The Project Was Funded. The Debt Did Not Move: What $18 Billion of AI Loans Reveals About Distribution Risk

A loan can be legally valid and fully funded, with signed financing documents and construction already under way, while its intended distribution remains incomplete. The borrower has access to capital, but the debt may not yet have reached the investors expected to hold it. Funding and the placement of risk therefore need separate records. Project Jupiter, the large New Mexico data-center development tied to Oracle's AI infrastructure build-out and its relationship with OpenAI, makes that distinction visible.

In November 2025, Reuters reported that a consortium of around 20 banks was providing approximately $18 billion of project-finance debt for the development. The lead banks intended to syndicate the exposure more broadly to other banks and institutional investors. Reuters: Banks lend $18 billion for Oracle-tied data center project By 18 September 2026, the financing had become a different story.

Financial Times reporting cited by Reuters said syndicate banks including Santander and Jefferies were quoting the loans at approximately 89 to 91 cents on the dollar. Attempts to sell the debt to a wider investor base had stalled. Banks were therefore holding more Oracle-linked project debt on their own balance sheets than originally planned. Reuters: Oracle's $18 billion data center debt under pressure, FT reports

The headline is about AI infrastructure. The more reusable private-market lesson is about distribution risk.

A funded loan is not necessarily a distributed loan.

That matters because private debt does not stop having a lifecycle when the money is advanced. For many large financings, funding is only one stage.

What Project Jupiter shows about funding and distribution

The simplest lending model looks like this:

Lender → Loan → Borrower.

In that model, the same lender underwrites the borrower, advances the money and holds the resulting exposure. Large syndicated and project-finance transactions can look very different:

Borrower / project → Arranging banks → Committed financing → Funding → Syndication → Allocations → Transfers → Final holders.

The same loan can pass through several institutional balance sheets. The financing can be fully available to the borrower before the arranger has reached its preferred final hold position. That means two statements can both be true: Project financing: Funded and Loan distribution: Incomplete. A serious asset record needs both.

Reuters reported in November 2025 that around 20 banks were involved in the financing. The reported structure was approximately $18 billion. The loan was expected to have a four-year maturity with two possible one-year extensions. Reported pricing was around 250 basis points over SOFR. Lead banks planned to distribute the debt to other banks and institutional investors. Those details describe the financing at origination. By September 2026, the relevant questions had changed.

The market now needed to know:

• how much debt had actually been distributed;
• which banks still held exposure;
• what amount each arranger intended to retain;
• what amount remained available for placement;
• what price potential investors would require;
• whether buyers were willing to participate;
• what credit and project information had changed since origination; and
• whether a transfer could actually settle.

Those are post-closing questions. They belong in the loan lifecycle.

“$18 billion raised” does not describe who owns the risk

A financing announcement often says $18bn financing secured. That tells the market something important. It says funding capacity exists. It does not necessarily say: $18bn successfully placed with final investors. The original bank group may still hold large positions. Some exposures may have been sold. Some may remain with lead arrangers. Some investors may have declined allocations. Some positions may be offered at a discount.

Some transfers may still be pending. A single financing headline cannot describe that distribution state.

Track the arranger's intended hold and actual exposure

Once debt begins to move between institutions, the asset record should answer: Who originally committed? Who funded? Who was allocated the loan at closing? Who was expected to retain it? Who was expected to distribute it? Who owns it now? Which interests are being marketed? Which transfers have actually settled? The answer can change over time. That makes holder history part of the asset.

A bank can underwrite a financing with an intended strategy:

Underwrite → Fund → Distribute → Retain target hold.

If distribution fails, the actual path can become:

Underwrite → Fund → Attempt distribution → Investor demand weakens → More debt remains with arranger.

The legal loan may be unchanged. The bank’s balance-sheet position is not. The distinction matters for concentration, capital consumption, liquidity, risk limits, hedging, mark-to-market analysis, further underwriting capacity and eventual sale strategy. Distribution risk is therefore not merely a sales problem. It can become a credit-management and balance-sheet problem.

Funded loan

Describes money advanced and the borrower's resulting obligation. It does not establish that arranging banks reached their intended final hold.

