[APLD] Applied Digital: Q1 Earnings Preview — Will the Second Building Lift Base Rent?

Applied Digital (APLD) builds high-density, liquid-cooled AI data center campuses and leases whole buildings to CoreWeave and investment-grade hyperscalers under take-or-pay leases of roughly 15 years[1]. The Applied Digital Q1 earnings call is scheduled for 2026-10-08, when the company reports the fiscal first quarter of 2027, ended August 31, 2026[2]. The latest full disclosure is the fiscal 2026 annual report filed on July 29, 2026, for the year ended May 31, 2026, which does not break out the fourth quarter: full-year revenue rose to $611.3 million from $228.6 million a year earlier[3], only $99.8 million of it was HPC base rent[4], the operating loss was $236.5 million[5], and at year-end the company had signed about 1,410 MW of critical IT load worth roughly $36.2 billion of contracted revenue over the initial 15-year terms[1]. The only forward-looking figures from the company are two statements management made on the fiscal fourth-quarter call on July 27, 2026, both taken from a summary of that call: capital expenditures of about $600 million in the coming quarter, rising further as construction advances at new campuses[6], and a $1 billion annual net operating income run rate that management said it would reach within one year of that call, three years ahead of the original plan[7]. On the analyst side, the 6 analysts compiled by FXEmpire project first-quarter revenue of $137.1 million and earnings per share of -$0.27, and the same page records prior-quarter revenue of $258.75 million[8]; the Drillr earnings calendar showed lower estimates on the same day, with revenue of about $116.9 million and earnings per share of about -$0.30[2], a gap of roughly 15% that makes the single-quarter figures a poor precise reference. The 5 analysts compiled by stockanalysis.com project fiscal 2027 revenue of $837.7 million[9].

Three things are worth watching in this report. The first is whether the second building at Polaris Forge 1 has started paying rent: the $99.8 million of fiscal 2026 base rent roughly equals three fiscal quarters of rent on the first 100 MW building, or about $46 million for 100 MW in a full quarter[4], while the annual report says only that the second, 150 MW building is "partially operational"[10], so first-quarter base rent clearly above $46 million would show that energized space has turned into rent. The second is how thick revenue remains once tenant fit-out work fades: the $270.6 million of fiscal 2026 fit-out services revenue is recognized at cost plus a markup[11] against $258.1 million of cost[12], the quarterly revenue analysts project is well below the prior quarter, which is consistent with that revenue falling away, and the real question is whether the two recurring lines, rent and hosting, are themselves growing. The third is who pays for the construction: management's figure for quarterly capital expenditures is about $600 million[6] while minimum contracted lease payments for fiscal 2027 are only $451.1 million[13], so any new issuance of Series G convertible preferred stock or common stock in the quarter, together with the change in net interest expense, will show whether project-level financing can carry the company to the rent ramp on its own.

Company Background and Business Structure

Applied Digital is an AI data center developer and landlord that grew out of crypto-mining hosting. The company is based in Dallas, Texas, trades on Nasdaq and ends its fiscal year on May 31; it was formerly called Applied Blockchain, took its current name in November 2022, and began by providing energized hosting sites for crypto miners in North Dakota, a business it now describes as a legacy line[14]. Over the past two years it has redirected the same ability to find cheap power and build quickly toward AI compute, using a standardized campus design that delivers a building of about 150 MW in roughly 14 to 18 months[15], and it calls these campuses AI Factories. As of May 31, 2026, the company employed approximately 256 full-time employees[16], reported two segments, and held about 96% to 97% of ChronoScale, a separately listed cloud services company that it continues to consolidate[17][18].

The HPC Hosting Business develops and leases AI data centers, and it accounts for all of the company's growth. As of May 31, 2026, the company had signed about 1,410 MW of critical IT load across five campuses, representing roughly $36.2 billion of contracted revenue over the initial 15-year terms: Polaris Forge 1 in North Dakota is leased to CoreWeave for 400 MW and about $11.0 billion; Polaris Forge 2 is leased to an investment-grade hyperscaler for 200 MW and about $5.0 billion; Polaris Forge 3 and Delta Forge 1 in the South are 300 MW and about $7.5 billion each, both leased to high investment-grade hyperscalers[1][19]; and Delta Forge 2 is 210 MW and about $5.2 billion under a lease signed on June 5, 2026[20]. Only the first 100 MW building at Polaris Forge 1 and the partially operational second building are actually running[10], Polaris Forge 2 is expected to reach initial capacity in the second half of calendar 2026 and full capacity in early calendar 2027[21], and the remaining campuses are due for delivery between the second half of 2027 and the second half of 2028[19]. The business earns two kinds of revenue: rent, which begins once a building is ready for use and the tenant takes possession[22], and fees for fitting out tenant space, which are recognized at cost plus a contractual markup[23].

