APLD: The Rent Is Real. The Equity Math Is Harder.

The 60-second read

Five things to know

1. Rental execution is becoming measurable. HPC base rent reached $65.8 million and property NOI was $58.8 million. The 250 MW October milestone gives the next quarter a larger operating base.
2. The revenue mix matters more than the headline. Tenant fit-out contributed $183.5 million, versus $65.8 million of HPC base rent; tenant recoveries add revenue with matching cost.
3. Positive operating cash flow does not fund the build. Operating cash of $63.9 million compared with $2.075 billion of purchases of property, equipment and other assets. Financing remains central.
4. Common shareholders sit below meaningful other claims. Debt was $6.377 billion, redeemable minority equity $2.011 billion, and parent preferred equity $181.3 million at quarter end. The preferred partnership return can compound.
5. Future power is an option, not delivered earnings. Finland potential capacity and the related-party Base Electron power agreement expand possibilities. Neither substitutes for contracted rent, completed facilities or funding.

Three scenarios, one probability-weighted number

Analyst scenarios, not issuer guidance; values per common share

Signposts: what would change our mind

These thresholds are Charged Alpha tests, not company guidance. The 300 MW year-end objective is management’s. We do not invent a prior packet score: the legacy Q4 episode predates this available packet record.

The tape and the reference price

The stock traded at $24.82 in the retained after-hours ranking snapshot, about 4.24% above the $23.81 regular close. Those are different observations: the after-hours move is the initial response since the print, while the regular session had already finished. We do not present either as a completed next-day earnings reaction. Prices can move materially while a packet is being produced; every valuation comparison here uses the stated reference rather than silently switching quotes.

The provider’s market capitalization was $6.855 billion, based on a share count that trails the issuer’s latest cover page. The October 6 cover count is 302.387 million shares; multiplying that count by $24.82 gives roughly $7.51 billion of common equity. That calculation is more suitable for our current valuation bridge, although it is still a basic share count and does not include every possible warrant, award or conversion. Our forward models use 360 million shares as an explicit dilution assumption, not a reported count.

The price history gives perspective without proving value. The provider’s 52-week range was $19.005 to $50.73 and reported beta was approximately six. Both describe a volatile security; neither makes a decline cheap by itself. We keep the daily series as reported and avoid constructing a false reaction series across changing business classifications. Historic earnings dates and market observations are retained in the research dataset, but an unverified event-to-session match is not promoted into an investment statistic.

The print: separate rent from construction services

Revenue of $341.875 million was a large increase from the current release’s comparable $80.934 million. That comparison includes ChronoScale in continuing operations. The earlier Q1 release originally presented a different continuing-business perimeter; comparing this quarter with the old headline without reconciling that change would exaggerate the analytical precision. The current release is the authority for the stated 322% increase.

Services revenue was $262.755 million and data-center rental and other revenue was $79.120 million. Services include tenant fit-out and the activities inside ChronoScale. Rental and other revenue includes base rent and tenant recoveries. Recoveries have corresponding reimbursable costs, so treating that component as high-margin incremental rent would be wrong. Fit-out activity can be important to the customer relationship and the build schedule while still deserving a different valuation from a long-lived rental stream.

The quarter generated a GAAP operating loss of $62.379 million. Services cost was $245.709 million and rental-related cost was another $43.862 million, producing computed gross profit of $52.304 million before selling, general and administrative costs. A provider feed omitted that rental cost from its gross-profit field and mislabeled continuing net loss as operating income. We corrected the analysis to the actual filing; those errors remain documented rather than being silently smoothed away.

We do not claim a clean earnings beat against a verified pre-release consensus. The retrieved annual estimates and targets were not all updated for this release, and the definition changes complicate any casual adjusted-EPS comparison. The more useful conclusion is operational: the rental platform is now contributing, but the quality and ownership of earnings require work below the headline.

The operating engines are different businesses

HPC hosting produced $262.6 million of revenue and $33.4 million of segment operating profit. Its revenue comprised $65.8 million of base rent, $183.5 million of tenant fit-out and $13.3 million of recoveries. The recurring rent component is economically attractive, but it was only about one fifth of consolidated revenue. Fit-out was roughly 70% of HPC revenue. The most informative next-quarter question is whether rent grows faster than the costs and capital claims attached to delivering it.

