Five wedges into one predevelopment operating system.

The data center industry has invested heavily in operational technology. DCIM platforms track every rack. SCADA systems run every facility. Cooling optimization algorithms save hyperscalers hundreds of millions of dollars per year. The construction side is increasingly automated too — design, procurement, and project management all have substantial software stacks.

The predevelopment phase — site selection, interconnection filings, environmental review, permitting, incentive applications, regulatory compliance — is the one phase of data center construction where almost nothing has been built. It is still coordinated through spreadsheets, email, and law firms billing $1,000–$2,200 per hour. The result is that the predevelopment phase, which determines whether a project happens at all, is the slowest, most expensive, and most error-prone part of the work.

The consequence shows up in the numbers. $64 billion in announced data center projects have been canceled or delayed since 2023, according to S&P Global, with the proximate causes overwhelmingly being permitting battles, interconnection delays, and regulatory uncertainty. The cost of these delays is roughly $14.2 million per month per 60 MW project, per our own modeling. For a 300 MW campus, the delay cost rate hits $70 million per month.

The absence of tooling is not because the work is too hard to automate. It is because the incumbents who profit from the complexity — law firms, consultants, engineering firms — have no incentive to build the tools that would compress their billable hours. When your business model depends on charging $1,500 per hour to move PDFs, software that moves the PDFs for you is not an investment. It is a threat.

Cliffcenter is built by the team behind Cliffwide, an AI-native platform for affordable housing bond applications. Cliffwide's core technology is an AI form-mapping engine that ingests jurisdiction-specific application forms, maps them against a canonical project schema, and generates filings that pass audit review at the first submission.

The engine was built for affordable housing because affordable housing bond applications are the most regulation-heavy, form-heavy workflow in US real estate. LIHTC applications, tax-exempt bond issuances, state housing agency programs — each has its own form, its own compliance schedule, and its own evaluation framework. A successful housing bond application is a coordination problem: collect the right inputs, fill in the right forms, meet the right deadlines, and pass the right audits.

When we looked at data center predevelopment, we saw the same structure. Interconnection filings are forms. Incentive applications are forms. FERC ISA amendments are forms. Environmental permits are forms. NRC license amendments are forms. Every category is a form-mapping problem over a canonical project schema. The engine that already ships a large fraction of Cliffwide bond applications without a human in the loop is the same engine that could ship data center predevelopment filings.

We start with interconnection because it is the highest-leverage pain point. 2,600 GW of capacity sits in US queues. The median time from interconnection request to commercial operation is 5 years nationally and 9.2 years at CAISO. Every data center developer with a large-load request is navigating an ISO-specific process that has no standardized framework across the industry.

The product is a portfolio-level view of every active large-load interconnection across every US ISO and major utility. It encodes the process differences — cluster windows, readiness requirements, withdrawal penalties — so a project that moves across ISOs does not start from a blank form each time. It pre-generates submittal packages, tracks every deadline, and flags upstream upgrade risks before they appear in a study.

Why it works as a wedge: the product is self-contained. A developer can buy Queue Intelligence for a single project, for a single ISO, and see immediate value. No platform buy-in required. The sales motion is short. The competitive landscape is open because every existing interconnection tool serves the utility or ISO side, not the developer side.

Our current pilot pricing is $4,500 per project per month at the starter tier, with a developer tier at $18,000 per month and an enterprise tier for custom work. We launched ERCOT, PJM, and MISO first because those three cover the highest queue volume. NYISO and SPP shipped next.

Wedge two is the product most directly leveraged by our existing technology. Every state has data center incentive programs — sales tax exemptions, property tax abatements, job creation credits, workforce grants, infrastructure grants. The top fifteen states distribute $3–4 billion per year through these programs. Virginia's DCRSUT alone waived $732 million in sales tax in the last JLARC audit.

Incentive applications are form-mapping problems on canonical project data. The customer has an investment, a MW, a COD, a job count, a wage profile, and a candidate county shortlist. The atlas maps that project to every eligible program across every jurisdiction, projects the net-present-value of capture, and auto-prepares the application package.