The other two businesses are smaller, but both affect the consolidated statements. The Data Center Hosting Business operates about 286 MW across two North Dakota sites, Jamestown (106 MW) and Ellendale (180 MW), serving a single crypto-mining customer under a contract with about one and a half years remaining[14], and both sites were still running at full capacity as of May 31, 2026[24]. ChronoScale is the cloud services business the company separated on May 5, 2026; its shares trade on Nasdaq under the symbol CHRN, and it rents space in third-party facilities in Colorado, Minnesota and Utah to provide compute to customers[17]. ChronoScale is no longer a reporting segment, but its results remain consolidated, and it posted an operating loss of $37.0 million in fiscal 2026[25]. Customer concentration is high: one HPC customer contributed 59% of fiscal 2026 revenue and one hosting customer contributed 25%[26], and every site the company currently operates is in North Dakota[27].

Financial History and Current Position

Applied Digital's revenue has more than tripled over three years, yet the company has lost money every year. Annual revenue grew from $165.6 million in fiscal 2024 to $228.6 million in fiscal 2025 and $611.3 million in fiscal 2026[3], while net losses were $149.7 million, $231.1 million and $184.3 million respectively[5]. The jump in fiscal 2026 came mainly from the first year of HPC operations, including $270.6 million of fit-out services revenue[11] and $114.7 million of rental and other revenue, made up of $99.8 million of base rent and $14.9 million of tenant recoveries[4]; hosting contributed $154.4 million, compared with $144.2 million and $136.6 million in the two prior years[28].

On a GAAP basis the fiscal 2026 loss widened at the operating level, and the cause sits at the corporate level rather than in the segments. The operating loss grew from $72.2 million to $236.5 million, and the net loss attributable to common stockholders was $250.3 million[5]; selling, general and administrative expense rose from $107.9 million to $332.1 million, including an increase of about $198.3 million in stock-based compensation[29], and the cloud business recorded a $59.7 million held-for-sale impairment. On the company's own non-GAAP definitions, adjusted EBITDA was $107.2 million[25], and net operating income on HPC base rent was $90.4 million, a 91% margin[30].

Cash flow and the balance sheet show a company at the peak of its construction spending. Operating cash flow swung from an outflow of $115.4 million to an inflow of $89.7 million in fiscal 2026, while cash purchases of property and equipment rose from $681.6 million to $2,865.8 million[31] and financing activities brought in $6,876.8 million[32]. At year-end the company held $1.6 billion of unrestricted cash and $2.6 billion of restricted cash[33], net long-term debt of $4,959.5 million against $677.8 million a year earlier[34], and construction in progress of $2,776.2 million against $1,090.6 million a year earlier[35]. The company has published no new quarterly financial data since the annual report, so the 2026-10-08 disclosure will provide the first single-quarter figures of fiscal 2027.

Operating Model

Revenue comes from four sources of very different quality, and the variable that determines future revenue is the number of megawatts delivered and paying rent, not the number signed. The first source is rent on HPC campuses: base rent is recognized on a straight-line basis once a building is ready for use and the tenant takes possession[22], most operating costs, including power, are recovered from tenants and also counted as revenue[36], and this line totaled $114.7 million in fiscal 2026[4]. The second is fit-out services revenue, recognized at cost plus a markup[23]; at $270.6 million it was the largest source of the fiscal 2026 jump, but it is project-based and stops when a building's fit-out is finished[11]. The third is legacy crypto-mining hosting, with two sites and 286 MW at full capacity generating $154.4 million[28]. The fourth is the still-consolidated ChronoScale cloud business, which works out to about $71.6 million when fit-out and hosting revenue are subtracted from consolidated services revenue of $496.6 million[3]; the annual report says this revenue fell by $12.4 million year over year because of lower cloud service rates[37]. The lag from delivery to revenue is roughly the 14 to 18 months it takes to build one building[15], and signing a lease by itself generates no revenue.