Distributed exposure

Requires evidence of allocations, transfer documentation, any required consents, settlement and the current holder record. Investor interest alone is insufficient.

Keep quotes, realised prices and credit states distinct

The September 2026 reporting introduced another number: 89–91 cents on the dollar. That number should not replace: Face value and it should not automatically become Realised sale price. Reuters described the figure as quotes from syndicate banks based on Financial Times reporting. That distinction matters. A private-loan record may need several price fields:

• Original issue / funding price
• Par amount
• Current carrying value
• Indicative bid
• Indicative offer
• Executed trade price
• Settlement value

These are different facts.

89 cents is not automatically a realised 11% loss

Suppose a bank holds a loan with a face value of $100m and an indicative market quote is 89. It would be wrong to conclude automatically: Realised loss: $11m. The bank may not have sold the asset. Accounting treatment may depend on classification, valuation methodology and applicable standards. The loan may later trade at another price. The borrower may repay at par. The position may be transferred through another mechanism.

The quote is economically important. It is not the same thing as a completed disposal.

A useful private-market price record should contain:

• asset or facility;
• tranche;
• currency;
• quoted price;
• bid / offer status;
• quoting party;
• timestamp;
• size;
• market context;
• whether the quote is firm or indicative;
• whether a transaction followed;
• executed price if applicable; and
• settlement status.

Without that provenance: Price: 89 can become misleading. For illiquid loans, price quality is itself an asset-data problem.

A loan’s market price can change even before a payment default. Investors may reassess borrower credit, sponsor credit, tenant credit, construction risk, completion timing, project economics, refinancing risk, regulatory risk, concentration, market supply of similar debt, interest-rate expectations, required return or balance-sheet capacity. Therefore: Loan payment status and Loan market value must be separate fields. A performing asset can still be quoted below par.

Separate project, corporate and permitting evidence

Project Jupiter has faced public and regulatory controversy in New Mexico. Reuters reported concerns relating to local opposition, water and air-quality issues, and power infrastructure. The Financial Times reported project delays and financing pressure. But Oracle has publicly disputed interpretations suggesting that the relevant permitting dispute stops construction of the data-center campus. On 14 September 2026, Oracle said the data-center campus and proposed microgrid are separate facilities.

According to Oracle, the court stay concerns the microgrid air-permit proceeding and does not concern the county permits under which data-center buildings are being constructed. Oracle: Project Jupiter Statement on Construction Permitting That disagreement is itself a proposed useful evidence record lesson. A credit record should distinguish: Reported project risk from: Company response from: Regulatory order from: Final legal outcome.

A system should preserve the competing source-backed states rather than collapse them into a conclusion.

Large infrastructure financings often involve several parties. Project Jupiter is described as a data-center campus linked to Oracle’s broader arrangement with OpenAI. Oracle is expected to use or lease capacity. The project itself has its own development entity and physical infrastructure. The financing may sit at a project or special-purpose-vehicle level. The credit analysis therefore should not collapse everything into: Oracle loan. A more useful graph can be:

Project company → Data-center campus → Construction contracts → Power arrangements → Tenant / customer arrangements → Oracle-related payment exposure → Project-finance debt → Lenders.

Each relationship requires evidence.

Oracle’s own credit profile also changed. S&P Global Ratings downgraded Oracle’s long-term issuer credit rating from BBB to BBB- on 9 July 2026, leaving it one notch above speculative grade. Wall Street Journal: Oracle Gets Credit Rating Downgrade That corporate rating does not automatically determine the value or enforceability of the Project Jupiter loans. But it can affect investor appetite if Oracle-related economics are important to the project.

A structured record therefore needs to distinguish:

• Project borrower credit
• Tenant / offtaker credit
• Sponsor credit
• Guarantor credit
• Corporate rating
• Facility-level market price

These are connected risk indicators. They are not interchangeable.

A field such as Oracle rating: BBB- needs:

• rating agency;
• rating type;
• effective date;
• prior rating;
• outlook;
• affected entity;
• source; and
• later changes.

The same applies to project-level ratings where they exist. Asset state should be time-aware.

