800 MW data center
- Contractual CODDec 2029
- Deliverability43%
- P50 CODMar 2030
- Binding345 kV network upgrade
Deliverability intelligence for physical capital
Know what will actually get built.Evidence before confidence.
800 MW campusBefore you invest, finance, insure, build, or contract against physical infrastructure, know what the project actually depends on, and the probability it delivers on time, at cost, and at capacity.
The problemThe layer
Physical infrastructurehas a delivery problem.
We turn fragmented project evidence into a probabilistic view of physical delivery — then translate that risk into financial exposure.
You defineSteps 1 to 3
Objectives
A project can be financed and still fail to deliver.
A project can have land, a sponsor, financing, permits, contracts and a target COD, and still not have the infrastructure required to operate.
Illustrative. Not a published model output.
800 MW data center
An 800 MW data center deciding whether a Dec 2029 COD can actually energize. The run tests the node it sits on, who else is asking for that wire, and how interconnection, equipment, and upgrades move the date.
interconnection.csvequipment.pdf
- 01Find the binding path most likely to slip the COD
- 02Find the date where deliverability starts to fall
- 03Find the dependency that should receive capital first
A1 PowerA2 PermittingA3 Equipment
Q1 What exists?Q2 What does it depend on?Q3 What is likely to happen?
Dependencies
The project is only as strong as its dependencies. Each dependency is a named object on the path.
A node is the wire. An upgrade is shared. A slot has a lead. Atlas carries those objects — and who else sits on them — as conditions of the run, not as a blended score.
A1
Power
- Description
- Interconnection, generation, transmission, substations, transformers.
- Nodes held
- 4,200
A2
Permitting
- Description
- Zoning, environmental approvals, regulatory milestones.
- Nodes held
- 1,140
A3
Equipment
- Description
- OEM, production, shipment, installation, commissioning.
- Nodes held
- 860
A4
Construction
- Description
- EPC, schedule, labor, site progress.
- Nodes held
- 2,080
A5
Utilities
- Description
- Service agreements, upgrades, capacity, connection dates.
- Nodes held
- 640
A6
Offtake / financing
- Description
- Contracts, financing conditions, milestone dependencies.
- Nodes held
- 310
Questions
Questions run against the physical graph. Against a named construct.
Use the formats the desk already speaks. Add a slip, a hypothetical, a binding path, and more. Every follow-up runs against the same construct.
What exists on the path to energization?
- AThe node
- BNetwork upgrade
- CTransformer slot
What does the contractual COD actually depend on?
- APermitting clock
- BOEM lead
- CEPC progress
At which dates would this project still deliver the contracted MW?
- ADec 2029
- BMar 2030
- CAug 2031
In your own words, what would make you hesitate to treat the deck date as a base case?
Two similar campuses, one wire. What decides between them?
- AWho else is asking
- BObserved clearance
- CCapital already spent
Inside AtlasSteps 1 to 6
Atlas starts with evidence. Then it builds the physical world.
Stratum Atlas is a risk intelligence platform for physical infrastructure. It maps the physical dependencies behind infrastructure projects — the campus, the node it sits on, and the load book it is counted in — tracks their state over time, estimates the probability of delivery, and quantifies the financial consequences of delay, shortfall and cost overrun.
The explorer above is a sample of the records Atlas draws on — examples, not the full catalog. Atlas keeps each source at its available grain and cadence rather than smoothing away missing detail. Coverage gaps, irregular updates, and bias remain part of the model's limits.
01 · Discover
Discover
Atlas finds project, asset, utility, regulatory and infrastructure records.
What this holds
From fragmented evidence to a decision grade physical state.
Dependencies move together. A project is not a spreadsheet of independent columns.
Interconnection and equipment are not independent; neither are permits and construction. Atlas models those nodes jointly. Without those relationships, a project can match every one-way total and still energize a hall with no transformer behind it.
d = 2
What it depends on
Edges. A campus sits on a node; the node sits behind a 345 kV upgrade; the upgrade sits behind a transformer slot.
Why preserve it
Pairwise tables miss the three-way joint: this campus, this upgrade, this factory date.
Calibration works at two resolutions. The path, and the book it sits in.
Fit only the project and each path may look plausible while the portfolio totals drift. Fit only the book and the totals can match with no coherent asset underneath them. Atlas holds both: the people and the total they produce.
Predicted and observed mix
Blue is predicted. Ink is observed.
Effect estimated during calibration
Select a trait to remove its effect.
A run does not claim to predict one project's next act. Individual paths contain chance the profile does not observe. Atlas estimates how delivery changes across a population under the stated conditions.
Validation
Internal validation
Atlas withholds evidence from model development. Tests measure whether the system recovers delivery outcomes, preserves differences between projects, and carries learned relationships across time.
Validation
Validation in an engagement
Each engagement starts by naming the outcomes that matter and the performance needed for the decision. Those criteria are set before the results are interpreted. Gaps remain part of the report.
What Atlas measures1 / 3
See what the project actually depends on.
The campus, the node it sits on, the load book it is counted in. If the wire moves, what else moves?
Project example
Ask a different question.
A project can look close on paper and still be far from deliverable. The score stays on the campus. The wire is a different object.
View asTerminalMemoDeckSheet
See how we'd read this fileIllustrative. Not a published model output.
800 MW data center
- Contractual COD
- Dec 2029
- Deliverability Probability
- 43%
- P50 COD
- Mar 2030
- P90 COD
- Aug 2031
- Binding dependency
- 345 kV network upgrade
- Node
- Round Rock 345 kV
Every important number has a trail.
Atlas does not ask you to trust an unexplained score.
Material assessments connect back to the underlying evidence, observation date, source quality, and model version.
See the chain behind the numberWho uses it
Atlas sits with the people who have to decide.
Infrastructure investors
Infrastructure investments are built on schedules, contracts, engineering assumptions and expected capacity. Atlas turns those assumptions into an observable physical state — the campus, the wire it sits on, and who else is asking for that wire — and a probability distribution.
Lenders
Debt repayment depends on physical infrastructure delivering the capacity assumed in the financing case. Atlas models the delivery path — the node, the other asks on that wire, the upgrade they share — and carries physical delay through debt service and downside scenarios.
Insurance
Construction and operational risk often starts with physical dependencies that can be observed before a loss occurs. Atlas provides a structured view of project progress, critical dependencies and delivery uncertainty.
Developers
Your project schedule is a network of dependencies. Atlas names the node each site sits on, who else is asking for that wire, and which constraint should take the next dollar.
Large load
A lease is a claim. A printed load book is a pile. Named megawatts are footprints on a wire. Atlas holds them in separate columns, so a campus is not treated as live capacity and a metro is not treated as a published book.
Built for infrastructure investors, lenders, insurers, developers, and operators.
Ready when you are
Know what will actually get built.
Open the desk, or request access for your institution.