High-Coupon Credit for Data-Centre Developers: What Lenders Want

Introduction

A data-centre developer in the Gulf who takes a construction financing proposal from a private credit fund in 2026 will typically see an all-in cost of debt in the mid-teens: a cash coupon in the low teens, original issue discount, perhaps a payment-in-kind toggle, an equity kicker, and a security package that reaches into every contract and account the project owns. The first reaction of many sponsors is indignation. The asset class is the most discussed infrastructure theme of the decade; hyperscale demand is visible to everyone; the sponsor's own conviction is total. Why, then, is the debt priced as if the project were distressed?

This paper argues that the question is badly posed, and that reposing it correctly is worth several hundred basis points to the sponsor. A double-digit coupon on data-centre development debt is not a verdict on the sector. It is the sum of prices attached to specific, identifiable risk blocks: development and permitting risk, construction and delivery risk, merchant or short-tenor offtake risk, power availability risk, sponsor balance-sheet thinness, and jurisdictional enforcement uncertainty, plus a premium for the illiquidity and complexity of the instrument itself. Each block is priced, structured around, or declined on its own terms. A sponsor who understands how the credit committee decomposes the loan can remove several blocks before launch and materially compress the coupon. Conversely, where a block cannot be removed, the lender will take payment in structure rather than price: covenants, collateral, cash control and step-in rights. Understanding that exchange rate, between coupon and structure, is the second half of the sponsor's education.

The paper is written primarily for the sponsor: the data-centre and digital-infrastructure developer or operator in the Gulf who will borrow from international private credit, direct-lending and special-situations funds. It is written secondarily for those funds, who will recognise the framework and may find the sponsor-side articulation useful in their own origination conversations. The arithmetic throughout is anchored to a single reference project whose base-case calibrations, drawn from the author's transaction experience in the region, are recorded in Section 4 and Appendix A.

Research questions

RQ1. How does a private credit committee decompose a Gulf data-centre development loan into risk blocks, and how is each block priced in the coupon or absorbed in structure?

RQ2. What does the full security and structural package of a high-coupon data-centre financing consist of, and what economic function does each element serve?

RQ3. Which sponsor actions compress the coupon, in what sequence, and by how much?

Contributions

Institutional And Market Context

This section explains why the marginal lender to Gulf data-centre development is a high-coupon private credit fund rather than a bank, and what features of the Gulf environment shape the underwriting.

The AI demand shock and the developer population it created

The rapid growth in demand for artificial-intelligence training and inference capacity has changed both the scale and the population of data-centre development in the Gulf. Governments in the region have made digital infrastructure a stated policy priority, hyperscale cloud and AI firms have publicly announced regional ambitions, and land, power and capital have been directed towards the sector. The demand shock has done two things to the borrower population. First, it has pulled forward projects of a size that the region's developers have not previously delivered, so that even experienced sponsors are operating beyond their track record. Second, it has attracted new entrants: real-estate developers, family groups, telecommunications spin-outs and first-time platforms, many with thin balance sheets relative to project size and no history as borrowers of institutional scale. Both effects raise the unverifiable component of credit risk exactly as the theory of Section 2 describes, and both push the borrower towards the monitored, expensive end of the credit spectrum.

Why banks are not the marginal lender

Regional and international banks do lend to Gulf digital infrastructure, but predominantly to a narrow slice: operating assets with contracted cash flows, government-linked sponsors, or corporate facilities to investment-grade groups. Development-stage lending to independent sponsors sits poorly with bank constraints for several reasons. Construction-phase project risk consumes regulatory capital heavily and sits awkwardly in bank risk frameworks when the offtake is uncontracted. Banks are institutionally cautious about asset classes without long regional default histories. Internal single-name and sector concentration limits bind quickly when individual projects are large. Credit processes calibrated to amortising, covenant-standard facilities adapt slowly to milestone-drawn development debt with bespoke controls. None of this means banks are absent; it means that for the specific product a development sponsor needs, a large, fast, flexible construction facility against a partly merchant asset, the bank is rarely the marginal price-setter. The marginal lender is the international private credit, direct-lending or special-situations fund, and the pricing conventions of that market, absolute-return targets, original issue discount, call protection and equity participation, follow it into the sector.

Enforcement considerations across GCC jurisdictions

Creditor enforcement in the Gulf requires careful, jurisdiction-specific analysis. The GCC states have invested substantially in creditor-friendly reform: modern insolvency statutes, moveable-collateral registries, and, in the United Arab Emirates, common-law financial free zones, the Dubai International Financial Centre and the Abu Dhabi Global Market, whose courts apply familiar commercial law and whose judgments interact with onshore courts through defined gateways. At the same time, several practical realities shape lender behaviour. Enforcement of security over onshore land and buildings can involve procedures that are slower and less tested for this asset class than lenders would like, and foreign ownership of onshore real property is restricted in various ways across jurisdictions. Precedent for the enforcement of security over an operating data centre, as distinct from generic commercial property, is thin everywhere in the region. Court-supervised reorganisation regimes are young, and the behaviour of local courts in a contested, large-scale infrastructure insolvency is not yet richly evidenced. The rational lender response, predicted by Proposition 3, is not to abandon security but to relocate it: hold shares of the project company through a free-zone or offshore holding structure whose pledge is enforceable in a predictable forum, take assignment of contracts and receivables governed by English law, and control cash through accounts held with institutions and in places where account security is routine. Enforcement uncertainty is thereby transformed from an unpriceable block into a structured one, at some residual premium.

