Summary

  • Cipher’s approximately US$793mn average contracted annualised NOI covers October 2026 through September 2036 and includes three campuses. It is neither the first-year run-rate nor an AWS-only figure, and its clock stops with Barber Lake’s shorter base lease before the two 15-year AWS terms end.
  • Black Pearl and Stingray carry 400MW of gross capacity but 286MW of critical IT load. Black Pearl has begun first rent after an accelerated August 2026 delivery; Stingray’s 10MW network hall and 60MW data hall target rent starts in April and May 2027.
  • The two AWS-backed issuers have US$2.81bn of project notes. Black Pearl’s US$2bn notes follow an indenture amortisation schedule after completion; Stingray’s US$810mn notes amortise after completion according to project NOI. Contract value reaches equity only after delivery, rent, operating obligations, reserves and debt service.

An average is not a receipt

Cipher Digital’s portfolio slide is built to communicate scale. It reports 700MW of contracted gross high-performance-computing capacity, approximately US$11.4bn of contracted revenue and approximately US$793mn of average contracted annualised net operating income. Three campuses sit inside those numbers: Black Pearl, Barber Lake and Stingray.

The footnote does more analytical work than the headline. The NOI average runs from October 2026 through September 2036, ending with the base term of the ten-year Barber Lake lease. Black Pearl and Stingray each have a 15-year base lease with AWS. Their contractual clocks therefore continue after the common portfolio measurement window has stopped.

The annual presentation also shows a ramp, not an immediate plateau. Cipher displays US$97mn of NOI for 2026, US$686mn for 2027, US$727mn for 2028 and a rise to US$894mn by 2035. Those figures should not be averaged casually against calendar years because the company defines its headline window from October to September. They do establish one point without calculation: US$793mn is not the opening cash run-rate.

Nor does it belong entirely to AWS. Barber Lake’s 300MW gross lease with Fluidstack, supported in part by Google, is the third contributor. The two AWS campuses account for approximately US$5.5bn and US$2.0bn of disclosed base contracted revenue, or about US$7.5bn before rounding. The remaining portfolio economics come from a different tenant, a different guaranty structure, a shorter base term and a different project financing.

A useful model must therefore move down one level. It needs a separate lease, construction and debt ledger for each issuer.

Black Pearl has started, not finished

Black Pearl is the larger of the two AWS projects. Amazon Data Services, Inc. leases 300MW of gross capacity representing 216MW of critical IT load near Wink, Texas. The base term is 15 years, with three five-year extension options and a 3 per cent annual rent escalator. Amazon.com, Inc. guarantees the disclosed base-rent and operating-expense obligations.

Cipher presents approximately US$5.5bn of contracted revenue and an NOI margin of roughly 100 per cent. That margin is supported by a triple-net structure under which the tenant bears recoverable operating costs, taxes, insurance and tenant-driven change orders. It is a contractual allocation of costs, not proof that every dollar of rent has begun or that the project can never incur owner expense.

The original financing plan divided 216MW of critical IT into five delivery subphases. The first data hall and network hall were targeted for rack-ready completion at the end of September 2026, with further halls following through February 2027. Rent was expected to start as each designated phase became ready, from October 2026 through March 2027.

That schedule changed. On 24 July, Cipher amended the lease at the tenant’s request to bring the first delivery forward. Its 4 August update says first capacity was delivered in August, two months early, and rent commenced. This is meaningful evidence: Black Pearl crossed from contracted construction into at least partial cash generation.

It is not evidence that all 216MW is accepted and earning rent. The remaining halls still need their own delivery receipts. A portfolio model should record, for each phase, critical IT delivered, acceptance date, rent date, any service credit and cash collected. “Rent commenced” is the first row, not the last.

Black Pearl also has a construction-cost sharing boundary. Amazon agrees to reimburse costs above US$9.5mn per critical-IT MW under the disclosed lease terms. That protection can reduce the sponsor’s exposure to overruns. It does not erase schedule risk, the need to fund costs before reimbursement, eligibility disputes or termination rights tied to delay, casualty, grid outages and other specified conditions.

Stingray starts with two smaller gates

Stingray is Cipher’s second lease with Amazon Data Services. Its 100MW gross envelope supports 70MW of contracted critical IT load. The same 15-year base term, three five-year options and 3 per cent escalator create a family resemblance to Black Pearl. The project is nevertheless a separate commercial and financial instrument.

The first gate is a 10MW critical-IT network hall with rent targeted to begin on 1 April 2027. The second is a 60MW data hall with rent targeted for 1 May. Those are forward dates, not completed milestones. As of the August update, earthwork, grading, pad preparation and underground electrical work were progressing, with first delivery expected in the first half of 2027.

