AI Prepayments Fund Contracts. The GPU Pool Still Has to Earn.

A customer paying before delivery changes the financial test. It can reduce the capital an infrastructure supplier must finance before billing begins. It is stronger evidence of commitment than a prospective order or an annualised revenue aspiration.

Megaport's 29 September ASX filing supplies that evidence. Three new agreements through Latitude.sh carry about A$978.6 million in total contract value, with a weighted average term just over four years. They provide committed revenue irrespective of usage. Customers have committed approximately A$322.6 million in prepayments against A$500.3 million of equipment investment.

The new orders nevertheless do more than absorb capacity. Megaport is assigning GPUs previously ordered for its on-demand pool to those contracts, then procuring replacement equipment. The financial distinction is between capital supported by specific customer commitments and capital deployed to recreate capacity for future customers.

What the prepayments improve

One new customer is making an approximately A$281.5 million prepayment before service delivery to secure rapid access to already-ordered GPUs. Megaport says this covers most of the cash capex for that customer's contract. Across the three agreements, the disclosed prepayments equal about 64.5% of the stated equipment cost.

That percentage is BoomRisk's calculation, A$322.6 million divided by A$500.3 million. It measures the scale of committed customer funding relative to disclosed equipment investment; it is not a return, margin or claim that all the money has already arrived. The announcement does not provide a complete receipt-and-payment schedule.

Usage-independent commitments also move some demand risk away from the supplier. A customer's lower utilisation does not, under the disclosed fixed-term commitment, automatically reduce contracted revenue. The supplier still has to deliver the service, and payment remains dependent on the customer's ability to meet its obligations.

The customer mix adds useful evidence. Two contracts are with US technology providers supported by institutional shareholders; the third is with a publicly listed enterprise using the capacity for its own AI workloads. That direct end-user commitment matters. The counterparties are unnamed, so the public record does not permit an independent assessment of their credit or upstream funding. Institutional backing alone is not a guarantee.

The GPU allocation creates a second capital decision

Megaport estimates A$360 million of capex to replenish the on-demand GPU pool. It expects the replacement equipment to be procured and deployed by Q4 FY27, with revenue ramping in H1 FY28. Procurement and deposit payments have begun; remaining equipment payments fall due on shipment.

This is a different underwriting question. Allocating already-ordered GPUs to fixed-term customers can accelerate earnings and reduce exposure on those assets. Buying replacements restores the pool's capacity—and its dependence on subsequent demand, utilisation, pricing and operating costs.

Megaport describes the pool's purpose as attracting customers needing immediate infrastructure and converting shorter-term usage into longer-term deployments. The new contracts are evidence that this strategy can work. They do not establish that every replenishment cycle will produce the same customer commitment on the same terms.

The A$360 million programme must not be added mechanically to the announced A$500 million increase in FY27 capex guidance. Megaport says the guidance increase tracks the new contracts' equipment cost, and its updated A$1.78–1.88 billion group capex guidance includes both the strategic contracts and pool replenishment. The filing supports a distinction between investments, not an additional A$860 million increase in guidance.

EBITDA payback leaves other claims on the cash

The pool retains a 16–22 month EBITDA payback target. Megaport defines this as the estimated period for cumulative pool EBITDA to exceed the associated investment capex. It explicitly makes the target dependent on utilisation, duration of usage, hardware mix and operating costs.

This is a target for a specified operating measure, not demonstrated recovery of cash after every claim. EBITDA excludes interest, tax, depreciation and amortisation; Megaport's definition also excludes several other costs. The target does not by itself show repayment after financing costs, cash taxes, working-capital movements or later investment.

Depreciation is non-cash, but the eventual replacement of equipment can require cash. The distinction matters when today's contracted GPUs are expected to keep earning after their initial terms while new GPUs are being purchased to serve the next cohort. Neither useful life nor post-contract revenue is secured by a payback target.

A fast EBITDA payback would be economically meaningful if achieved. BoomRisk's question is whether that operating recovery also leaves enough cash to support the financing and repeated capital deployment required by the wider platform.

Fully funded describes a conditional capital plan

Megaport reports an executed bank facility upsized to A$845 million and approximately A$362.2 million of pro forma liquidity after the updated programme. The latter combines cash, equity proceeds, debt capacity and customer prepayments, less planned capex. It is not A$362.2 million of cash at bank.

The filing says immediately available drawdown covers current cash-flow requirements. Access to the full facility grows as revenue milestones are achieved and remains subject to agreed conditions and covenant limits. Its claim that the programme is fully funded must be read with those terms.

That creates an execution link worth monitoring: delivery produces billable revenue; revenue milestones expand financing availability; funding supports further deployment. The disclosed arrangement does not demonstrate a funding shortfall. It does make the pace of conversion from committed capacity into billing relevant to the capital plan.

The evidence against a weak-demand interpretation

Earlier strategic deployments are already provisioned, customer-accepted and billing. The new contracts are expected to begin contributing as hardware becomes operational through Q3 FY27 and reach their full run rate in Q4. Megaport also says it has secured power and space for both the new contracts and replacement pool equipment.

Those disclosures deserve weight. Prepayments, fixed commitments, an enterprise end-user and actual billing strengthen the case that this buildout is responding to customers. This paper does not infer that Megaport's demand is artificial, that prepayments are circular or that its financing signals distress.

The reporting boundary remains important: pro forma group ARR includes unbilled committed revenue. It therefore cannot be treated as current recognised revenue or used directly to prove realised returns on the pool. A growing contractual pipeline and a profitable completed investment are different observations.

What would change the assessment

• Cash receipt and delivery. Confirmation that committed prepayments arrive alongside equipment payments, and that customer acceptance starts billing within the disclosed deployment window.
• Pool conversion. Evidence that replacement GPUs achieve utilisation and pricing consistent with the payback target, including the balance between short usage and fixed-term contracts.
• Cash recovery. A bridge from pool EBITDA to cash after interest, tax, working capital and subsequent equipment spending, with the relevant asset cohort identified.
• Financing availability. Revenue milestones met, debt actually drawn, remaining availability and covenant headroom through the deployment and revenue ramp.
• Counterparty durability. Further disclosure allowing customer credit, payment protections and concentration to be assessed without assuming that institutional backing transfers the risk.

Timely acceptance, collected prepayments, durable pool conversion and cash recovery after financing would strengthen the investment case. Delayed billing, weaker pool economics or reduced borrowing headroom would weaken it. Neither outcome follows from the contract headline alone.

This issuer-level analysis does not change BoomRisk's rules-based composite. The live reading checked on 30 September was 3.125, displayed as 3.13/5 Elevated cycle risk. Research can expose a financial dependency before evidence qualifies for a scored input.

Megaport's customer commitments reduce a real part of the buildout's funding risk. The next test sits in the capital deployed to replenish the pool: whether recurring capacity investment can recover its cash while keeping the funding chain intact.

BoomRisk's analysis distinguishes company disclosures, company targets and editorial inference. ARR and EBITDA are non-IFRS measures. Prepayments are commitments rather than a verified schedule of cash received; payback and future revenue are targets or expectations. This research is not an investment recommendation or market forecast.