Follow distribution through documentation and settlement

“Syndication” is often treated like a checkbox: Syndicated: Yes. That is too crude. A distribution process can contain:

Underwriting → Initial bank commitments → Allocation strategy → Investor marketing → Orders / indications → Allocation decisions → Assignment documentation → Transfer approval → Settlement.

At each stage, different amounts can exist. For example:

Record item	Value or state
Total facility	$18bn
Initial underwriting	$18bn
Target arranger hold	X
Target distribution	Y
Investor indications	Z
Allocated for transfer	A
Legally assigned	B
Settled	C

A single boolean loses the entire lifecycle.

Suppose an investor agrees to buy part of a loan. The lifecycle may still require assignment documentation, borrower or agent procedures, eligibility checks, KYC / AML, transfer fees, consent where required, settlement instructions, cash movement and lender-register update. Until the transfer completes, the existing holder may still carry the exposure. A private-market system therefore needs to distinguish:

• Investor interest
• Allocation
• Executed transfer document
• Settled transfer
• Authoritative lender record updated

That is the debt equivalent of trade-versus-settlement separation.

This matters because private loans do not move like exchange-traded equities. A sale can be economically agreed before legal ownership moves. A transfer can be documented before cash settles. Cash can settle before every internal risk system updates. The agent’s lender register may become the authoritative record only after defined steps. That creates a lifecycle:

Trade agreed → Transfer documents → Conditions satisfied → Cash settlement → Agent record updated → New holder recognised.

The authoritative current holder should not be inferred from the earliest step.

Project Jupiter also shows how quickly a newly originated loan can develop secondary-market information. The financing begins in the primary market:

• Commitment
• Funding
• Syndication

Then the same loan can enter a secondary context:

• Bid
• Offer
• Discount
• Transfer
• Settlement

A proposed robust evidence record should not create a new unrelated asset simply because the transaction moves from primary to secondary distribution. The same loan identity should continue.

Reconcile values, dates and institution-specific positions

A single loan can simultaneously have:

• Original commitment
• Funded principal
• Outstanding principal
• Accrued interest
• Accounting carrying value
• Indicative market price
• Executed sale price
• Recovery value

These values answer different questions. A platform that stores them in one generic field called: Value will eventually produce errors.

Private-debt valuation can move quickly. For every price or balance, the record should answer: Value as of when? An 89 quote on 18 September 2026 should not silently become today’s value months later. A time-aware system needs value, value type, currency, as-of timestamp, source, methodology, confidence or status, and superseding value.

This is one of the strongest lessons from the Project Jupiter story. At origination, a bank may intend to hold $500m while underwriting $2bn with the expectation that $1.5bn will be distributed. If only $700m is sold, the actual retained exposure is different from the planned hold. That should not require reconstructing old emails months later. A structured financing record can store:

• Underwritten amount
• Target final hold
• Target distribution
• Allocated distribution
• Settled distribution
• Actual current hold

This turns distribution risk into measurable asset data.

A $18 billion facility is one transaction. Each lender experiences it differently. For Bank A:

Record item	Value or state
Commitment	$2.0bn
Funded	$2.0bn
Target hold	$500m
Distributed	$700m
Current hold	$1.3bn

For Bank B:

Record item	Value or state
Commitment	$500m
Funded	$500m
Target hold	$500m
Distributed	$0
Current hold	$500m

The facility-level record and holder-level records need to coexist.

If arrangers retain more debt than expected, the consequences can extend beyond one asset. Balance-sheet capacity may affect new underwriting, sector concentration, capital allocation, risk appetite, pricing, syndication strategy, warehouse capacity and hedging. That makes post-closing distribution state relevant to portfolio management. For a credit manager, the question is not merely: Is this loan good? It can also be: Why do we still own this much of it?

A loan can have low expected default risk and still be difficult to distribute. Reasons may include price disagreement, portfolio concentration limits, market volatility, sector saturation, rating constraints, tenor, documentation, transfer restrictions, project uncertainty, investor liquidity, competing supply or required return. So: Default risk and Distribution risk must remain separate.

The concepts overlap, but they are not identical. Liquidity risk asks: Can I sell this asset, in what size, at what price and how quickly? Distribution risk asks: Can the arranger move the intended amount of newly originated exposure to the target investor base? A highly liquid secondary market can reduce distribution risk. But distribution can fail for reasons beyond secondary-market liquidity. The proposed evidence record should preserve both dimensions.