The legacy data-center hosting business generated $37.8 million of revenue and $13.3 million of segment operating profit on $111.9 million of segment assets. Its crypto-hosting sites were operating at full capacity, making it a useful earnings contributor rather than the main source of the reported growth. It also carries customer and commodity-cycle exposure that differs from the AI campus leases. A single technology multiple applied to every revenue dollar would miss this distinction.

ChronoScale contributed $41.482 million of consolidated revenue but is excluded from management’s core adjusted metrics. Applied Digital owned approximately 96% at quarter end. Its reported operating loss was $25.354 million, and the quarter included GPU hardware sales. We do not assign a large speculative value to the Microsoft deployment announcement or count that prospective business again inside the HPC real-estate valuation. Our modest noncore value assumption is deliberately separate.

At quarter end, live critical IT load was 175 MW. The October 1 delivery lifted it to 250 MW, and management expects 300 MW by the end of calendar 2026. The company’s leased footprint is much larger at 1.41 GW. Leased capacity, delivered capacity, power potential and options are four different measures. The $86 billion figure includes every renewal option; $36 billion is the initial base-term contracted revenue. Neither is current annual revenue.

Property returns versus the common-share result

The property-level economics are the strongest part of this release. HPC base rent of $65.8 million, plus the $0.854 million noncash lease-incentive adjustment and less property operating costs, taxes and insurance, produced $58.829 million of NOI. The reported margin was 89%. That metric helps assess an operating facility, but it excludes corporate overhead, financing, development spending and claims owed to preferred partners. It is not free cash available to common shareholders.

The adjusted net-loss reconciliation starts with a $168.001 million consolidated continuing loss. It adds back the ChronoScale operating loss and associated net interest, core stock compensation and employment taxes, transaction and litigation expenses, lease-incentive amortization, derivative and investment marks, restructuring and other adjustments. The result is a $4.053 million adjusted loss. That near-break-even figure uses a different business perimeter and different expense treatment from the $221.028 million continuing loss attributable to common holders.

The total common loss was $237.082 million after discontinued operations. Preferred dividends and the allocation or accretion associated with noncontrolling interests are consequential. A reader cannot simply apply the adjusted EPS to an equity multiple and assume the financing structure has disappeared. Noncash does not always mean economically irrelevant: stock compensation changes ownership, while preferred returns paid in kind can increase the claim ahead of common equity.

Below operating income, interest expense was $77.383 million, interest income $35.821 million, derivative losses $49.511 million and investment losses $11.352 million. The marks partly relate to Babcock & Wilcox securities. We do not extrapolate those market losses into recurring operating margins, but neither do we ignore the capital at risk. Separate operating progress, financing cost and investment volatility before judging whether the common-equity earnings engine has turned.

Nine earnings-quality checks

The quality scorecard is mixed because a construction ramp can create both real progress and noisy accounting. Operating cash flow became positive, yet a sizable increase in accrued liabilities helped the quarter’s cash conversion. Receivables also rose sharply. We therefore avoid declaring the cash model proven from one quarter of positive operating cash. The next reports must show collections, rental growth and operating cash recurring as the asset base expands.

GAAP stock compensation was $66.499 million, approximately 19.45% of revenue. That is larger than property NOI. The adjusted reconciliation removes a narrower core amount plus employment taxes because ChronoScale is separately excluded. Those two SBC numbers should not be compared as if one is an error. For a common shareholder, both the ownership effect and the cash taxes paid upon vesting matter even when the accounting expense is noncash.

Receivables of $114.110 million imply about 30.7 days when divided by the current quarter’s revenue and multiplied by 92 days. This is a rough consolidated indicator, not a billing-cycle measurement: fit-out milestones, leases and hardware sales do not share the same collection pattern. Inventory days are not meaningful for the central real-estate thesis and are not invented. The effective tax rate is also unhelpful in a loss quarter with valuation allowances and subsidiary-specific items.

The old quarters require caution. Cloud services moved between discontinued and continuing presentation over time. We preserve the nine original earnings releases and their reconciliation tables, and explicitly label the historical headline series as originally reported. We do not splice those numbers into a supposedly consistent growth curve. This discipline matters more than producing a superficially complete but misleading chart. The packet’s decision rests primarily on current source-verified rental, capital and ownership economics.

