Edition 25 ended with a question: what changes when the counterparty behind a bankable contract is a neocloud rather than a hyperscaler? Hut 8's 2026 financings offer an unusually useful way to investigate it.
1 The deal
In April and June 2026, two Hut 8 project subsidiaries issued US$7.5 billion of investment-grade senior secured notes to help finance data centres under construction. [1] The notes are non-recourse to Hut 8's parent company, subject to the applicable carve-outs. Both mature in 2042. Yet the tenants behind the expected rent are different.
At River Bend in Louisiana, the immediate tenant is Fluidstack, a neocloud. Google provides a financial backstop for specified rent payments and other obligations under Fluidstack's lease. At Beacon Point Phase 1 in Texas, the lease is with an unnamed high-investment-grade company, with no separate external lease backstop disclosed.
Different counterparty architectures. The same broad capital-market outcome: investment-grade, non-recourse project debt. That does not prove the two projects carry identical risks. It challenges the simpler assumption that a neocloud tenant necessarily prevents long-dated institutional financing. The better question is what sits between the tenant and the noteholder.
Same market, different credit chains
| River Bend — Louisiana | Beacon Point Phase 1 — Texas | |
|---|---|---|
| Critical IT capacity | 245 MW | 352 MW |
| Tenant | Fluidstack (neocloud) | Unnamed high-investment-grade company |
| Lease | 15-year triple-net | 15-year triple-net |
| Additional support | Google backstop of specified lease obligations | No separate external lease backstop disclosed |
| Bond rating | BBB− (S&P and Fitch) | Baa2 (Moody's) |
| Senior secured notes | US$3.25bn at 6.192% | US$4.25bn at 6.129% |
| Relative issuance spread | Reference transaction | 20bp tighter than River Bend, as reported by Hut 8 |
| Maturity | 15 November 2042 | 30 November 2042 |
| Amortisation begins | May 2028 | May 2030 |
| Note guarantees | Not guaranteed by Hut 8, Fluidstack or Google | Not guaranteed by Hut 8 or the tenant |
Both financings are non-recourse to Hut 8's parent, subject to applicable carve-outs. The most important distinction is not the coupon. It is the difference between a lease-level backstop and a note guarantee. At River Bend, Google supports specified obligations under Fluidstack's lease — Google does not guarantee the project bonds. Under the disclosed structure, noteholders have claims against the project issuer and its secured assets and accounts; Google's support behind the lease should not be mistaken for an unconditional promise by Google to repay the bonds. At Beacon Point, the high-investment-grade tenant is directly obligated under its lease, but does not guarantee the project notes either.
Rent is expected to be the principal source of debt service in both cases. First-priority project security, pledged issuer equity and debt-service reserve accounts form part of the financing protection.
What can we observe? Both projects reached the investment-grade bond market using different payment-support arrangements. What cannot we observe from the public filings alone? Each investor's internal cash-flow stress tests, recovery assumptions, or how much of the financing price was attributable specifically to tenant credit rather than other project characteristics. That boundary matters for the investment conclusion.
The two financings also received different ratings and issuance spreads, despite their remarkably similar coupons. That distinction matters. But it does not automatically tell us how much of the difference reflects tenant credit rather than other financing characteristics.
2 The engineering read: four clocks, running together
The contract, physical asset and financing do not begin producing or demanding cash on one synchronised timetable. A credit committee must track four overlapping clocks.
Delivery clock
Power interconnection, construction, testing and commissioning determine when contracted MW can actually be handed over. Debt can be raised while those steps remain incomplete. If energisation slips, the financial impact depends on the lease, completion protections, reserves and who contractually bears the delay.
Tenant-revenue clock
A neocloud may rely on compute customers whose service commitments and cash generation differ from the data-centre lease. The relevant questions are the duration, concentration, performance conditions and renewal risk of those commitments — not a blanket assertion that all neocloud contracts last only months.
Lease clock
The lease determines what rent is due, when it starts, how it escalates and under what circumstances obligations can end. Fifteen years of contractual rent is meaningful. It is not equivalent to fifteen years of certain collections.
Debt-service clock
Interest, principal instalments, reserve requirements and covenant triggers have their own dates. River Bend's scheduled amortisation begins in May 2028, while Beacon Point's begins in May 2030. The notes are structured to amortise through maturity, reducing dependence on a large scheduled bullet refinancing. That does not eliminate cash-flow shortfall, covenant, restructuring or residual-value risk.
This is where an engineering constraint becomes a financing constraint: delayed ready-for-service → potentially delayed rent or support activation → tighter debt-service cash flow or greater reserve use → less headroom for distributions and downside shocks. This is a conditional transmission chain, not an assertion that either Hut 8 project has experienced a delay or covenant breach. The actual contractual allocation determines where the economic loss lands.
