AI Hardware’s Customer-Financing Balance Sheet

AI equipment suppliers can sell two things at once: hardware and time to pay for it. The first produces a product margin. The second creates a financing asset, a funding requirement and a claim on future customer cash. As those businesses grow together, an operating-performance measure can improve while the cash committed to customers also rises.

Dell’s latest quarter makes that distinction unusually visible. HPE provides a useful comparison, but not a second copy of the same business model. The analytical task is to separate commercial performance, investment in customer credit and the funding of that investment—not to choose whichever cash-flow headline best supports an optimistic or pessimistic story.

Our conclusion is constructive: customer finance can expand access to useful AI infrastructure, but its scalability depends on the price and durability of the funding behind it. Adding financing investment back to a performance metric does not supply that funding.

1. Dell’s three cash-flow numbers

For the quarter ended July 31, 2026—Dell’s fiscal 2027 second quarter—the company reports the following bridge. Amounts are USD billions; free cash flow and adjusted free cash flow are Dell-defined non-GAAP measures.〔1〕

The two adjustments total $7.163 billion. This is a reconciliation, not an unexplained cash inflow. The financing-receivables item removes the period’s customer-financing cash-flow impact. The $0.496 billion item is Dell’s net capex-and-depreciation impact for operating and embedded leases. It includes noncash depreciation, so is neither a pure cash add-back nor the $0.451 billion increase in the net operating-lease-equipment balance.〔1〕〔2〕〔3〕

An equipment company that finances a customer can have a profitable sale and a delayed cash recovery. Measuring its commercial engine before that financing investment can be useful. Measuring its ability to fund the investment requires the cash-flow statement and financing arrangements as well. Neither perspective replaces the other.

The practical mistake is to treat the adjusted number as cash already received and freely available. The opposite mistake is to regard the investment in financing assets as proof that the underlying sale destroyed value. Customer credit may earn an adequate return; that is a proposition about future collections, funding costs and losses, not a conclusion contained in the adjustment.

2. Credit assets are not collections

Dell’s net financing receivables rose from $13.950 billion at the preceding quarter-end to $20.430 billion. That $6.480 billion balance increase is not the $6.667 billion cash-flow adjustment. The latter is also recoverable from the reported cash statements: the first-half adjustment of −$6.404 billion minus the first-quarter adjustment of +$0.263 billion gives −$6.667 billion for Q2.〔2〕

These are different records. A balance incorporates the accounting treatment of the asset pool, including applicable noncash movements; a cash-flow adjustment bridges profit to cash. Neither amount is gross loan originations or gross cash collected. A researcher should preserve the mismatch until the relevant reconciliation is available, not assign the difference to defaults, currency or asset sales without evidence.

Dell reports $7.5 billion of quarterly financing originations; its performance-review definition includes third-party originations. That does not establish that the entire amount financed AI servers.〔6〕 The 10-Q relates increased financing demand to infrastructure offerings and distinguishes sales-type leases, which create financing receivables, from operating leases, where equipment remains in its property-and-equipment base.〔3〕

This distinction changes what to examine. A receivable is primarily a claim on contracted customer payments. Equipment retained under an operating lease adds a different exposure to utilization, asset management and end-of-term economics. Contract terms can redistribute those risks, so neither label alone is a complete credit analysis.

The same caution applies to ordinary working capital. More inventory can reflect a larger production program, a deliberate procurement buffer or slower conversion into delivered sales. More receivables can reflect higher sales or slower collection. Supplier credit can finance a growing inventory position without reducing the underlying obligation to suppliers. Closing balances cannot identify which explanation dominates.

For AI hardware, the technology schedule makes this especially important as an analytical proposition. Procurement, system integration, power availability, acceptance and customer payment need not occur together. A delay can leave a commercially valuable order requiring additional financing. It can also expose an equipment owner to a product transition before the anticipated cash has arrived. Measuring those effects requires dated contract and deployment information, not an assumed link between every inventory dollar and obsolete hardware.

3. HPE requires its own bridge

HPE’s fiscal 2026 third quarter also ended July 31, 2026, but its fiscal-year label and year-to-date reporting window differ from Dell’s. Its quarterly cash reconciliation is:〔4〕

The simple $0.896 billion difference is not the company’s $0.958 billion measure. It is also not designed to be economically interchangeable with either of Dell’s free-cash-flow measures. Asset ownership, financing activity and the definitions of the deducted or added-back items matter.