A large syndicated facility does not need to move as one $18 billion block. Different lenders can hold different pieces. Different transfers can occur on different dates. Different prices can apply. Different investors can have different restrictions. That means the correct model is not Loan owner: Bank consortium. It is closer to:

Facility → Lender position 1 → Lender position 2 → Lender position 3 → ...

Each position has owner, amount, effective date, transfer history, encumbrance, pricing, settlement and source.

In syndicated lending, an administrative or facility agent may maintain the lender register or equivalent authoritative ownership record under the transaction documents. A platform such as DaDepo should not pretend to replace that legal record. Instead, the proposed record could reference: Authoritative agent record and preserve last reconciliation date, current holder, position amount, pending transfer, source document, reviewer and discrepancy state. This is the difference between asset information infrastructure and pretending to be the legal ledger.

A proposed record for the project and the loan

Large project debt can depend on many linked objects.

Record layer	Information to retain
Project identity	project name; location; site; development entity; ownership; construction status; expected completion; capacity; permits.
Financing	facility; tranche; commitment; funded amount; maturity; extension options; margin; fees; amortisation; covenants; security.
Commercial support	tenant; customer; lease; capacity contract; offtake arrangement; minimum payments; guarantees; termination rights; concentration.
Infrastructure dependencies	power; grid connection; gas; fuel cells; water; permits; easements; construction contracts; commissioning.
Distribution	arrangers; underwriters; initial lenders; target hold; target syndication; investor marketing; allocations; settled transfers; current holders; retained exposure.
Market state	indicative bid; indicative offer; executed trades; rating; credit spreads; valuation; liquidity; date.

The debt asset sits inside this graph.

The Financial Times has reported that Project Jupiter is behind its original development schedule. That is relevant to credit. It should not be represented automatically as Loan default: Yes. A project can be delayed without breaching a debt covenant, and a covenant can be breached without a payment default. A waiver can cure or modify a breach, a construction milestone can be moved, and a committed completion date can differ from a public target. The record should retain each distinction.

The system needs separate fields for:

• Project schedule status
• Contractual milestone status
• Covenant compliance
• Event of default
• Payment status

That separation prevents narrative reporting from becoming an incorrect legal conclusion.

Project Jupiter has attracted local opposition concerning environmental impacts. But a proposed evidence record should preserve who says what. For example:

Record item	Value or state
Concern	Water use
Source	Local opponents / public reporting
Status	Contested / under review

Record item	Value or state
Concern	Air permit
Source	Court / regulatory proceeding
Status	Proceeding stayed

Record item	Value or state
Company response	Data-center construction permits unaffected
Source	Oracle statement, 14 Sep 2026

This allows a credit manager to see the issue without the platform declaring the legal outcome.

Evidence-led AI and a proposed distribution record

The following AI uses describe proposed assistance within this evidence model. They do not establish that DaDepo currently automates these workflows. Extraction and inconsistency flags would remain suggestions for review, not legal, credit, compliance or valuation decisions.

AI can be useful because syndication evidence is often scattered across:

• commitment letters;
• facility agreements;
• lender schedules;
• allocation notices;
• trade confirmations;
• assignment agreements;
• agent notices;
• pricing sheets;
• valuation reports;
• emails;
• market indications;
• portfolio reports; and
• settlement records.

AI can help extract lender names, committed amounts, funded amounts, target allocations, actual allocations, transfer dates, principal balances, indicative prices, settlement dates, rating changes, project milestones and source references. Across a large portfolio, AI could flag:

• loans still showing the original underwriter as full holder after distribution;
• a transfer agreement with no corresponding settlement;
• current-holder totals that do not equal outstanding principal;
• an indicative quote stored as an executed trade;
• a par amount stored as current market value;
• a project delay incorrectly labelled as default;
• a rating change attached to the wrong entity;
• a stale price with no as-of date;
• an intended allocation presented as a completed transfer; or
• a loan marked “distributed” while the arranger still holds much more than its intended final position.

That can make private-credit operations materially more reliable.