Power as the binding physical constraint

In most Gulf markets electricity is generated, transmitted and sold by state-linked utilities, and large new connections are allocated through utility and government processes rather than through a liberalised market. For a data-centre developer this converts power from an operating cost into a development risk: the project's viability depends on a grid connection and a supply arrangement whose timing and terms rest on administrative decisions. Tariff levels for large industrial users are, in general, internationally competitive, and several jurisdictions offer credible renewable or clean-energy supply structures, which matters to hyperscale tenants with sustainability commitments. But from the lender's chair the question is narrower: does this project hold an executed, unconditional connection and supply arrangement for the load it needs, on dates consistent with the construction schedule? Until the answer is yes, power availability is a distinct risk block with its own premium.

How The Lender Underwrites

The decomposition at committee

Development-stage data-centre facilities in the Gulf currently price at all-in coupons of 14 to 16 per cent for a thinly capitalised sponsor with partly merchant offtake, and the credit paper that supports such a loan is organised in a recognisable way. The committee memorandum does not debate the sector; the sector view is settled in a paragraph. It decomposes the transaction into the six risk blocks of Section 4, prices each block that can be priced, assigns structure to each residual that cannot, and tests the whole against three quantitative gates: loan-to-cost, debt service cover through the ramp, and recovery on the worst day. Table 2 summarises the decomposition as the committee sees it; Figure 1 presents the author's decomposition of a representative 15 per cent all-in coupon into its components.

Table 2. The lender's risk-block decomposition of a Gulf data-centre development loan

Figure 1. Decomposition of a representative 15 per cent all-in coupon on Gulf data-centre development debt

Two features of the decomposition deserve emphasis before the blocks are examined individually. First, the base rate and the illiquidity margin, together 5.25 percentage points in Figure 1, are not negotiable by the sponsor: they are the cost of the money and of the bespoke, buy-and-hold instrument through which it arrives. Everything above them, 9.75 percentage points in the representative build-up, is block premia, and each block premium is negotiable against evidence. Second, the block premia in Figure 1 are the residual premia after standard structure has been applied. The lender does not price a naked merchant project at 300 basis points of offtake premium; it prices at 300 basis points a project whose uncontracted revenue is already caught by a cash sweep and whose receivables are already assigned. Structure is in the price, which is why giving structure away cheaply is a sponsor error examined in Section 6.

Block by block

Development and permitting. This block behaves as Proposition 2 predicts at its extreme: it is barely priced at all, because it is barely priceable. A committee cannot quantify the probability that a zoning variance arrives or that an environmental clearance is granted; it can only verify whether these events have happened. In practice the block is handled with conditions precedent rather than coupon: land rights executed, master-plan consent issued, building permit granted, all before first draw. A residual premium of 50 to 100 basis points appears where minor conditions remain open on a defined timetable. Where material permits are outstanding with no defined path, the lender declines, and the sponsor who shops an unpermitted project to twenty funds is not conducting a process; it is educating the market against itself.

Construction and delivery. The committee's question is not whether the sponsor can build a data centre; it is who stands behind price and date, and with what money. The strongest answer is a fixed-price, date-certain engineering, procurement and construction wrap from a contractor of substance, with liquidated damages for delay sized to cover debt service during the overrun period and performance security in the form of bonds or parent guarantees. Gulf construction markets can deliver this answer, but at a price in the EPC margin, and sponsors frequently self-perform or split packages to save that margin. The lender prices the difference: an unwrapped or multi-package structure with interface risk carries 150 to 300 basis points against a fully wrapped one, and in addition attracts tighter structure, milestone-based draws certified by the lender's technical adviser, retention accounts, and a contingency line that the sponsor funds first. Long-lead equipment is a specific committee concern: generators, switchgear and cooling plant have extended delivery times, and a credit paper that shows executed supply agreements with staged payments and security over equipment in transit reads very differently from one that shows quotations.

Offtake and revenue. This is the largest block, the heart of Proposition 4, and the place where the data-centre asset class divides in two. A campus fully leased to a creditworthy hyperscale or sovereign-linked counterparty, for a term at least as long as the debt, on terms where termination requires making the lender whole, is an infrastructure credit: the loan is underwritten to the counterparty's covenant and the premium collapses towards zero. A merchant campus, built to specification in anticipation of demand, is underwritten to a market forecast the committee cannot verify, and carries 200 to 400 basis points together with the heaviest structure in the package: assignment of all offtake receivables, revenue accounts under lender control, a full cash sweep on uncontracted revenue, and covenants restricting the terms on which future capacity may be let. Between the two poles sit the common intermediate cases, anchor leases over part of the capacity, shorter-tenor agreements with strong renewal economics, and reservation agreements that are firm in commercial intent but soft in legal force. The committee reads these documents with care, because the label matters less than three clauses: the termination rights, the credit standing of the actual signing entity, and the conditionality between the offtake and project milestones. A reservation agreement terminable for convenience compresses nothing.

