Control tower for AI campus teams buying and commissioning behind-the-meter generation while utility power is delayed.

1. The category now has institutional balance-sheet backing, which means customers will commit to large bridge-power programs rather than treat them as emergency experiments.
2. Data-center deployments are explicitly called out in the sources, so the buyer already has a named use case and budget owner instead of a hypothetical future market.
3. Propell’s manufacturing expansion shows the next choke point is not invention but matching scarce factory output to site readiness and delivery sequencing.
4. A 7.5 GW order book and explicit execution-risk language mean software that removes schedule slippage can defend budget by protecting revenue, not just saving admin time.

Catalyst. VoltaGrid’s financing, Propell acquisition, and explicit data-center focus show that deployable generation capacity is scaling fast, which shifts the urgent customer pain to procurement and delivery coordination.

Build a software system for owners and EPCs that treats deployable generation like a managed supply chain instead of a generic construction project. The product creates a digital bill of megawatts for each site, ties equipment slots to civil, electrical, and permitting prerequisites, and surfaces which dependency is most likely to delay first power. It also standardizes factory acceptance, shipping, site acceptance, and black-start testing across vendors so teams can compare real progress against the promised energization date. Over time, the company builds the best benchmark dataset on how modular generation projects slip, which vendors hit schedule, and which design choices shorten time to revenue.

What's different. Generic project-management tools do not understand that a single missed controls cabinet, gas-conditioning skid, or FAT signoff can strand tens of megawatts of AI capacity. Traditional microgrid vendors sell hardware or engineering services, but they rarely provide a neutral control tower across OEMs, EPCs, site owners, and financiers. This company wins by owning the schedule-risk dataset for bridge-power deployments and turning it into benchmarking, financing, and insurer visibility that hardware vendors cannot easily replicate.

Jobs to be done

flowchart LR
  Buyer[AI campus developer] --> Pain[Utility delay threatens first power]
  Pain --> Product[Bridge-power commissioning OS]
  Product --> Outcome[Generator fleet lands on time]

Executive takeaways

• Bridge power for AI campuses has shifted from emergency fallback to scheduled primary infrastructure because utility timelines now frequently exceed construction timelines and on-site generation adoption is accelerating [1][3][6][10][11][12].
• The real bottleneck is execution certainty, not invention: VoltaGrid, Propell, Oracle, and Vantage examples all point to manufacturing slots, integration capacity, fuel readiness, and commissioning sequence as the scarce assets [1][2][3][4].
• Buyers already use Oracle, Procore, InEight, and consultant-led war rooms, but those tools are horizontal and do not natively model generator blocks, FAT/SAT milestones, permitting gates, and first-power dependencies across multiple vendors [23][29][30][31][32][33][34][35][39].
• Regulation is not peripheral. EPA engine rules, state air permits, and large-load/co-location processes in ERCOT and PJM mean any credible product needs audit-ready evidence, not just prettier task lists [19][20][21][22][23][24][25][26][27][28].
• The beachhead is narrow but economically meaningful: if the startup becomes the neutral system of record for 15-20 live bridge-power programs, it can expand into permanent microgrids, lender reporting, and insurer-grade execution benchmarking [7][8][12][17][18][36].

Market definition

This market is workflow software for behind-the-meter bridge-power delivery on AI campuses: reservation of generation blocks and balance-of-plant equipment, site-readiness gating, factory and site acceptance, fuel and permitting readiness, and executive risk reporting for 20-100 MW temporary-to-permanent power packages. It excludes generic construction PM, pure DERMS, and hardware-only generation contracts [1][3][10][11][12][29][30][33].

Customer and buyer

Primary user is the owner-side power-delivery or development program lead, often alongside the owner’s engineer running a live bridge-power schedule. The economic buyer is usually the VP Power Delivery, COO, or Chief Development Officer at a campus developer whose tenant revenue or financing depends on a first-power date. Secondary users include EPC program-controls teams, commissioning leads, and lenders or insurers that need a defensible schedule of record [4][10][11][28][30][31][36].