This is our fastest wedge to first customer. The Cliffwide engine ports cleanly. The market is underserved — economic development consulting is a relationship-driven industry where firms like CBRE and JLL charge $100,000 to $500,000 per engagement and deliver in slides. Our Atlas tier is $2,200 per month; our filings tier is $15,000 per filing flat-fee, with no success fees. We undercut the incumbents by 50 to 70 percent, ship in a fraction of the time, and leave the customer with source artifacts they can reuse.

The state briefs we publish — Virginia, Texas, Georgia, Arizona, Ohio, Tennessee, Iowa — are the public-facing proof that we know the programs. Each brief covers the incentive structure, the active pressure points, the eligibility thresholds, the clawback exposure, and the specific playbook we use to file in that state.

Wedge three is the highest-revenue-per-customer wedge, and the hardest to execute. Behind-the-meter deal structuring is the workflow that produced the Talen-AWS restructure — 18 months of legal work, billions in deal value, and no software because the law firms billing $1,000 to $2,200 per hour have no structural reason to automate.

The FERC December 2025 co-location order turned this wedge from a theoretical market into an urgent one. Forty-six existing BTM arrangements must re-file under the new framework. Every new deal lands in rules that are still being written. The paper hearing reply round closes April 17, 2026. The transition deadline runs through December 2028.

BTM Workflow encodes the regulatory graph as a live data structure. It tracks every filing in the co-location proceeding, every ruling, every stakeholder comment. It generates Gantt charts and workplans from the graph. It supports deal rooms with redlining, precedent search, and regulator-facing exhibits. It integrates with Queue Intelligence so that co-location and transmission assumptions stay in sync.

Our pricing is $75,000 per deal, $420,000 per year for a Program plan, and custom for Enterprise. The customer is the power company or hyperscaler structuring the deal — Constellation, Vistra, Talen, Microsoft, Meta, Google, Amazon, or the mid-tier law firm bidding for work the AmLaw 100 refuses to staff at the customer's price point.

Wedge four is the data moat. Approximately 120 US coal plants are scheduled to retire in the next five years. Each is a $1 to $10 billion conversion opportunity with existing grid interconnections. Homer City is the template: a 1,884 MW retired coal plant in Pennsylvania that is being converted into a $10 billion gas-plus-data-center campus.

The problem is that the data to identify viable sites lives in ten different databases that do not talk to each other. EIA Form 860 for retirement schedules. NRC licensing data for nuclear sites. ISO/RTO queue feeds for interconnection capacity. EPA ECHO for environmental assessments. County GIS layers for zoning. USGS for water availability. LandGate covers energy broadly and did a useful BTM white paper in April 2026. Transect handles environmental due diligence. Cushman & Wakefield's Athena is an internal platform only available to CBRE's clients. None integrates the full set of data layers.

Site Intelligence is the index we wished existed when we started modeling the wedge thesis. Every retired coal plant, every decommissioning nuclear site, every brownfield with heavy-industrial zoning, scored on a 0 to 100 fit metric that blends POI voltage, interconnection headroom, remediation status, zoning, and local ordinance risk. The three public indices — coal, nuclear, and brownfield — are samples of the full 800+ site database.

The customers are acquirers and aggregators: private equity building predevelopment pipelines, power companies with underused assets, hyperscaler land teams. Our Browse tier is $3,500 per user per month. Our Diligence tier is $14,000 per month with unlimited workspaces. Enterprise is custom.

Wedge five is the services-as-software layer. The data center predevelopment consulting market is $7.73 billion today and projected to grow to $12.4 billion by 2035, per IMARC. Traditional consultants at CBRE, JLL, and Newmark charge $250 to $500 per hour and deliver in PowerPoint. Our AI-augmented team delivers the same outcomes at 50 to 70 percent of the cost, three to five times the speed, and every deliverable is a structured artifact the customer can keep and reuse.