Funding is the link between delivery and dilution

Cash, cash equivalents and restricted cash totaled $3.678 billion at quarter end. Unrestricted cash was $2.950 billion; the remainder should not automatically be treated as freely distributable to shareholders. Debt carrying value was $6.377 billion. Using only unrestricted cash gives net debt of approximately $3.427 billion. Using all cash would reduce the number, but that is a more generous funding assumption and is not our primary equity bridge.

The company spent $2.075 billion on property, equipment and other assets in one quarter while generating $63.921 million of operating cash. The computed difference is negative $2.011 billion. This is predominantly a development-stage funding issue, not proof that a completed rented facility has negative economics. It nevertheless requires outside capital or existing liquidity. Financing cash inflow was $1.543 billion and aggregate cash declined by approximately $475 million.

The new $1.59 billion secured notes carry a 7% coupon and finance the third Polaris Forge 1 building while repaying a bridge facility. Lower-cost or longer-dated financing is helpful, but it still puts a contractual claim ahead of common equity. Cash interest paid of $242.852 million exceeded the income-statement interest charge; payment timing and capitalized interest matter during construction. We do not annualize a single quarter’s cash-interest payment as if it were the steady-state coupon bill.

Redeemable noncontrolling interest was $2.011 billion, and parent preferred carrying values totaled $181.302 million. The TopCo preferred units accrue at 12.75% annually, compound semiannually and have contractual step-ups under specified circumstances. Paying in kind can preserve current cash while raising the future claim. The partnership also owns common equity in a project structure. This is why our valuation reserves separately for future debt, preferred claims and minority participation rather than subtracting only today’s bank debt.

Three routes, with the funding assumptions exposed

Our base fair value is $20, with an $8–$40 scenario range and a probability-weighted value of $22. This is a HOLD with conviction three out of five and very high uncertainty. The reference price already requires substantial successful delivery. We would want an approximately 30% margin below base value, or $14 under unchanged assumptions, before calling the valuation attractive enough for a BUY. Better funding or operating evidence can change that threshold; a lower quote alone does not repair a broken thesis.

The first route capitalizes prospective rental income. We assume 1,410 leased MW ultimately generate $1.5 million of annual rent per MW with an 85% NOI margin. That produces $1.798 billion of stabilized NOI. A 7% capitalization rate gives $25.682 billion of property value. We subtract $10 billion of future net debt, $2.8 billion of preferred claims and a $1 billion reserve for minority and structural participation, then add $500 million of noncore equity value. Dividing by 360 million future shares and discounting three years at 12% gives $24.48 per share.

The second route assumes $1.6 billion of stabilized EBITDA at a 15-times multiple. Applying the same future claims, noncore value, shares and discount produces $21.16. These are independent operating yardsticks but share financing assumptions, so their agreement is not independent proof. The future debt and share counts explicitly include the risk that construction needs additional financing. They are analyst assumptions, not management forecasts or exact campus-by-campus financing plans.

The third route discounts cash actually available to common equity. We assume five annual cash flows of negative $250 million, negative $100 million, positive $100 million, $350 million and $650 million, followed by normalized $850 million growing at 3%. At a 14% equity discount rate and 360 million shares, the result is $12.36. This lower result shows how much terminal asset valuations can outrun near-term common cash generation. All debt service, preferred claims and dilution must already be reflected in those equity cash flows; subtracting debt again would double count.

The current economic enterprise-value bridge is about $13.12 billion after using the latest cover share count, adding debt, preferred and redeemable minority carrying values, and subtracting unrestricted cash. That is not a standard quoted EV. It is a transparent way to show the claims a casual market-cap comparison misses. Against quarterly adjusted EBITDA of $64.4 million, it demands a very large ramp; we do not use that temporary annualized multiple as the sole valuation. The sensitivity grid makes the key tradeoff explicit: stronger property valuations help, while each additional financing claim reduces what remains per share.