Buying triggers

• Utility interconnection slips beyond the campus build schedule, turning first power into a board-level delivery risk. [6][10][12][13]
• A tenant contract or financing close depends on bridge-power energization before permanent utility service arrives. [1][3][4][26]
• The developer commits to on-site generation or non-firm service, creating new permitting, fuel, and commissioning workstreams. [11][19][23][25][36]

Willingness to pay

Willingness to pay should be high because bridge-power programs protect tenant revenue, avoid stranded campus capital, and already require heavy spend on consultants, permitting, and complex program controls. Even the horizontal substitutes in this workflow are sold as enterprise custom-quote systems rather than lightweight seat tools, so a domain-specific control tower can plausibly command six-figure ACVs on live programs. [3][6][10][11][24][30][31][36]

Category dynamics

Growth signal 13.1% low-case annualized U.S. data-center electricity-demand growth (2023-2028)

Validation signals

• VoltaGrid’s $1B raise plus Propell acquisition show investors now treat deployable power for AI campuses as core infrastructure.
• Oracle’s 2.3 GW deployment and Vantage’s 1 GW partnership show live, multi-gigawatt demand for modular behind-the-meter power.
• 92% of surveyed industry leaders cite grid constraints as the top obstacle and 44% report utility waits exceeding four years.
• Bloom survey coverage indicates power access is now the top siting driver and primary on-site generation is expected to rise from 13% to 38% by decade end.
• ERCOT and PJM are formalizing large-load and co-location rules, confirming bridge power is now a planning-regulated workflow rather than an improvised workaround.

Regulatory & technical constraints

• Stationary engines used for bridge power can trigger EPA NSPS and NESHAP compliance, including electronic reporting requirements.
• Texas and similar state pathways may permit some stationary assets faster, but runtime assumptions and recordkeeping still materially affect feasibility.
• ERCOT requires formal large-load and proposed net-metering documentation for very large loads, which makes ad hoc planning risky.
• PJM’s treatment of co-located and behind-the-meter large loads is changing, so bridge-power operating assumptions may need to be reworked as rules evolve.
• Generic construction suites do not natively model FAT, SAT, energization, and fuel-readiness dependencies across multiple power vendors.

The competition is fragmented across horizontal project-control suites, consultant-led coordination, and power vendors that bundle their own milestone views. Oracle, Procore, and InEight can hold schedules, costs, and documents; VoltaGrid and Enchanted Rock can bundle delivered power and operational dashboards; and advisors can stitch together permitting and utility evidence. The gap is a neutral, megawatt-block system of record built specifically for cross-vendor bridge-power delivery risk and first-power readiness [1][2][29][30][31][32][33][34][35][36][39].

Why incumbents do not win by default

• Hardware and power-service vendors. VoltaGrid and Enchanted Rock win when the buyer wants delivered generation capacity, but they do not win by default as neutral workflow software because they are economically aligned to their own assets, contracts, and preferred execution paths.
• Horizontal construction suites. Oracle and Procore are credible systems of record for generic capital projects, yet their abstractions stop at schedule, cost, and document control and do not encode engine-level commissioning or power-package dependencies.
• Project-controls specialists. InEight gets closer on schedule rigor and capital-project controls, but it still sells a horizontal platform where bridge-power playbooks must be configured rather than natively modeled.
• Consultants and owner’s engineers. Specialist advisors remain sticky because bridge-power programs cross permitting, utility, and commissioning disciplines, but services alone do not compound into benchmark data or reusable risk models.

This company should start as a bridge-power commissioning OS for Texas and ERCOT-led AI campus developers bringing 20-80 MW of behind-the-meter reciprocating generation online before utility service is ready. The buyer pain is operational, not conceptual: first power still depends on spreadsheets, OEM portals, consultant war rooms, and generic project controls even though a missed FAT, permitting gate, fuel-readiness item, or controls shipment can delay tenant revenue by months. The first product should be a neutral system of record that ties generator blocks, balance-of-plant dependencies, air-permit and large-load gates, FAT and SAT evidence, and executive risk reporting into one campus-level control tower. Go-to-market should stay narrow and coherent: sell a paid live-project pilot to the VP Power Delivery or COO right after a tenant commitment or financing close makes a first-power date board-critical, price by active campus and MW band, and win distribution through owner’s engineers and permitting advisors already inside the same delivery loop. The strategic reason to start in ERCOT is not market size alone but repeatability: Texas has visible bridge-power activity, explicit large-load process rules, and a permit path that can be productized sooner than PJM’s moving co-location regime. The moat is the cross-project dataset on which vendors, milestones, and sequencing choices actually predict first-power slippage and first-pass commissioning success. The main disconfirming risk is that projects remain too bespoke and data too fragmented for software deployment to stay lighter than consultant-led coordination. Exact delivered cost per MW, customer concentration by developer, and how much milestone data can be pulled automatically from OEM and EPC systems remain missing, so the first year must prove both repeatable integrations and pilot-to-production conversion.

Problem

• AI campus developers using bridge power still coordinate generator reservation, site readiness, permitting, fuel planning, FAT, SAT, and first synchronization through spreadsheets, email, OEM portals, and weekly war rooms, so no one has a trustworthy view of which dependency will delay first power.
• Horizontal tools such as Primavera, Procore, and InEight can track generic schedule and document status, but they do not natively model megawatt-block dependencies, cross-vendor commissioning gates, or audit-ready regulatory evidence for behind-the-meter generation.