The profit structure is thin at both ends and thick in the middle, and segment profit does not yet cover corporate-level costs. Rent is thick: $99.8 million of base rent produced $90.4 million of net operating income[30], because tenants bear power and similar costs and non-recoverable operating expenses were only about $8.5 million[38]; after the depreciation included in rental cost of revenue, HPC segment profit was $39.1 million, compared with a loss of $12.1 million a year earlier[39]. Fit-out services are thin, with $270.6 million of revenue against $258.1 million of cost[12], a gross margin of about 4.6%, and hosting sits in between with segment profit of $48.3 million[40]. The two segments together earned $87.5 million while the consolidated operating loss was $236.5 million; the gap comes from $332.1 million of selling, general and administrative expense[29], ChronoScale's $37.0 million operating loss[25] and the $59.7 million impairment, and after deducting the share attributable to redeemable noncontrolling interests, such as distributions on Macquarie's preferred equity, and preferred dividends from the $184.3 million net loss, the net loss attributable to common stockholders was $250.3 million[5]. Each additional dollar of rent therefore lifts net operating income almost one for one, but whether that narrows the operating loss depends on whether corporate-level costs come down at the same time.

The main thread of cash flow is that money goes out first and rent comes in later, so cash recovery clearly trails cash investment. Operating cash flow turned positive at $89.7 million in fiscal 2026, but cash purchases of property and equipment reached $2,865.8 million[31], which the annual report attributes to higher construction-related payments[41], and another $556.4 million of construction costs sat in payables at year-end; financing inflows for the year were $6,876.8 million, including $1,825.0 million from Macquarie's purchase of subsidiary preferred equity[32]. The $2.6 billion of restricted cash at year-end is mainly proceeds from project notes and can only be spent on the corresponding campuses[33]; the core of the debt is $2.35 billion of 9.25% notes due December 2030 and $2.15 billion of 6.75% notes due March 2031[34], and on June 16, 2026, after the fiscal year ended, the company issued $1.59 billion of 7.00% notes to refinance a bridge facility[20]. On the other side, the minimum contracted lease payments listed in the annual report are only $451.1 million for fiscal 2027 and $1,453.7 million for fiscal 2028, with a 15-year total of $35,844.9 million[13]; this is a cash-basis contractual schedule that cannot be reconciled quarter by quarter with straight-line GAAP rent, and it is not revenue guidance.

Industry and Competitive Position

The AI data center industry is currently constrained on the supply side, and competition turns on who can secure power, finish buildings and raise capital fastest. The industry estimates the company cites in its annual report are that hyperscaler capital expenditures on AI infrastructure will exceed $700 billion annually by 2026 and that U.S. data center construction spending has tripled since 2022[42]. The annual report reduces competition to four things: securing large-scale, reliable and cost-competitive power and interconnection, designing and delivering high-density facilities, attracting and retaining customers, and accessing capital on attractive terms. It groups rivals into established data center operators such as Digital Realty and Equinix, hyperscalers that build their own capacity, and power-advantaged developers that moved from crypto mining into HPC, naming IREN, Cipher, TeraWulf, Hut 8, Riot, CleanSpark, Core Scientific and Galaxy Digital among others[43].

Applied Digital's advantages are a repeatable construction capability and the top-tier tenants it has already signed, and its weaknesses are small scale and high concentration. The company positions itself around a standardized, repeatable campus design that delivers a building of about 150 MW in roughly 14 to 18 months[15], and it has already signed about 1,410 MW of long-term leases with CoreWeave and investment-grade hyperscalers[1]. Its weaknesses are equally clear: its size and balance sheet are far smaller than those of established operators, it has only three tenants, and all of its operating assets are in North Dakota[27]. The comparison available in the current material has clear limits, because the annual report names rivals without giving unit construction costs, rent levels or delivery times for any of them, so where the company's build speed and cost of capital stand relative to peers can only be described qualitatively for now.

Core Debates

Only a small slice of Applied Digital's 1,410 MW of signed leases is paying rent today — will the second building at Polaris Forge 1 show up in this quarter's base rent?

This debate matters because Applied Digital's main tension is timing, not demand. The leases are take-or-pay and non-cancellable, and a tenant that terminates for convenience must pay the full remaining contractual value[44], so whether tenants will pay is not the central issue; the company, however, only begins recognizing rent once a building is ready for use and the tenant takes possession[22], a signed but undelivered building produces no rent, and that is why base rent recognized in fiscal 2026 was only $99.8 million against roughly $36.2 billion of contracted revenue across five campuses[1][4]. The financial transmission is direct: once a building is delivered and the tenant takes possession, rent-paying critical IT load rises, straight-line base rent and tenant recovery revenue rise with it, and because non-recoverable operating expenses are small[38], net operating income and HPC segment profit increase almost in proportion. The 91% net operating income margin on this rent is the company's only thick and durable source of profit[30], and the speed of its ramp determines when the company-level loss starts to narrow.