Wall Street: distinguish dated targets from a fresh response

The retrieved consensus was Buy, with thirteen buys and one hold, and an average price target of $61.65. The range was $22 to $90. That wide dispersion is informative about uncertainty, but it is not a probability distribution and does not demonstrate that the current print has been incorporated. The latest target-news records in our snapshot predate this earnings release. Calling the aggregate a post-earnings endorsement would overstate the evidence.

The dated examples illustrate the spread: UBS initiated at Buy with a $38 target on September 23; Redburn’s September 21 initiation carried a $22 target and Neutral stance; Wells Fargo’s September 17 initiation showed a $50 target and Overweight. We attribute these to the retained provider records, which themselves link to the original reports or news notices. We have not read proprietary analyst models and do not imply agreement with assumptions that were not available.

Our $20 base value is more cautious than that consensus. The difference is mainly the bridge from completed property economics to cash and ownership available to APLD common shareholders. A bullish analyst can reasonably use lower financing leakage, a faster build, more valuable campuses or optional expansion. Our case requires those improvements to appear in actual delivery, funding and dilution evidence before assigning them full value.

Annual provider consensus should also be treated cautiously. The FY2027 revenue estimate was roughly $758 million across six analysts and the FY2028 estimate roughly $1.900 billion across six. These are retrieved estimates, not verified post-release revisions, and they may reflect inconsistent business definitions. We show them as context, not as a clean beat/miss benchmark or a management promise. A useful next check is whether dated revisions explain both the rental ramp and the capital structure rather than only lifting the top line.

Management: delivery credit, financing accountability

Management deserves credit for moving capacity into service. Delivering the first 75 MW phase on July 1 and the second on October 1 is an observable operating result. It is more useful than a generalized statement about AI demand. The year-end 300 MW objective is the next clear test, and we separate the portion already delivered from the portion still expected. A credible operating record can lower the uncertainty premium, but only when the associated economics reach common shareholders.

Capital allocation is more complicated. The company has multiple funding layers, investments in other public securities and a related-party power-development relationship. Base Electron is approximately 10% owned by Applied Digital, which also had a $58.5 million demand-note principal receivable from it at quarter end. The October 4 power agreement concerns a prospective 1,200 MW gas-fired facility. It could solve a genuine constraint; it also adds execution, counterparty and related-party oversight questions.

The agreement provides a financing checkpoint on March 31, 2027. Until financing and construction advance, we treat the prospective power as contingent infrastructure, not operating rent. Finland’s potential power agreement is similarly an option on expansion rather than an immediate earnings asset. Assigning the same valuation to unbuilt optional capacity and an operating leased campus would erase the very execution work shareholders are paying management to complete.

Our questions for management are specific. How much incremental capital remains to deliver the already leased footprint? Which cash balances are restricted for particular projects? What share of stabilized cash will be available after preferred returns and minority participation? How much future equity issuance is embedded in the development plan? Finally, can management publish a consistent recurring-rent and cash-to-common bridge across quarters? Those answers would improve the model more than another large headline capacity number.

The risks that can change the answer

The highest-impact risk is a mismatch between delivery commitments and the capital required to meet them. Long-term leases can support financing, but they do not remove construction overruns, delayed power, equipment bottlenecks or customer acceptance conditions. When a project is late, financing costs can continue while rental cash is postponed. In a highly levered development model, that timing mismatch can matter more than a small change in a headline revenue growth rate.

The second risk is financing and dilution. Preferred accruals and common issuance can let the company build valuable assets while reducing the fraction of that value captured by each existing common share. This is not an argument that preferred financing is inherently bad. It is a reminder to judge the final per-share economics rather than the gross size of a campus. Our forward share count and capital-claim reserves attempt to reflect this uncertainty without pretending to know the exact future financing documents.

Customer concentration, hardware cycles and power availability remain important. A small group of hyperscale tenants can support bankable contracts but also create dependence. ChronoScale adds exposure to compute equipment and a different economic model from property leasing. The Base Electron relationship introduces a related-party concentration in a prospective power source. The market value of securities investments can create additional earnings volatility unrelated to rental execution.

The bull case is substantial: operating milestones arrive on time, the rent base expands, cheap reliable power becomes scarce elsewhere, and funding improves as campuses mature. The bear case is equally concrete: construction cash needs outrun funding, preferred claims compound, the share count rises, and rent arrives too late to protect common value. Our HOLD acknowledges both. We would change our view on evidence about rental cash, financing terms and ownership—not simply on a larger capacity announcement.