The $0.745 billion investment is consolidated. HPE’s PP&E includes equipment held for lease; this cash-purchase line does not separate operating-company investment from leased-equipment investment. It must not be relabeled pure operating-company or AI-only capex.〔7〕

HPE reports Financial Services revenue of $0.883 billion and net financing receivables of $9.701 billion. Its fiscal 2026 presentation places Financial Services within Cloud & AI. That reporting structure does not make the financing portfolio exclusively AI-related, nor does the receivable balance equal total portfolio assets or the amount of new financing originated during the quarter.〔7〕 HPE’s preceding-quarter presentation defines net portfolio assets using both financing receivables and operating-lease assets, less related reserves.〔9〕

HPE’s quarterly financing-receivables cash-flow adjustment is −$0.506 billion: the nine-month −$0.224 billion less the six-month +$0.282 billion. This net adjustment, not gross originations, is already included in its operating cash flow. The published HPE FCF reconciliation does not add it back in the way Dell’s adjusted measure removes its own financing-receivables impact.〔7〕〔8〕

This is an important limit on a broad industry claim. The disclosures establish that customer finance is part of these suppliers’ activities. They do not establish a common AI financing intensity, identical credit standards or a uniform reliance on short-term funding. A financing-assets-to-AI-revenue ratio would be misleading if its numerator contains unrelated financing and its denominator includes only selected AI sales.

The right comparison is therefore a set of reconciled business questions: what was sold, which payments remain to be collected, which assets stay with the supplier, and which liabilities fund those exposures? A single cross-company cash-conversion ranking would conceal much of that structure.

4. Funding a receivable is a separate transaction

Dell initially cash-funds the DFS offerings reflected in its cash-flow adjustments and says that funding is largely subsequently replaced by DFS-related debt.〔3〕 The separately reported $7.5 billion originations metric includes third-party originations and is not identical to Dell-funded cash deployment.〔6〕 Dedicated funding arrangements and consolidated borrowing flows must therefore be examined separately.

The debt labels require care. Dell calls its $9.618 billion line DFS debt; “direct” is our distinguishing label. Applying a 7:1 debt-to-equity ratio to $23.612 billion of DFS owned assets gives approximately $20.660 billion of DFS-related debt. Deducting the $9.618 billion leaves $11.042 billion of allocated core debt—not another separately identified legal instrument.〔6〕 Dell describes a majority of the $9.618 billion DFS debt, not all DFS-related debt, as non-recourse. Its relevant securitization entities remain consolidated.〔3〕

That sequence matters. A loan or lease can be originated before the corresponding longer-term financing closes. The interim liquidity requirement exists even if management reasonably expects the asset to be financed later. Conversely, a company can arrange funding in advance and avoid that particular gap. The sequencing has to be observed.

Three quantities should remain separate:

1. Customer financing investment: principal deployed into new assets less cash principal recovered, subject to the exact cash-flow definition.
2. Funding cash: proceeds from borrowing, repayments and other financing transactions, recorded over the same period.
3. Funding exposure: the liabilities and obligations remaining after those cash flows, including their maturity and collateral terms.

None is a substitute for the others. A large gross borrowing can refinance an old liability without financing new customer assets. A net borrowing amount can hide substantial refinancing turnover. An accounting add-back can remove an asset-growth effect from a performance measure without changing either the funding cash or the outstanding exposure.

Asset transfers also need precise language. A transfer to a special-purpose entity, a sale recognized for accounting purposes, a secured borrowing and a transfer of economic credit risk are different questions. They should be answered from the specific documents. The words “securitized” or “non-recourse” do not, by themselves, prove that all risks or all possible parent obligations have disappeared.

This does not make customer finance suspect. It makes it a financing business. Its growth is most useful when the customer receives a payment schedule it can support, the supplier earns adequate compensation and the funding provider accepts clearly specified exposure. A structure that improves only one party’s reported metric has not yet demonstrated that result.

5. An illustrative cash bridge

The following model is independent of the companies above. Every amount is a hypothetical unit. It is a period-end bridge, not a forecast, a peak-liquidity calculation or an estimate of either company’s borrowing terms.