The available evidence points to higher rent, but it does not settle which quarter the increase lands in. The annual report says the first building at Polaris Forge 1, about 100 MW, became operational in October 2025, the second, 150 MW building is partially operational, and the third, about 150 MW, is under construction with a ready-for-service date in calendar 2027[10]. The $99.8 million of fiscal 2026 base rent was recognized only from the second fiscal quarter and roughly equals three fiscal quarters of rent on 100 MW; prorating the Polaris Forge 1 contract of about $11.0 billion, 400 MW and 15 years gives average rent of about $46 million for 100 MW in a full fiscal quarter[1]. The minimum contracted lease payments the company lists for fiscal 2027 are $451.1 million, far above the rent recognized in fiscal 2026[13], which implies that the company's own delivery plan has several buildings starting rent during fiscal 2027. An alternative reading is equally valid: partially operational can mean only energization and commissioning, rent starts when the tenant formally takes possession, and if that happens late in the quarter base rent could barely move; the $451.1 million is also a cash-basis contractual minimum that does not map quarter by quarter onto straight-line GAAP rent. The annual report neither breaks out the fourth quarter nor gives the operating megawatts of the second building, so the only usable reference today is the full-year figure, and operating capacity can only be stated as 100 MW plus an unquantified increment.

The first-quarter disclosure can narrow this uncertainty through four readings. The most direct is whether base rent is clearly above the roughly $46 million that corresponds to 100 MW; the next is whether the company gives, for the first time, a specific operating megawatt figure for the second building and ready-for-service dates for the remaining capacity; then whether the net operating income margin on base rent stays near 91% instead of slipping noticeably as new space comes into service; and finally whether tenant recovery revenue moves in the same direction as base rent. There are three observable falsifiers: base rent stays at the 100 MW level, which would mean the second building is energized but not paying rent; the company pushes back the ready-for-service date for the rest of the second building or for the third building; or the net operating income margin falls below 88%, which would mean non-recoverable operating expenses at the new building are higher than at the first.

Revenue more than doubled last year, but nearly half of it was pass-through tenant fit-out work — once that fades, how thick is what remains?

This debate matters because large swings in total revenue may say nothing about the health of the business. Total revenue jumped from $228.6 million to $611.3 million in fiscal 2026[3], and the largest piece of the increase was $270.6 million of fit-out services revenue[11], which is recognized at cost plus a markup[23] against $258.1 million of cost[12], leaving gross profit of only about $12.5 million. That revenue follows each building's fit-out period and stops when the work is done, so total revenue could well fall sharply in a given quarter without the business getting worse. The financial transmission is two lines layered together: fit-out volume rises and falls with the delivery schedule and cost-plus services revenue swings widely, but with a gross margin of only about 4.6%, gross profit is barely affected when total revenue falls; if rent with its 91% net operating income margin and hosting with its roughly 31% segment margin rise over the same period, gross profit and segment profit improve, and whether that narrows the operating loss depends on corporate-level selling, general and administrative expense.

The available evidence shows that the recurring revenue base is real, but it is still less than half of total revenue and is buried under corporate-level costs. Rental and other revenue of $114.7 million plus hosting revenue of $154.4 million came to $269.1 million in fiscal 2026, about 44.0% of total revenue[28]. On the profit side, HPC segment profit was $39.1 million and hosting segment profit was $48.3 million[40]; hosting profit looks lower than the prior year's $63.9 million[39], but the prior year included a $24.6 million gain from the sale of the Garden City site[45], so it improved on an underlying basis, and both sites were still at full capacity as of May 31, 2026[24]. What dragged down consolidated profit was $332.1 million of selling, general and administrative expense, including an increase of about $198.3 million in stock-based compensation[29], together with ChronoScale's $37.0 million operating loss and the $59.7 million held-for-sale impairment[25]; the two segments earned $87.5 million combined while the consolidated operating loss was $236.5 million[5]. Another reading also has to be kept in view: fit-out revenue is thin, but it means buildings are being fitted out, so a steep drop in fit-out revenue could also signal a gap before the next batch of deliveries, and it should be read alongside the delivery milestones in the first debate.

