses in U.S.-led markets [1][2][3][15][17][19]. It excludes generic CRE CRM, colocation leasing marketplaces, and downstream EPC/construction execution systems [4][5][24][26].

Customer and buyer

Core ICP is the development/origination team inside data-center developers, land aggregation vehicles, and infrastructure investors pursuing powered campuses; the economic buyer is typically the VP/Head of Development or infrastructure investments, while the daily user is a site origination or development manager [3][5][10][15]. Budget likely comes from land origination, preconstruction diligence, or development operations rather than IT, and procurement friction is high because utilities, counties, and consultants still own parts of the truth set [6][15][16][17][28].

Buying triggers

• A new site-option budget or named tenant/AI-lab inquiry forces fast first-pass kill/go screening before exclusivity is granted. [1][10]
• Utility intake for a large-load or data-center request creates a deadline to package credible power and timeline assumptions. [15][16]
• County standards updates or expansion into a new market make reusable site-readiness checklists suddenly valuable. [17][18][29]

Willingness to pay

No public SaaS price points surfaced, but buyers already pay for consultant-led site selection, utility coordination, and large campus/JV commitments; that supports willingness to pay for software that reduces false-positive site options and creates reusable diligence packs. [5][6][15][19][20][26][27][28]

Category dynamics

Growth signal Rapid expansion; AI-campus demand is rising faster than utility and permitting capacity, but no single retained fetchable CAGR was more decision-useful than the operator and advisor activity set.

Validation signals

• Coatue launched a powered-land venture tied to AI data-center demand, validating upstream land as an investable workflow.
• Dominion maintains a dedicated data-center requests process, a strong sign that specialized large-load demand is operationally real.
• Loudoun publishes data-center-specific standards and locations material, signaling localized process complexity in the largest U.S. market.
• Applied Digital says it brought its first 50 MW online at Ellendale, underscoring the scale and capital intensity of AI-campus development.
• Crusoe is publicly showcasing the Abilene AI data center, reinforcing the move toward bespoke AI infrastructure campuses.
• Digital Realty has added DGX-H100-ready capacity and a build-to-suit JV, showing incumbents are still investing aggressively in AI-ready formats.

Regulatory & technical constraints

• Utility large-load and data-center request processes are explicit enough that any product must support formal evidence packaging, not just lightweight scoring.
• County zoning and data-center standards vary materially by locality, making templated but market-specific workflows essential.
• Power-ready is not the same as shovel-ready; over-automating site confidence without source evidence risks credibility loss.
• AI-campus technical needs are tenant-specific around density, cooling, and timing, so parcel scoring must stay configurable.
• The product will sell into infrastructure budgets and diligence processes, implying enterprise procurement timelines and security expectations.

The competitive set is fragmented across parcel-data platforms, power-intelligence layers, renewable siting tools, and incumbent advisors. Acres handles parcel/ownership/reporting workflows; LandGate layers power, fiber, and energy-market intelligence; Paces is closest to a power-development operating workflow; CBRE and Colliers sell services-heavy site-selection and capital-markets advisory [3][5][7][10][13][14][28]. The gap is a neutral, memo-first system of record built specifically for AI-campus kill/go decisions before full engineering spend [8][9][11][12][15][18].

Why incumbents do not win by default

• Brokerage and advisory firms. CBRE/Colliers win relationships but not by default as software owners; their deliverables are services-heavy, project-specific, and less likely to compound into a reusable underwriting dataset.
• Horizontal land intelligence platforms. Acres and similar parcel-data products solve ownership and mapping, but they stop short of integrating utility timing, zoning nuance, and tenant-fit into one investment memo.
• Renewable siting tools. Paces is strong on power-development workflow, yet AI-campus buyers also need land committee memos, water/zoning evidence, and anchor-tenant fit rather than generation siting alone.
• Operators and cloud platforms. Equinix, Digital Realty, Crusoe, and CoreWeave sell capacity or campuses, not a neutral pre-option underwriting layer for third-party buyers comparing raw parcels.

Power-site Underwriting OS targets the pre-development workflow where AI campus developers and land vehicles decide whether a 100MW+ parcel is worth optioning. The researched market shows urgent, funded demand for powered land, but the workflow is still fragmented across brokers, consultants, utility emails, and spreadsheet memos. The initial beachhead is Northern Virginia origination teams because Dominion's dedicated data-center intake and Loudoun's explicit standards make the pain observable, repeatable, and productizable. The MVP is a memo-first, human-verified evidence system that turns raw parcels into kill/go decisions across power, zoning, water, environmental, and tenant-fit criteria. Go-to-market starts with paid pilots triggered by new site-option budgets or named tenant inquiries, then converts successful teams to annual platform contracts plus per-parcel fees. Based on research estimates, the initial U.S.-first SAM is about $31.0M and the three-year SOM is about $6.0M, so venture upside depends on expanding from underwriting into option management, utility engagement, permitting coordination, and lender diligence. The deliberate choice is to avoid generic CRE software, full interconnection automation, and downstream EPC tools until the company has repeated proof that its memo workflow reduces dead-end parcels and speeds IC approval. The biggest gap is not market demand but whether enough utility and county evidence can be standardized early enough to earn buyer trust at software-like margins; that must be proved in pilots rather than assumed.