Cipher shows approximately US$2.0bn of contracted revenue over the 15-year base term. Its Stingray presentation also gives a range of roughly US$2.0bn to US$5.7bn of lease payments. The upper value assumes all three five-year extensions are exercised at the ceiling rental rate. It is optional-term arithmetic, not contracted base revenue that should be added to the portfolio today.

Stingray is also described as a triple-net lease with approximately 100 per cent NOI margin and minimal owner operating expense. Amazon covers eligible construction costs above a disclosed cap of US$10.5mn per critical-IT MW. The cap is US$1mn per IT MW higher than Black Pearl’s. That difference is a reminder that the campuses do not share one construction budget merely because they share a tenant.

The parent guaranty strengthens the rent promise. It does not pour concrete, energise the site or certify a rack-ready hall. Before April 2027, the decisive evidence remains construction progress, commissioning and tenant acceptance.

Four hundred gross megawatts become 286 billable ones

Black Pearl and Stingray total 400MW on a gross basis. Their contracted critical IT load totals 286MW: 216MW plus 70MW. The difference is 114MW, or 28.5 per cent of the gross envelope.

That gap is not necessarily unused capacity. It covers the electrical conversion, cooling, pumps, controls and other facility systems required to deliver reliable power to computing equipment. Gross MW describes the utility and site envelope. Critical IT MW describes the load available to the tenant’s machines. Both matter, but they answer different questions.

Construction cost caps are stated per critical-IT MW. Rent phases are tied to rack-ready premises and critical IT delivery. Portfolio capacity is promoted in gross MW. An analyst who divides contract value by 400MW obtains a site-envelope ratio; one who divides by 286MW obtains a compute-load ratio. Neither should be labelled simply “per MW” without naming the denominator.

The same discipline applies to time. Contracted revenue describes nominal payments across a term. Annualised NOI describes a presentation measure over a specified window. Cash rent describes what the tenant actually paid after commencement. Project distributions describe what remains after the issuer’s expenses, reserves and creditors. Combining those four layers produces a larger number, not a more accurate one.

The debt clocks diverge after completion

Black Pearl Compute issued US$2bn of 6.125 per cent senior secured notes due in February 2031. The proceeds finance remaining construction, debt-service reserves, fees and about US$232.5mn of reimbursement for Cipher’s prior equity contributions. The notes are project obligations, not a US$2bn pool of unrestricted corporate cash.

Their principal begins amortising only after Black Pearl is complete, then follows the semi-annual schedules in the indenture. The first rent can begin before full completion and before scheduled amortisation starts. During that interval, rent, interest, reserves, construction and reimbursement sit on different lines of the same project waterfall.

Stingray Compute issued US$810mn of 6 per cent notes due in June 2031. Its debt also waits for completion before amortisation, but the principal paydown is based on consolidated project NOI and a target debt-service coverage ratio. Stingray therefore converts operating performance more directly into mandatory deleveraging.

The two principals sum to US$2.81bn. The arithmetic gives scale; it does not create one cross-collateralised loan. Each issuer has its own collateral, completion support, reserves, lease and remedies. Black Pearl’s early rent cannot automatically cure a Stingray delay, and Stingray’s lower coupon cannot be applied to Black Pearl’s balance.

This is the economic sequence that matters: construction spending creates a rack-ready phase; tenant acceptance starts rent; rent becomes project NOI after permitted costs; NOI services interest and reserves; principal amortisation reduces the creditor claim; only residual cash can support equity value. The US$793mn average sits near the top of that sequence. Equity receives the bottom.

The next disclosure should be a bridge

Cipher has proved that long-duration AWS commitments can fund large project-debt issues. It has also accelerated one Black Pearl milestone. The next useful disclosure is not another combined megawatt total. It is a bridge from signed lease to residual cash.

For Black Pearl, the bridge should show critical IT accepted in the quarter, average rent-producing MW, cash rent, project NOI, construction reimbursements, reserves and the date amortisation begins. For Stingray, it should retain the two April and May gates until acceptance occurs, then show how NOI-based principal repayment reacts to actual performance.

The company should also explain the tail. Its common NOI window ends in September 2036 with Barber Lake’s base term, while the AWS leases have years left under their initial terms. Those later cash flows may add value, but they should be presented separately rather than allowed to blur the definition of the ten-year portfolio average.

The scale slide is a useful map. The investment case will be settled by the clocks beneath it.

Sources