AI should not independently decide:

• that a loan transfer is legally effective;
• that an investor became lender of record;
• that borrower consent was unnecessary;
• that a security interest transferred correctly;
• that a quoted price is executable;
• that a quote represents fair value;
• that an arranger has realised a loss;
• that a project delay constitutes default;
• that a permit will or will not be granted;
• that a covenant has been breached;
• that the borrower is insolvent;
• that the loan is impaired; or
• that an investment should be bought or sold.

Those require authoritative records, legal analysis, accounting treatment, valuation and professional judgment.

A useful structured record could contain the following linked sections.

Record layer	Information to retain
Facility identity	facility identifier; borrower; project company; facility type; currency; total commitment; signing date; closing date; first funding date; maturity; extension options; interest benchmark; margin; amortisation; governing law; agent; authoritative facility record.
Security and support	security package; collateral; guarantees; sponsor support; tenant / offtaker arrangements; account security; insurance; intercreditor arrangements; perfection status; release conditions; source documents.
Funding state	committed amount; funded amount; undrawn amount; draw date; conditions precedent; draw purpose; outstanding principal; accrued interest; repayment history; current payment state.

Underwriting state

For each arranging institution:

• underwritten amount;
• funded amount;
• target final hold;
• target distribution;
• internal risk limit;
• underwriting date;
• expected distribution window; and
• current retained exposure.

Distribution event

For each syndication or transfer process:

• distribution event identifier;
• launch date;
• amount offered;
• target investor class;
• marketing status;
• indications received;
• price guidance;
• allocations;
• transfer documentation status;
• consent status;
• settlement status; and
• completion date.

Holder position

For every lender:

• legal lender;
• economic holder where different;
• position amount;
• effective date;
• purchase price;
• transferor;
• transferee;
• settled amount;
• pending amount;
• agent-record status;
• encumbrance; and
• last reconciliation date.

Record layer	Information to retain
Market pricing	instrument / tranche; bid; offer; mid where used; price type; quoted size; quoting institution; quote time; firm / indicative; executed price; trade date; settlement date; source.
Credit state	payment status; covenant status; waiver status; rating; rating agency; rating date; outlook; borrower credit; sponsor credit; tenant / offtaker credit; default state; restructuring state; recovery state.
Project state	construction phase; completion percentage; milestone; schedule; delay; power availability; permit state; regulatory proceeding; litigation; operational start; capacity delivered; revenue commencement; source.
Valuation	par amount; outstanding principal; carrying value; valuation; indicative bid; indicative offer; executed price; methodology; as-of date; reviewer; source.
Provenance	facility agreement; commitment letter; lender schedule; syndication memorandum; allocation notice; assignment agreement; agent notice; lender register; trade confirmation; settlement confirmation; rating release; regulatory filing; project update; public statement; AI-extracted field; professional review; effective date; last-updated timestamp.

This turns “$18 billion of loans” into a traceable set of current asset positions.

Maintain holder continuity and the distribution history

A lender position can move through:

Underwritten → Committed → Funded → Marketed → Allocated → Transfer documented → Settled → Final hold.

Or:

Funded → Marketed → Investor demand insufficient → Retained by arranger → Repriced → Re-marketed.

Or:

Allocated → Transfer conditions fail → Allocation cancelled → Exposure remains with seller.

These are real lifecycle states.

A system should avoid: Distribution status: Done without knowing what “done” means. A stronger model can distinguish:

• Marketing complete
• Allocations agreed
• Transfer documents executed
• Cash settled
• Agent register updated
• Arranger at target final hold

Only the last of those may answer the balance-sheet question the arranger actually cares about.

One of the strongest automated controls is simple. If outstanding principal is $18bn then current lender positions should reconcile to that amount, subject to clearly identified mechanics such as undrawn commitments, participations or other structures. If Sum(current lender positions) != authoritative outstanding principal the system should flag the discrepancy. That is more useful than a beautifully formatted syndication report with stale allocations.

When Bank A sells $100 million of a facility to Investor B, the underlying loan does not need a new economic identity. The holder position changes. A durable record can preserve:

Record item	Value or state
Facility ID	constant
Holder	Bank A -> Investor B
Position	$100m
Effective date	[date]
Settlement	complete

This is asset-level continuity.