Problem

• Teams optioning AI campus land still rely on brokers, consultants, and spreadsheet memos, which makes first-pass diligence slow, inconsistent, and hard to reuse across IC, utilities, lenders, and tenants.
• One false-positive parcel can waste months of option time and consultant spend because power timing, zoning, water, environmental, and tenant-fit failures often surface after exclusivity begins.

Solution

• Provide a memo-first underwriting workspace that captures parcel evidence, applies standardized kill/go checklists, and outputs an investment-grade readiness memo for 100MW+ AI campus sites.
• Start with human-verified confidence scoring for power, zoning, water, environmental, and tenant-fit readiness, then compound learning into a proprietary dataset of why sites fail or advance.

Why we win

• The product is designed around the actual buying artifact, an investment and diligence memo, rather than generic parcel search or mapping.
• Northern Virginia offers a dense proving ground with explicit utility and county workflows, which lets the company build repeatable templates faster than a broad national launch.
• Each screened parcel produces structured failure data that brokers, consultants, and horizontal parcel tools do not systematically retain in a customer-owned workflow.

Milestones

flowchart LR
  Wedge[Northern Virginia 100MW+ site underwriting] --> MVP[Memo first underwriting OS]
  MVP --> Proof[Paid pilots and faster kill go decisions]
  Proof --> Expansion[Option management utility and lender workflows]

Founding team

Experiment roadmap

Risk assessment

What must be true

• Target customers screen enough 100MW+ parcels per year for underwriting software to be a repeated workflow rather than a one-off project.
• VP-level development buyers will approve budget before full engineering if the product materially reduces false-positive site options.
• Northern Virginia pilots can cut first-pass diligence time by at least 50% without hurting decision quality.
• Consultants and advisors will accept exported memo workflows instead of positioning themselves as the only trusted source of diligence.
• Expansion into option management, utility engagement, and lender diligence can at least double account value beyond the initial underwriting wedge.

Open diligence questions

• How many live parcels did each target account screen, option, and abandon in its last two site programs
• What exact memo or evidence package does the investment committee require before approving land spend
• Which utility and county data fields are reliable enough pre-application to support confidence scoring
• Who owns budget, security review, and procurement for pre-development software in these teams
• Do consultants treat the product as throughput infrastructure or as a threat to billable diligence

Model sanity

• Revenue engine. Base-case revenue comes from reaching 15 paying accounts by Q4Y3, with early pilots converting into $210K platform contracts and older accounts expanding toward $300K ARR.
• Must go right. The pilot-to-platform conversion rate has to stay near the BP's 50%+ target so the company can fund expansion without hiring ahead of proof.
• Model breaks if. The biggest risk is a one-quarter slip in the relationship-driven sales cycle, which the sensitivity table shows can erase about $403K of Y3 revenue and push downside cash negative.
• Next-round proof. The next seed round is justified once the team reaches 7 customers, launches Texas templates, and shows the first paid expansion-module revenue on top of the core underwriting workflow.

Scenarios

Sensitivity

flowchart LR
  Leads --> PaidPilots
  PaidPilots --> AnnualContracts
  AnnualContracts --> ExpandedAccounts
  ExpandedAccounts --> Revenue
  Revenue --> GrossProfit
  GrossProfit --> Cash

Flags: Gross margin assumes the human-verified evidence workflow can be standardized to 70%; if services intensity persists, the model is too optimistic on cash. · The customer plan reaches 15 accounts in a narrow buyer universe, so a few slipped logos materially change the outcome. · No second financing round is modeled; the business stays cash-positive through Y3 only because hiring remains deliberately lean and Q4Y3 turns slightly EBITDA-positive.

• Consultant channel resistance. Engineering and diligence consultants may see the product as a threat to billable study work. Mitigation: Position the platform as the front door for earlier screening and package consultants into the workflow for stamped follow-on work.
• Utility data opacity. Developers may not trust automated conclusions if power and interconnection data are incomplete or inconsistent. Mitigation: Start as an evidence and workflow system with human verification, then layer confidence scoring only where the supporting data are explicit.
• Narrow buyer concentration. The initial customer set is small and deal-driven, which can slow revenue growth. Mitigation: Start with land vehicles and mid-market developers on per-site pricing, then expand into lenders, tenants, and broader campus workflows.