Infrastructure Partnership OS

Make the operating agreement the first machine.

A transaction and execution system for companies combining sites, solar, storage, balance-sheet capacity, AI engineering and real workloads. Structure the bargain, prove the economics, then buy the estate.

Compose a deal Inspect the control room

€6,500: fixed Phase 0 Partnership Blueprint
15 days: target decision cycle after complete inputs
G0 → G6: stage gates from mandate to expansion

REFERENCE TRANSACTION · GATE 0 mandate design

Economic object

Governed AI infrastructure node

Purchase rule: Workload before hardware
Control rule: Rights before relationships
Scale rule: Evidence before debt

AssetCosite · power · equipment

Host + asset schedule

OperatorCoruntime · skills · operations

MSA + licence

House bankterm debt · lease · diligence

Finance documents

Energy partnerPV · storage · metering

EPC / O&M / supply

Technology estatecompute · network · warranty

Purchase / lease

Counsel + taxrights · tax · compliance

Issue opinions

Capital→Assets→Workloads→Evidence→Expansion

01 / OBJECTProductive capability

02 / UNITContracted workload

03 / CONTROLExplicit authority

04 / PROOFReconstructable evidence

What is being built

One economic system across software, buildings and capital.

The opportunity is not “a GPU farm.” It is a governed capability that improves the host company first, creates a reusable operating asset second and becomes externally sellable only when demand and responsibility are legible.

Operating intelligence

Attach to ERP, shop, CRM, documents and facilities without replacing the systems that already run the business.

margin and inventory exceptions
lead-to-offer workflows
owner decision briefs

Physical productivity

Treat buildings, energy, storage, network and compute as one governed productive estate rather than unrelated purchases.

metered site economics
asset acceptance and warranty
capacity allocation

Financeable evidence

Translate the operating programme into a lender-readable sources-and-uses model, implementation plan and proof trail.

baseline and downside case
supplier quotations
approval and contract map

Repeatable expansion

Scale only when one node proves utilization, value, controls and exit mechanics. The second node is an evidence decision.

offtake before expansion
portfolio control plane
operator licensing

Transaction architecture

Five structures. Only one should be first.

The default is a customer-funded productivity node: the operating company buys capability for named workflows. More complex structures enter only when utilization, offtake and governance can carry them.

#	Structure	Economic bargain	Entry evidence	Failure condition
01	Customer-funded productivity node	Implementation fee + managed operations; optional capacity allocation priced separately.	Named internal workloads with accountable owners and a measurable baseline.	The only demand is speculative third-party GPU rental or the buyer cannot name an operating problem.
02	Hosted capacity agreement	Fixed host fee + metered energy + optional upside share above a defined utilization hurdle.	Metered power economics, network resilience, insurance position and committed workload.	The operator expects the property company to underwrite venture demand without minimum commitments.
03	Financed digital transformation	Milestone implementation revenue + recurring operations; debt serviced by company cashflow.	Approved business case, supplier quotations, implementation plan and debt-service headroom.	Property-backed or subsidised debt is being used to finance unbounded product experimentation.
04	Offtake-backed infrastructure node	Reserved capacity + usage + managed services with minimum commitments.	Bankable offtake, clear service boundary, tested unit economics and redundancy plan.	Revenue depends on spot-market compute prices or a single revocable customer.
05	NodeCo / project company	Multi-customer capacity, enterprise services, energy optimisation and asset-management fees.	At least one proven node, diversified demand, audited economics and professional governance.	The parties are using a JV to avoid defining price, ownership, liability or exit rights.

Compile the starting structure Inspect the contract stack

Why each party enters

The bargain must survive six different balance sheets.

Partnership language is cheap. The system becomes real when each participant can state what it contributes, receives, controls, proves and can exit.

Asset and operating company

Turns underused site, roof, energy, data and existing workflows into higher asset productivity—without becoming an AI lab.