Let C be operating cash before the modeled customer-financing principal flows. This subtotal includes the other assumed operating receipts and payments, including interest and ordinary costs; those items are not charged again below. Let K be cash capital expenditure, O new cash principal originations and P principal collections.

After capital expenditure and customer financing, cash is:

Cash after investment = C − K − O + P

For this example, C=100, K=10, O=100 and P=30. The subtotal before customer financing and after capital expenditure is 90. Net principal investment is 70, leaving 20 before external financing.

If the principal flows are classified in operating activities, modeled CFO is 30 and CFO−K is 20. If the identical flows are classified in investing activities, CFO is 100 and CFO−K is 90, but investing activities contain the separate net principal outflow of 70. Total cash after both investments is still 20. This is a presentation experiment with fixed economic flows—not permission to choose accounting classifications or a conclusion about the appropriate treatment of an actual transaction.

Now add financing assets and their funding. Begin with assets of 300 and secured debt of 240. With originations of 100 and collections of 30, ending assets are 370. Assume all are eligible for the modeled borrowing base, the advance rate is 80%, and the supplier immediately uses the full ending capacity. Ending debt is therefore 296, producing modeled net borrowing of 56.

Cash movement before other financing and payouts is 20+56=76. With opening unrestricted cash of 20, ending cash is 96. The 90 subtotal did not become an equal cash surplus: under the assumptions, 14 of the net customer-financing investment remained funded without additional secured debt.

6. Why a stock-level funding change matters

Now reduce the advance rate from 80% to 60%, applying it to the entire ending eligible pool. Keep customer originations, collections and the operating subtotal unchanged. Continue to assume immediate rebalancing to full capacity.

Ending debt becomes 370×60%=222. Relative to opening debt of 240, the model implies net repayment of 18 to reach that assumed balance. Period cash movement falls from 76 to 2. The reduction is 74, not merely 20% of the net new assets of 70. The change affects the existing asset pool as well as its growth.

This is a contractual scenario assumption, not a claim that lenders to Dell or HPE can or will demand such a repayment. If an actual facility applies a new advance rate only to new assets, has fixed-term debt or permits a different adjustment timetable, the scenario must change.

Add a noncash asset reduction of 20 while retaining the 60% assumption. Ending assets are then 350; ending modeled debt is 210, with net repayment of 30. Period cash movement becomes −10. The reduction itself was not a cash principal payment. Its incremental modeled cash effect is 12, arising through the borrowing-base assumption.

The last case still has positive ending cash. It does not establish default, insolvency or a funding shortfall after all available resources. With zero opening cash, the same example would instead have a modeled ending gap of 10. Neither version reveals the maximum cash need inside the period.

Full utilization is an active assumption too. If the opening assets were 300 but opening debt only 150, with no originations or collections and an 80% ending advance rate, this rule would still create 90 of modeled borrowing. That is additional leverage, not cash produced by selling hardware. The model deliberately makes that choice visible rather than embedding it in a headline cash metric.

7. Designing finance that supports productive AI growth

The commercial objective should not be to maximize the amount of customer credit written. It should be to enable useful purchases on terms that remain workable across the customer, supplier and funding provider. That requires matching the instrument to the actual constraint.

Procurement and delivery finance. Where the gap is between component payment and customer acceptance, a short bridge may fit better than a long amortizing loan. Its sizing should follow dated cash requirements and realistic completion buffers. An order backlog is evidence of commercial demand only within its terms; it does not automatically provide cash on the date a supplier invoice falls due.

Customer payment flexibility. Where the customer benefits from equipment before generating the cash to pay for it, installments or leasing can align payment with use. The expected cash benefit must be assessed separately from the financing schedule. A longer payment term cannot repair a permanently uneconomic deployment, although it can make a productive deployment possible sooner.

Funding of retained customer assets. When a supplier keeps the receivable or equipment exposure, the funding arrangement should be evaluated alongside that asset’s collections, maturity and eligibility. The relevant questions include the share funded with durable capital, the cost of unfinanced exposure and what happens when assets cease to qualify. Those inputs belong in the economics of the financing product.