This wedge exists for two reasons. First, most of the customer base for wedges one through four cannot wait for the full platform to mature. They need the outcome this quarter. Services lets us ship the outcome today while the software catches up. Second, every services engagement becomes training data for the next automation. Scale AI started with humans labeling data by hand and productized the workflow over time. We do the same thing for predevelopment.

Pricing is flat-fee by engagement type. Site memos are $25,000 for 10 working days. Full predevelopment engagements start at $125,000 and cover end-to-end workflow through ISA and break ground. Embedded engagements are custom. Every engagement produces artifacts in the customer's Cliffcenter workspace, so when the engagement ends, the customer has a working instance of the software running in their environment.

The five wedges are ordered by time to first customer, not by revenue potential. Incentive Optimization is the fastest because the Cliffwide engine ports most directly. Queue Intelligence is second because the pain is most acute and the competitive landscape is most open. BTM Workflow, Site Intelligence, and Predevelopment Services follow because each requires more domain investment or more technology investment than the prior two.

Each wedge is defensible on its own. You do not need to buy the full platform to buy Incentive Optimization. You do not need to buy Queue Intelligence to buy BTM Workflow. The wedges integrate — data flows between them, customers expand naturally from one to the next — but the initial sale is unit-economic on a single wedge.

Over time, the shared regulatory graph becomes the product moat. A queue filing and a BTM deal and an incentive application and a site diligence memo all touch the same graph of regulatory objects. An operator who has Cliffcenter workspaces for three of the five wedges gets a fourth for marginal cost. The platform effect compounds on top of the wedge sales.

The obvious risk is that a well-funded adjacent vendor — PermitFlow (Y Combinator; $54M Series B in December 2025), GridUnity, LandGate, or CBRE — decides to build a data-center-specific workflow on top of their existing business. We assume this will happen. Our defense is velocity: we ship every week, we publish our research, and we build customer relationships that compound faster than a pivot would take.

The second risk is that the big law firms build internal tools that preserve their billable surface. Latham & Watkins has a ~50-lawyer data center group. Baker Botts brands itself as 'the market leader in the data center ecosystem.' Neither has shown a structural willingness to cannibalize its own billable hours, but both could. Our defense is that the mid-tier firms — the ones losing deals to Latham because they cannot staff them — adopt us as a force multiplier.

The third risk is that the regulatory landscape stabilizes, reducing the velocity advantage of a dedicated software company. We believe this is unlikely on a 3-year horizon. The regulatory rewrite is too deep, the political pressure is too high, and the industry growth is too fast for the rules to calcify. On a 10-year horizon, if the landscape does stabilize, the product naturally evolves from a regulatory intelligence platform into a transactional platform that handles the stable workflow.

The fourth risk is that hyperscalers build internal tools and refuse to buy software. This is the least concerning risk, because the hyperscalers are not the only customers. The 'next 200' developers — crypto-to-AI converters, regional colos, coal plant converters, enterprise builders, tribal land developers — collectively represent more new capacity than the hyperscalers, and they cannot afford internal platforms.

We publish this thesis for three reasons. First, to explain our shape to prospective customers and investors who want to understand why we are building five products instead of one. The short answer is that no single product can earn the right to become the platform, and trying to ship the platform first is a path to shipping nothing.

Second, to invite correction. The research desk treats every claim in this brief as revisable if a customer or competitor shows us better data. The wedge thesis is a hypothesis that we will keep updating as we learn. The April 2026 revision reflects the post-FERC-order landscape and a year of additional market data.

Third, to recruit. Every role at Cliffcenter is built around a wedge or a cross-cutting capability that serves multiple wedges. If you want to help write the next version of this thesis with actual customer data behind it, the open roles are published at /careers.

The predevelopment market will be built by somebody. We think we have the shortest path from existing technology to a platform the industry needs. If we are right, the next eighteen months will validate each wedge one by one. If we are wrong, we will publish that too, revise the thesis, and keep shipping.