Once individual loans have clean holder histories, they can be grouped into:

• bank books;
• private-credit funds;
• CLOs;
• securitisation pools;
• continuation vehicles;
• insurance portfolios;
• warehouses; or
• debt-sale portfolios.

The portfolio record should not erase the loan-level record. That connects Project Jupiter directly to the broader DaDepo thesis:

Institutionalisation depends on portable asset identity.

A stalled syndication tells a future buyer something. It can reveal:

• the price at which prior investors declined;
• concentration concerns;
• market-clearing uncertainty;
• documentation friction;
• risk appetite;
• financing supply;
• holder pressure; and
• potential future secondary supply.

That information belongs in the asset history. It should not be hidden because the original financing technically closed.

Suppose the banks eventually sell the debt at 95 or 100. The earlier 89–91 quote still matters historically. The asset lifecycle should preserve:

Origination at / near par → Distribution stalled → Indicative quote 89–91 → Later sale / transfer.

Current state and historical state should coexist.

Distribution risk appears across private markets.

Leveraged loans

Banks can underwrite acquisitions and become stuck with exposure when markets weaken.

Commercial real-estate finance

Construction or bridge loans may be originated with an expectation of refinance or distribution that later becomes difficult.

Infrastructure debt

Long-duration project loans may need institutional take-out after initial bank funding.

Private credit

Managers can originate assets expecting warehouse financing, syndication, club participation or later portfolio sales.

Receivables finance

A funder may acquire invoices expecting to refinance them through a warehouse or securitisation.

Mortgage finance

Originators can accumulate loans intended for sale or securitisation.

NPL portfolios

A buyer can finance acquisition with an expectation of later refinancing, sub-participation or portfolio disposal.

Litigation claims

A funder can accumulate exposures that may later be syndicated or transferred to another capital provider. The same infrastructure principle applies: Origination is not Distribution and neither is Settlement.

AI data-center finance already illustrates another DaDepo principle. In an earlier Insight, the central point was:

The collateral is not enough.

A project loan may depend on permits, power, leases, construction milestones, commissioning and revenue commencement. Project Jupiter adds the next layer:

Funding is not the end of the loan lifecycle.

Even after money is advanced, the debt itself can need to move through a market. The two lessons fit together:

Project readiness → Loan origination → Funding → Distribution → Secondary ownership → Settlement → Ongoing monitoring.

A future DaDepo model would need to be capable of following the asset through all of them.

A bounded contribution from DaDepo

DaDepo's supported contribution is private intake of supported documents and preparation of a draft for user review. Its Parties and contractual roles evidence table can record the party, specific contract, source-described role, document and clause reference, relevant source-stated date and review status. Suggested rows require the user's review and confirmation. The wider relationship models, automated monitoring, reconciliation and decision workflows discussed here are proposed capabilities, not a description of existing DaDepo functionality.

DaDepo does not need to become a syndicated-loan desk. It does not need to arrange Project Jupiter financing. It does not need to quote loans. A proposed contribution would be an evidence model for reviewers. A proposed DaDepo extension could help connect:

Project → Loan → Funding → Arranger → Underwritten exposure → Target distribution → Investor allocation → Transfer → Settlement → Current holder → Current market indication.

That can make it easier to answer:

• what the exact loan is;
• how much has been funded;
• who legally holds it;
• who was expected to hold it;
• how much an arranger intended to distribute;
• what has actually been distributed;
• which transfers remain pending;
• which prices are indicative;
• which prices came from executed trades;
• what project changes have occurred;
• which credit changes have occurred;
• what the current holder exposure is; and
• which source supports every answer.

The platform does not create market liquidity. The proposed model would make the loan’s movement through the market inspectable.