Solution

• Provide a live-project control tower that creates a digital bill of megawatts for each campus and ties every generator block, switchgear package, controls cabinet, fuel-readiness check, permit gate, FAT milestone, and SAT milestone to the promised energization date.
• Start with evidence capture, critical-path risk scoring, and executive readiness reporting for bridge-power programs, then expand into benchmark analytics, lender and insurer reporting, and bridge-to-grid handoff once the core workflow is trusted.

Why we win

• The company is neutral across OEMs, EPCs, owner teams, and financiers, which matters because hardware vendors and EPCs are aligned to their own assets and cannot credibly be the system of record for cross-vendor delivery risk.
• The wedge sits inside an urgent, measurable workflow where a schedule slip delays tenant revenue and financing milestones, making the ROI easier to defend than generic construction productivity software.
• Each live project compounds proprietary data on slot reservations, FAT and SAT outcomes, permit delays, and vendor performance that horizontal PM suites and consultant war rooms do not naturally retain in reusable form.

Milestones

flowchart LR
  Wedge[ERCOT bridge-power wedge] --> MVP[Neutral commissioning control tower]
  MVP --> Proof[On-time readiness proof and benchmark data]
  Proof --> Expansion[Multi-campus, lender, and microgrid expansion]

Founding team

Experiment roadmap

Risk assessment

What must be true

• At least 15-20 beachhead campuses per year face first-power risk severe enough to fund a standalone control layer.
• A standard milestone schema for 20-80 MW reciprocating-engine programs covers most early pilots without heavy custom services.
• VP Power Delivery or COO buyers will buy a neutral overlay without requiring replacement of Primavera, Procore, or OEM portals.
• Pilot customers will permit the company to retain anonymized benchmark data on milestone slips, vendor performance, and commissioning outcomes.
• Lender, insurer, or executive reporting needs create expansion revenue beyond the initial campus workflow.

Open diligence questions

• Which developers and owner’s engineers actually control budget when the trigger is a utility delay plus signed tenant demand?
• How many live bridge-power programs in the 20-80 MW band does a typical beachhead account run in a two-year period?
• What exact milestone data can be pulled automatically from OEM, EPC, and commissioning systems versus entered manually?
• Why will Primavera, Procore, InEight, or consultant-led controls teams not satisfy the first customer well enough?
• How stable is the ERCOT-first thesis if permitting or large-load rules tighten faster than expected?

Model sanity

• Revenue engine. The base case is driven by growing from 3 paying campuses at Y1 exit to 20 at Y3 exit on a $360K blended annual campus value.
• Must go right. The first three pilots must convert into repeatable multi-campus rollouts before implementation work behaves like standardized software rather than consulting.
• Model breaks if. If sales cycles stretch toward nine months or gross margin stays below the high-60s, the downside case points to another raise before the next proof point.
• Next-round proof. The next financing is justified by reaching 8-10 production campuses, showing at least one in-account expansion, and proving benchmark or reporting modules lift ACV.

Scenarios

Sensitivity

flowchart LR
  Leads --> PaidPilots
  PaidPilots --> ActiveCampuses
  ActiveCampuses --> Revenue
  Revenue --> GrossProfit
  GrossProfit --> Cash

Flags: The model assumes a very concentrated buyer universe, so a small number of delayed campuses can move revenue materially. · Revenue per FTE sits above generic SaaS benchmarks by Y3, which only works if implementations keep standardizing instead of turning into a services arm. · Unit economics use heuristic CAC and churn because the company has no observed cohort history yet. · Gross margin reaching the low 70s depends on export-based ingest and reusable ERCOT compliance templates staying sufficient for most deployments.

• Projects stay too bespoke. Early bridge-power programs may differ so much by site and vendor mix that software standardization is hard. Mitigation: Start with reciprocating-generation packages in a narrow MW band and encode the common FAT, SAT, and site-readiness milestones before expanding.
• OEMs bundle lightweight tooling. Generator packagers or EPCs may add basic milestone dashboards and reduce urgency to buy standalone software. Mitigation: Win on neutral cross-vendor coordination and lender-grade risk reporting that single-vendor tools cannot provide credibly.
• Utility delays ease faster than expected. If grid upgrades accelerate, some developers may choose to wait instead of funding bridge-power projects. Mitigation: Expand quickly into permanent microgrid and resilience deployments where the same commissioning and delivery-control workflow still matters.