Business-owned hardware and data
Margin, inventory and sales intelligence
Bounded implementation and exit

AI operator

Earns fixed underwriting, implementation and managed-service revenue while retaining reusable platform and skill IP.

Anchor customer and reference node
Contracted capacity rather than implied access
Reusable delivery system for the next client

Bank and finance partners

Underwrite a defined operating programme with assets, quotations, implementation controls and observable cashflow—not an AI narrative.

Sources-and-uses by asset life
Accepted baseline and downside case
Clear borrower, security and repayment logic

Execution control

Capital unlocks one gate at a time.

Each gate has a named question and required proof. The next purchase is unavailable until the current evidence has an owner and an acceptance state.

Mandate

Are the parties, objectives, authority and protected boundaries explicit?

Signed Phase 0 scope, data-room protocol, decision owners and exclusions.

Baseline

Can the value pools and operational pain be evidenced from current systems?

ERP/shop/finance extracts, workload logs, energy data and accepted baseline.

Underwriting

Does one structure dominate on economics, control and downside?

Sources-and-uses, sensitivity model, contract matrix and risk register.

Pilot authority

Is the pilot bounded enough to buy, integrate and stop without ambiguity?

Approved SOW, budget ceiling, acceptance tests and human approval matrix.

Shadow operation

Does the system produce reliable recommendations before production write access?

Dual-run results, exceptions, model/tool receipts and operator sign-off.

Production acceptance

Has the node created verified value without violating its authority boundary?

Acceptance report, measured economics, incident review and ownership register.

Expansion

Is additional debt, hardware or another site justified by contracted demand?

Utilization history, offtake, DSCR case, supplier terms and exit plan.

Non-negotiable doctrine

Prevent the expensive category errors.

Most infrastructure failures begin before installation: demand is inferred, roles remain social, asset life is mismatched to finance and automation receives authority it has not earned.

Workload before hardware

A server is not demand. Name the workflow, data owner, service envelope and value baseline before selecting equipment.

Contracts before assumptions

Who owns the machine, software, output, uptime obligation, residual value and failure is written before family trust or enthusiasm is tested.

Evidence before debt

Debt belongs behind accepted business value or contracted offtake—not ahead of an unpriced experiment.

Shared primitives, isolated organisations

Runtime standards can be common while identity, data, memory, budgets, queues and authority remain tenant-specific.

Local where it changes control

Private inference is purchased for confidentiality, latency, resilience or measured cost—not as a symbolic claim of sovereignty.

Rights decay with exposure

Guarantees, priorities, reserved capacity and special approvals expire when the corresponding risk or capital exposure ends.

Reference architecture

Attach to the existing company before replacing anything.

Read-first connectors map ERP, shop, CRM, identity, documents, finance preparation and facilities. Agents operate in shadow mode; production write-back is earned by accepted evidence.

Physical estate

Warehouse, power, PV, storage, meter, network, access, cooling, equipment, warranty and insurance become one asset graph.

serialised asset register
energy and capacity ledger
maintenance and incident evidence

Intelligence estate

Persistent workers, deterministic workflows, local inference and metered cloud models are routed by workload and authority.

tenant-separated identity and memory
model and tool version receipts
reversible automation first

Control estate

Contracts, policies, approvals, SLAs and evidence receipts determine what the system may observe, recommend, prepare or execute.

human approval for consequential action
append-only decision history
portable data and explicit exit

Inspect the system and skills See the operator model

Reference frameworks: NIST AI RMF, NIST Generative AI Profile, and NIST Zero Trust Architecture.

Phase 0 · fixed decision product

Partnership Blueprint

Commission the underwriting before the machines: asset and workload map, IT integration design, three viable deal structures, contract matrix, sources-and-uses model, procurement sequence, risk register and 90-day pilot mandate.

Professional fee: €6,500 ex VAT
Target cycle: 15 business days after complete inputs
Purchase exposure: No hardware, debt or production access included
Acceptance: Decision pack + executive review

Commission the Blueprint Review boundaries first