Creating or reviewing a loan-distribution Asset Passport does not mean that DaDepo has:

• arranged a syndicated loan;
• underwritten a facility;
• committed capital;
• funded a borrower;
• approved a drawdown;
• verified legal title to every lender position;
• acted as facility agent;
• maintained the authoritative lender register;
• approved a loan assignment;
• obtained borrower consent;
• perfected transferred security;
• quoted a loan;
• determined fair value;
• calculated an accounting carrying value;
• executed a loan trade;
• settled a transfer;
• rated a borrower;
• determined covenant compliance;
• determined default;
• assessed project feasibility;
• determined permit validity;
• valued collateral;
• recommended a loan purchase or sale; or
• provided lending, arranging, syndication, brokerage, legal, accounting, tax, regulatory, project-finance, investment or valuation advice.

Important: DaDepo provides technology and information tools. It does not provide loan arranging, underwriting, syndication, brokerage, facility-agency, lending, custody, legal, accounting, tax, regulatory, investment or valuation advice or services unless a specific service is expressly identified and lawfully provided. Loan ownership, transferability, security, settlement, default, covenant status and valuation depend on governing documents, authoritative records, applicable law, accounting standards, market evidence and qualified professional review.

Review questions before calling a loan distributed

Before a large funded loan is described as successfully distributed, ask:

1. Asset identity: Which exact facility and tranche are being discussed?
2. Borrower: Who is the legal borrower?
3. Project: Which project or assets does the financing support?
4. Commitment: What is the total committed amount?
5. Funding: How much has actually been advanced?
6. Outstanding principal: What remains outstanding today?
7. Arrangers: Which institutions arranged or underwrote the financing?
8. Initial lenders: Which institutions funded at closing?
9. Target hold: How much did each arranger intend to retain?
10. Target distribution: How much was intended for other investors?
11. Marketing: When was syndication or distribution launched?
12. Investor universe: Which banks, funds, insurers or other investors were targeted?
13. Indications: What demand was received?
14. Allocation: Which amounts were allocated?
15. Transfer documents: Which assignments or participations were executed?
16. Consent: Were required borrower, agent or other consents obtained?
17. Settlement: Which transfers actually settled?
18. Authoritative record: Does the agent record reflect the new holder?
19. Current holder: Who owns each position today?
20. Current retained exposure: How much does each arranger still hold?
21. Distribution shortfall: How much exceeds the intended final hold?
22. Par amount: What principal amount is attached to the position?
23. Indicative bid: What current bid exists?
24. Indicative offer: What current offer exists?
25. Quote provenance: Who supplied it and when?
26. Executed trade: Has any transaction occurred at that price?
27. Accounting value: What carrying value is used and by whom?
28. Rating: Which entity is rated, by which agency and as of what date?
29. Project status: What construction and operational milestones have been achieved?
30. Permits: Which permits are effective, challenged, pending or disputed?
31. Covenants: Is the facility compliant with its contractual requirements?
32. Payment state: Are interest and principal being paid as required?
33. Security: What collateral and guarantees support the loan?
34. Secondary liquidity: Can positions be transferred today and under what restrictions?
35. Reconciliation: Do current lender positions reconcile to authoritative outstanding principal?
36. Current state: Can every material field be traced to a current source and date?

If the answer to:

“Who owns the funded risk today, how did it get there, and at what price could it actually move?”

is unclear, the loan-distribution record is incomplete.

Funding creates the loan; distribution locates the risk

That is the lasting lesson from Project Jupiter. The project obtained financing. The debt was created. The borrower received access to capital. But the intended movement of that debt through the institutional market has become a separate problem. The financing therefore contains several truths at once:

• Loan exists
• Funding occurred
• Distribution was intended
• Distribution is incomplete
• Banks retain more exposure than planned
• The debt is being quoted below par

None of those facts should overwrite the others. They describe different parts of one lifecycle.

A loan is not distributed when it is funded. Distribution is a sequence of allocations, transfers, holder changes, prices and settlement events that deserves its own asset history.

The project was funded. The debt did not fully move. That is why private debt needs a placement lifecycle.

Sources

• Reuters: Oracle's $18 billion data center debt under pressure, FT reports
• Financial Times: Oracle’s $18bn data centre debt under strain amid local pushback
• Reuters: Banks lend $18 billion for Oracle-tied data center project, Bloomberg News reports
• Oracle: Project Jupiter Statement on Construction Permitting
• Oracle: Project Jupiter to Provide Environmental Data to the Public
• Wall Street Journal: Oracle Gets Credit Rating Downgrade