3 The capital allocation read
From contractual backlog to debt-service capacity
The same underwriting discipline appears at a different financing layer: computing infrastructure and its customer contracts, rather than a data-centre property's lease income. In March 2026, CoreWeave entered into its US$8.5 billion DDTL 4.0 facility to finance GPUs and related infrastructure required for a customer contract. [2]
The facility received ratings of A3 from Moody's and A (low) from Morningstar DBRS [3] — ratings associated with the secured transaction and its borrowing entity, not investment-grade corporate ratings of CoreWeave Inc. The financing has a specific legal perimeter: the borrowing subsidiary's obligations are secured by substantially all assets of the borrower and its subsidiaries, while borrower subsidiaries provide guarantees. CoreWeave Inc.'s parent-level guarantee is limited to specified bad acts.
The disclosed facility also requires a minimum 1.15× debt-service coverage ratio (DSCR). [2] Testing begins in the first full calendar month after the earlier of the loan commitments being reduced to zero, or 30 June 2027. That is a contractual testing trigger, not evidence that the borrower currently has 1.15× coverage or that all neocloud facilities use the same covenant threshold.
Put simply, lenders cannot treat a headline backlog as though it were cash already available in a debt-service account. They need to examine which contracted receipts can service debt, what collateral is pledged, what expenses and principal payments fall due, and what happens when service performance or customer payments fall short. CoreWeave reported approximately US$104 billion of revenue backlog at 30 June 2026 [2] — a figure that includes remaining performance obligations and estimated future revenue under committed customer contracts, subject to delivery and service-availability requirements. Backlog is not cash collected, realised earnings or a DSCR calculation.
The Hut 8 projects and CoreWeave's DDTL cannot be priced against each other as though they financed the same asset — the former are property-level project notes, the latter finances GPU and customer-contract-backed infrastructure. The shared principle is narrower and more useful: investment-grade credit quality can be achieved at the financing-structure level without requiring the ultimate operating parent to hold the same investment-grade corporate rating.
What the pricing cannot tell us
Hut 8 reported that Beacon Point's notes priced 20 basis points inside River Bend's issuance spread. That issuer-reported comparison is informative, but it does not isolate the cost of a neocloud counterparty. The notes were issued about 40 days apart, against different assets, capacities, amortisation profiles, ratings and market conditions. Beacon Point's Baa2 rating from Moody's is also one notch above River Bend's BBB− ratings from S&P and Fitch — a cross-agency equivalence, not a same-agency comparison, which makes the rating gap alone no easier to isolate than the spread.
The 6.192% and 6.129% coupons are not, by themselves, a controlled experiment in tenant-credit pricing. The defensible conclusion is that a supported neocloud lease and a direct high-investment-grade lease both enabled access to investment-grade, non-recourse project financing in these transactions. How much the support package reduced funding costs, changed debt sizing or altered recovery assumptions is not disclosed with sufficient precision to quantify. For investors, the financing result is observable. The full internal underwriting process is not.
Investment Lens: which MW deserves the capital?
Edition 25's five Financeable MW gates remain unchanged: counterparty, contract, delivery, collateral and risk allocation. Edition 26 changes the questions inside those gates — payment, protection, timing, resilience and recovery.
The return implication is directional, not a published project-level IRR estimate. Stronger protection may support debt capacity or reduce downside risk, but its value depends on cost, enforceability and contractual limits. Fully amortising debt can reduce refinancing dependence while requiring more principal cash payments during the operating life. Equity value depends on both sides of that trade-off.
4 What this means for the broader market
APAC: plenty of financing, less comparable tenant-level disclosure
In July 2026, AirTrunk announced US$2.325 billion of green financing for its JHB2 hyperscale data-centre campus in Johor, describing it as its largest single-asset financing to date. [4] A Reserve Bank of Australia staff analysis published on 23 September 2026 estimated that Australian data-centre operators had raised at least A$35 billion in 2026 through 8 September, with syndicated lending representing the largest financing channel. [5]
Those figures demonstrate the scale of infrastructure capital formation across parts of APAC. They do not establish a comparable APAC funding spread for neocloud versus direct investment-grade tenants — the borrowers, assets, guarantees and security packages differ, and some crucial tenant-level terms are not publicly disclosed. The disciplined APAC question is therefore not simply which market funds neoclouds more cheaply. It is: where are the payment obligations, credit support and recovery rights sufficiently disclosed to make a meaningful comparison?
Beyond the rating
The River Bend and Beacon Point comparison does not suggest that the market has stopped caring about tenant credit. It shows that tenant credit is one input into a transaction-specific assessment of contractual payment, enforceable support, collateral, repayment timing and recovery. A direct investment-grade lease provides one path. A neocloud lease with a defined external backstop provides another. A GPU-backed facility is a different financing solution altogether. None eliminates the need to ask what happens when the real project departs from the financing model.
A contract can make the MW financeable at closing. The real test is who keeps it financeable after closing.
And if a data-centre operator depends on a handful of large customers, how much of that contracted growth becomes resilient cash flow for the operator's own investors? That is the question I'm taking forward next.