Technology and end-of-term exposure. Financing that retains equipment value risk needs an explicit view of reuse and recovery. A faster product transition may affect useful life or resale opportunities; better software compatibility or a credible secondary deployment may improve them. Neither direction should be assumed for every AI asset. The contract should make clear which party receives the upside and which bears the loss.

Risk transfer. Selling assets or obtaining credit support can redistribute an exposure, but the transaction has a price and conditions. Any representation, servicing, repurchase, guarantee or residual obligation should be examined before calling the transfer complete. A reduction in one company’s visible balance sheet is not, by itself, evidence of a reduction in the system’s total economic risk.

These designs can be growth-enabling. Their benefit is strongest when they reduce a real timing or risk-bearing mismatch at an acceptable cost—not when they merely shift an obligation out of a favored financial subtotal.

8. A repeatable research record

A useful continuing record would maintain three linked views for each issuer and reporting period. The first is the operating-cash and capital-expenditure bridge using the company’s exact definitions. The second is the customer-financing asset record: new originations, principal collections, net balances, noncash movements and any separately disclosed retained equipment. The third is the funding record: proceeds, repayments, remaining debt, commitments and relevant eligibility or maturity terms.

The dates matter as much as the labels. Quarterly flows should not be mixed with year-to-date amounts; beginning and ending balances should bracket the same period; a later financing announcement should not be inserted into an earlier cash statement. Missing allocations should remain missing. In particular, corporate or financial-services totals should not be relabeled AI-specific merely because AI demand is growing.

For every apparent improvement, ask a competing question. Did collection become faster, or did customer financing move to a different vehicle? Did funding become more durable, or did the company simply borrow more at the end of the quarter? Did the customer obtain productive capacity sooner, or just receive a longer period before payment? A credible answer needs an observable event, not only a narrative.

The SEC staff’s discussion of free cash flow is relevant here: the term has no uniform definition, and it should not be presumed to represent cash left after all mandatory uses. This report applies that analytical discipline; it does not allege a disclosure violation by either company.〔5〕

AI finance creates more possibilities when it connects a sound commercial engine to a durable funding structure. The customer-financing balance sheet is where that connection can be examined.

Sources and scope

Evidence cutoff: September 10, 2026, Asia/Singapore. Dell’s period is FY2027 Q2 and HPE’s is FY2026 Q3; both end July 31, 2026. Dollar amounts above are corporate or specified financing-business measures, not independently measured AI cash flows. Calculations in Sections 5–6 are independent assumptions, not company estimates. The companion model specifies the inputs and cash identities.

[1] Dell Technologies. Second Quarter Fiscal 2027 Financial Results. September 1, 2026. Non-GAAP cash-flow reconciliation.

[2] Dell Technologies. Q1 FY2027 Form 10-Q, filed June 9, 2026, together with the Q2 10-Q in source 3. Balance sheets and operating-cash-flow statements support the signed quarterly derivation; Q1 results, May 28, 2026, provide related summary tables.

[3] Dell Technologies. Form 10-Q for the quarter ended July 31, 2026. Filed September 8, 2026. Financing receivables, debt and liquidity discussions.

[4] HPE. Fiscal 2026 Third Quarter Results and Financial Tables. September 2, 2026. Cash-flow and free-cash-flow reconciliation, balance sheet and segment data; newsroom edition.

[5] U.S. Securities and Exchange Commission, Division of Corporation Finance. Non-GAAP Financial Measures, Question 102.07. Question dated May 17, 2016; page states last update December 13, 2022. Consulted September 10, 2026.

[6] Dell Technologies. Q2 FY2027 Performance Review. September 1, 2026. Debt and DFS definitions, pp. 27–29; related balances in the Q2 10-Q.

[7] HPE. Form 10-Q for the quarter ended July 31, 2026. Filed September 3, 2026. Cash-flow statement; Notes 1–2 and 5–6.

[8] HPE. Form 10-Q for the quarter ended April 30, 2026. Filed June 2, 2026. Six-month cash-flow statement, used with source 7 to derive the Q3 adjustment.

[9] HPE. Q2 FY2026 Earnings Presentation. June 1, 2026. Financial Services panel and net-portfolio-assets definition. Used for the stated definition, not a claim that a precise Q3 portfolio total was disclosed.