A billable megawatt and a productive megawatt answer different financial questions. The first establishes that an infrastructure owner has a customer obligation generating income. The second concerns the economics of what the customer does with the capacity. Confusing them can make contracted infrastructure growth look like evidence of returns throughout the AI stack.

DayOne's 5 October Form F-1, followed by its 6 October IPO announcement, makes this distinction unusually testable. It discloses the capacity definitions, the timing of billing relative to power consumption, and the cash and financing burden behind expansion. The proposed IPO has no determined offering size or price range; this paper makes no valuation judgement.

What 99.1% utilisation establishes

At 20 September, DayOne reported 962MW of capacity in service and 953MW of Billings. Its utilisation rate is Billings divided by capacity in service: 99.1%. Billings means customer-committed capacity already generating income under legally binding contracts. It is a measure of commercial occupancy, not GPU activity, token demand or customer profitability.

That is strong evidence for the landlord. It shows completed capacity becoming income-generating rather than waiting for a customer. It cannot establish the economics of the equipment inside: customers own and operate their IT equipment, while DayOne supplies space, power and cooling. The portfolio serves cloud and AI workloads; the filing does not isolate AI-only capacity or revenue.

The timing difference is observed, not hypothetical. DayOne says substantially all Billings begin generating revenue when capacity is ready for service. In the third quarter of 2025, 156MW commenced billing while customer power consumption was still ramping. Service income and physical use therefore need not rise at the same pace.

Contracting can place some utilisation risk with the customer before that customer's returns are known. This does not imply idle GPUs, artificial demand or poor tenant economics. It identifies what the landlord's demand signal measures—and where its evidentiary reach stops.

The next capacity is committed, but not yet earning

The same September table reports 2,281MW of Bookings, comprising 953MW of Billings and 1,328MW of non-income-generating Backlog. DayOne separately reports 1.1GW of Reservations. Those reservation rights sit outside Bookings, Billings and Backlog; customers may never exercise them. Adding them to firm income-generating capacity would erase a consequential distinction.

DayOne expects substantially all current Bookings to be delivered by the end of 2028 and estimates approximately US$11.4 billion of remaining completion cost, after expenditure through June. It believes available capital resources can fully fund current Bookings. These are management estimates, not proof of completed delivery or an independent finding of a funding gap.

For the owner, the intervening test is delivery and collection. Contractual resource access can still be delayed, sometimes with limited recourse against utilities. Customer contracts permit early termination with specified costs and penalties, and contain remedies for delivery or service failures. The filing warns that termination compensation may not replace the revenue lost. A long term is protection with conditions.

A development yield is a project measure

DayOne estimates a mid-teens development yield across booked projects. It defines that yield as annualised adjusted gross profit at stabilisation divided by cumulative capital expenditure and estimated completion cost. Adjusted gross profit excludes depreciation and share-based compensation within cost of revenue.

The measure can test the operating earning power of completed infrastructure against its construction cost. It does not measure equity returns after corporate overhead, interest, tax, working capital or subsequent capital needs. Nor is a stabilised estimate a realised portfolio result.

This boundary matters during construction. First-half 2026 total interest cost was US$145.4 million, of which US$83.1 million was capitalised; reported interest expense was US$62.3 million. The amount charged through the income statement is therefore narrower than the financing cost incurred. Capitalised interest belongs in the project's cost and eventual recovery test; it must not be subtracted again from a cash-flow figure that already captures its payment.

Billing growth and cash recovery are different checkpoints

First-half revenue rose 238% to US$512.0 million as June Billings reached 666MW, compared with 213MW a year earlier. Operating cash flow improved to a positive US$14.0 million. Against that, payments and prepayments for property, equipment, land-use rights and construction deposits were US$3.11 billion.

This is a growth-funded construction programme, not a like-for-like test of mature-site free cash flow. The expenditure spans expansion and advance payments; the September capacity snapshot is later than the June financial period. Dividing one period's cash flow by another date's bookings would not produce a defensible asset return.

There is meaningful counter-evidence to a distress reading. DayOne reported US$1.98 billion of June cash and upsized its Johor loan facilities in September. The financing nevertheless keeps delivery financially consequential: the parent undertakes to provide equity to complete financed projects and fund related development overruns. Contracted rent reduces demand uncertainty without eliminating completion exposure.

Collections also depend on the contractual payer. The largest customer supplied 69.2% of first-half revenue, but counterparties are typically local affiliates. The filing cautions that recourse to the broader parent may be limited, except where affiliate guarantees apply. A global customer's identity cannot substitute for the actual obligor and protection.

What would change the assessment

  • Delivery to billing: backlog becomes income-generating capacity within budget and timetable, without material service credits or contract reductions.
  • Billing to cash: collections keep pace with service fees, with working-capital needs and customer concentration disclosed.
  • Project earnings to investor recovery: stabilised results reconcile to completion cost, total financing cost, overhead and subsequent capital requirements.
  • Tenant economics: independent evidence of workload use and customer monetisation supports the capacity commitment. The landlord's utilisation rate cannot provide this evidence alone.

Timely delivery and durable collections would strengthen the infrastructure investment case even before downstream compute returns become transparent. Weaker customer economics would matter through credit, contract renegotiation and future bookings; it need not immediately appear in today's billed utilisation.

BoomRisk's live composite was 3.125, displayed as 3.13/5 Elevated cycle risk, when checked during this publication run. This issuer-level analysis changes no score, weight or methodology. Its contribution is three separate tests: capacity bills; customer use earns; the full capital structure recovers its cash. Evidence at one checkpoint must not stand in for evidence at all three.

Sources and methodology

  • DayOne — Form F-1, filed 5 October 2026Definitions and KPI table, pages 20 and 79; customer and termination risks, pages 30–31; non-GAAP reconciliation, pages 18 and 84; billing/power explanation, page 86; cash flows, liquidity, interest and funding undertakings, pages 88–91; customer contract terms, page 113. Financial results are for the six months ended 30 June; capacity observations are dated 20 September unless stated otherwise. Interim financial statements are unaudited. Dollar amounts are US dollars.
  • DayOne — proposed IPO announcement, 6 October 2026Confirms filing and proposed listing; share count and price range were undetermined. No assumed IPO proceeds or reported valuation enters this analysis.
  • BoomRisk methodologyFixed rules for the composite; issuer research does not override scored inputs.
  • BoomRisk live dashboardPoint-in-time reading checked on 8 October 2026 SGT; 3.125 rounds to 3.13.

BoomRisk distinguishes issuer disclosures, management estimates and editorial inference. Utilisation and development yield use DayOne's definitions; adjusted gross profit is non-GAAP. No customer compute-utilisation or AI-only return series is inferred. This research is not an investment recommendation, IPO assessment or market forecast.