The Sponsor'S Playbook

Section 5 described the underwriting from the lender's chair. This section turns the same framework around: if the coupon is an additive stack of block premia, then the sponsor's capital-markets strategy is a portfolio of de-risking actions, each with a cost, a duration and a pricing effect, to be sequenced by return on effort. This is Proposition 1 applied as strategy.

Which actions buy what

Table 4 sets out the principal de-risking actions and the pricing effects they carry in current market terms. The ranges are bounded by the Figure 1 build-up: an action cannot compress a premium by more than the premium it addresses, and no combination of actions reaches the base rate plus margin floor of roughly 5.25 per cent, because that floor is the price of the instrument, not of the project.

Table 4. The coupon-compression playbook: de-risking actions and their pricing effects

Source: author's analysis. Effects are not fully additive; see Section 6.2.

The single most consequential row is the first, as Proposition 4 states. A pre-let to a hyperscale or sovereign-linked tenant, on a term matching the debt with lender-friendly termination provisions, does not merely compress a premium; it changes which funds can bid. Infrastructure-debt capital that will not look at a merchant Gulf development will compete for a contracted one, and competition does the rest of the pricing work. The realistic combined effect of executing the top four or five rows before launch is 400 to 600 basis points: the difference between borrowing at 15 per cent and borrowing at 9.5 to 11 per cent, which on USD 325 million over a five-year life is of the order of USD 65 to 95 million of interest, several times the cost of every de-risking action on the list combined.

Sequencing and interaction

The actions are not fully additive, and their sequence matters more than sponsors expect.

Power before offtake. No serious tenant signs a lease against unallocated power, and no lender prices a lease conditional on power as a firm lease. The grid connection and supply agreements are therefore the first milestone, even though their direct pricing effect looks modest in Table 4: they unlock the row above them.

Offtake before EPC finalisation where possible. A signed anchor tenant changes the specification risk in the EPC and lets the sponsor buy the wrap for the right building. Sponsors who lock the EPC first sometimes pay twice, once for the wrap and again for the variations.

Structure early, not late. The offshore holding structure and English-law document architecture cost little if built before the first contract is signed and a great deal if retrofitted across executed onshore agreements. This row should be completed before any term sheet is requested.

Interaction discount. Executing the offtake row shrinks the pricing effect of the LTC row, because a contracted project supports higher leverage at the same coupon; conversely, in a merchant financing extra equity is the strongest remaining lever. A sponsor should treat Table 4 as a menu whose prices move as items are ordered, and re-quote the market after each major milestone rather than negotiating the end-state coupon at the start.

The fundable pack

Credit committees see many Gulf data-centre proposals and fund few. The difference between a pack that gets priced and a pack that gets a polite decline is consistent, and it is not production values. A fundable pack answers the committee's questions before they are asked, in the committee's own order.

Conclusion

This paper set out to answer three questions: how a private credit committee decomposes a Gulf data-centre development loan, what the security and structural package consists of and why, and which sponsor actions compress the coupon.

The answers form a single argument. A double-digit coupon on data-centre development debt is not a judgement on the sector; it is the sum of prices attached to six identifiable risk blocks, development and permitting, construction and delivery, offtake, power, sponsor depth, and enforcement, stacked on a base rate and an illiquidity margin that no sponsor can negotiate away. The committee prices what it can verify and takes structure for what it cannot: share pledges enforceable in predictable forums, assignment of receivables, direct agreements and step-in rights, completion and overrun support, reserve accounts and cash waterfalls, milestone-based funding. The high-coupon toolkit, original issue discount, PIK toggles, equity kickers, call protection, is the same exchange conducted in the currency of return rather than control.

For the sponsor, the practical content of the paper is the inversion of that underwriting. Power first, because it unlocks everything above it; offtake second, because it is the largest compressible premium and changes which lenders can bid; the construction wrap, the funded support package and the offshore security architecture behind them; the fundable pack that answers the committee in its own order; and a negotiation that trades coupon for call flexibility and writes the de-risking ratchet into the contract. Executed in sequence, these actions compress the coupon by 400 to 600 basis points and, more importantly, convert the sponsor from a price-taker educated by term sheets into a counterparty that prices its own risk.

For the credit fund, the same framework identifies which parts of today's return are durable, the premia on complexity, speed, monitoring and enforcement engineering, and which parts are a wasting asset that refinances away as sponsors learn. Both sides of the table are better served when the decomposition is explicit. The coupon is a price list. This paper has tried to print it legibly.

Declarations

Funding. The author received no external funding for this research.

Conflicts of interest. The author is the managing partner of an independent capital advisory firm that advises managers and allocators on transactions of the type discussed in this paper. No client information has been used, and no live mandate is referenced.

Data availability. The paper uses no proprietary dataset. The base-case calibrations used in the exhibits are recorded in Section 4 and Appendix A.

Disclaimer. This paper is an educational and structural analysis. It is not investment advice, an offer, or a solicitation of any kind.