A dated decision calendar

The next operational checkpoint is the end of calendar 2026, when management expects 300 MW of delivered critical IT load. That objective gives investors a measurable test rather than an open-ended promise. It also establishes the right order of questions: first delivery, then rental recognition and collections, then the cash and equity claims associated with funding that capacity. A facility announcement by itself does not finish the sequence.

The next quarterly report should cover the period ending November 30, 2026. We expect the normal reporting window around January 2027, but the exact date is not confirmed in the sources reviewed here. That report should capture more of the October capacity addition. We will compare base rent with the $65.8 million starting point, property NOI margin with the reported 89%, and common shares with the October 6 cover count. We will also look for working-capital reversals rather than assuming positive operating cash repeats automatically.

March 31, 2027 is the specific financing checkpoint associated with the Base Electron power agreement. A timely financing close would advance the project, while a missed close or termination would change the probability assigned to that power pathway. Neither outcome alone determines the value of existing operating campuses; the distinction prevents optional expansion from dominating the entire thesis.

The valuation can change before any scheduled date. New financing, changes in preferred rights, material tenant developments or unexpected share issuance can alter the residual value per share immediately. Our signposts are therefore a monitoring framework, not a promise to ignore intervening information. At the current reference price, the burden of proof remains on stronger cash-to-common economics. At a lower price, the same risks still need to be checked before applying the stated margin-of-safety entry level.

Financial history, sources and method

The research archive for this new quarter contains the actual SEC earnings exhibits, periodic reports and their complete extracted tables. The historical headline table below uses the values originally published for each quarter. It intentionally does not claim consistent continuing-business definitions: the cloud business was reclassified across periods and was later consolidated through ChronoScale. For the current year-over-year comparison, use the current Q1 FY2027 release’s $80.934 million prior-year revenue, not the old $64.216 million headline.

Quarterly share counts and balance-sheet values must also be read with their dates. The August 31 balance-sheet count was 297.067 million, weighted-average shares were 291.558 million, and the October 6 cover count was 302.387 million. These are not interchangeable. The first is a point-in-time balance-sheet measure, the second is an earnings-per-share denominator, and the third is the latest disclosed outstanding count used in our current equity bridge. The 360 million future model denominator is an explicit analyst assumption.

The three-year cash table uses the fiscal 2026 annual report, whose comparative cash-flow statement covers fiscal 2024 through 2026. Free cash flow here means operating cash less purchases of property, equipment and other assets; it is a computed analytical measure and includes growth construction. It is not the company’s adjusted EBITDA or NOI. All monetary tables are in US dollars, so no foreign-exchange conversion is applied. Missing or nonmeaningful historical measures are not filled with invented values.

The current auditor identified in the annual report is CBIZ CPAs, with predecessor Marcum history disclosed. The annual report lists approximately 256 full-time employees at May 31, 2026. The security is US-listed common stock, not an ADS; there is no ADS ratio in our calculations. Applied Digital consolidates variable-interest project entities and its controlled ChronoScale subsidiary, so consolidated assets and revenue do not imply unencumbered ownership of every dollar of cash flow. We incorporate preferred and minority interests explicitly.

The prior Q4 episode framed the company around backlog and debt. This quarter’s thesis is different: the recurring-rent contribution can now be measured, and the question is how that operating progress converts into value after funding and ownership claims. No numerical performance score is invented for a legacy packet that is not available in the current packet record. The dated signposts in this report establish a transparent basis for the next review.

Primary sources. Q1 FY2027 earnings release, Exhibit 99.1; August 31, 2026 Form 10-Q; FY2026 Form 10-K; issuer release. Historical quarters use their dated SEC earnings exhibits. Market prices, target records and consensus context come from the retained FMP snapshot; they are not issuer guidance.

Glossary. NOI is property-level net operating income. Fit-out is work performed to prepare space for tenants. PIK means paid in kind, which can increase a financing claim without a current cash payment. A capitalization rate converts annual property income into an indicated asset value. FCFE is cash flow to common equity after the assumed financing and ownership costs. A scenario probability is an analytical judgment, not a frequency promised by the market.