What matters in the first quarter is the numerator, not just the ratio. The readings worth tracking are the quarterly amount of fit-out services revenue and whether the company says which building's fit-out period it relates to, whether rent plus hosting rises clearly from about 44% of total revenue with the numerator itself growing, whether consolidated gross profit is at least the fiscal 2026 quarterly average of about $39.4 million, and the quarterly amount of selling, general and administrative expense, especially stock-based compensation. There are three observable falsifiers: fit-out revenue falls while rent and hosting do not rise, in which case the improvement in revenue quality is only a denominator effect; the fit-out gross margin turns negative, which would mean the company is absorbing tenant cost overruns; or hosting segment revenue or profit drops clearly below the fiscal 2026 quarterly average, which would mean the recurring base is getting thinner.

Applied Digital spent nearly $2.9 billion on construction last year, funded mainly by project notes and Macquarie's preferred equity — can that structure carry it to the rent ramp without further diluting common shareholders?

This debate matters because how much of the rent ultimately reaches common shareholders depends on the cost and seniority of the money used to build. The business model spends first and collects rent later, with a construction period of roughly 14 to 18 months in between[15]; cash purchases of property and equipment were $2,865.8 million in fiscal 2026[31], while minimum contracted lease payments for fiscal 2027 are only $451.1 million[13]. Three layers of capital fill the gap: senior secured notes issued by campus subsidiaries, including $2.35 billion at 9.25% and $2.15 billion at 6.75%[34], followed after year-end by $1.59 billion at 7.00%[20]; subsidiary preferred equity purchased by Macquarie, which has funded $1.8 billion and holds 13.5% of TopCo 2's fully diluted common equity[46]; and parent-level convertible notes and Series G convertible preferred stock[47]. The first two layers are paid ahead of common shareholders, and the third dilutes them directly. The financial transmission runs as follows: the 1,410 MW already signed requires continued construction spending, so debt and preferred balances rise; interest income falls as the proceeds are spent, interest stops being capitalized as buildings enter service, and net interest expense, distributions attributable to redeemable noncontrolling interests and common share dilution all rise, taking a share of the profit from rent before common shareholders do.

The available evidence shows that the cost of financing is falling, but dilution is still occurring and today's low interest expense will not last. As of May 31, 2026, the company had $1.6 billion of unrestricted cash and $2.6 billion of restricted cash[33], net long-term debt of $4,959.5 million[34] and construction in progress of $2,776.2 million[35]. Net interest expense was only $29.5 million in fiscal 2026, lower than the year before, but two temporary factors explain it: unspent proceeds generated $55.7 million of interest income, and $39.6 million of interest was capitalized during construction[48], and both will reverse once the cash is spent and the buildings enter service. The cost of financing itself is coming down, with coupons on the three note issues moving from 9.25% to 6.75% and 7.00%, and part of the CoreWeave lease has been moved to a wholly owned CoreWeave subsidiary with a springing guaranty from the CoreWeave parent and a $50 million letter of credit[49], which adds a layer of protection for the related project notes. On the other side, cumulative Series G issuance has reached $1.04 billion, most of it already converted into about 52.8 million common shares, under a commitment capped at $2 billion[47]. The annual report gives no capital expenditure figure for fiscal 2027 and says only that significant investment will continue throughout the year[50]; on construction progress, Blockspace relays that B. Riley analyst Nick Giles, after visiting Polaris Forge 2, reported that the shell of Building 1 was complete and that the company expects the first data hall to finish in the fourth quarter of 2026[51], which is a sell-side site observation and not a company disclosure.

The first quarter should show which layer the money came from and how each of the three components of interest moved. The readings worth tracking are cash purchases of property and equipment in the quarter and the construction costs left in payables at quarter-end, whether Polaris Forge 2 is still expected to reach initial capacity in the second half of calendar 2026[21], the separate changes in net interest expense, interest income and capitalized interest, and whether the company issued new Series G convertible preferred stock or common stock during the quarter. There are three observable falsifiers: new Series G or common stock issuance in the quarter, which would mean project-level financing does not cover the spending; net interest expense above $15 million in a single quarter without a matching rise in base rent; or a clear decline in unrestricted cash, which would mean the parent is covering shortfalls for the projects.

Risks and Falsifiers

High customer concentration is the first risk, and it exposes rent, the servicing of the project notes and later refinancing all at once. One HPC customer contributed 59% of fiscal 2026 revenue and one hosting customer contributed 25%[26]; looking ahead, the 1,410 MW of leases sits with just three tenants, CoreWeave and two investment-grade hyperscalers, and all 400 MW and roughly $11.0 billion of contracted revenue at Polaris Forge 1 comes from CoreWeave[1]. The concern would weaken if CoreWeave delivers the $50 million letter of credit as agreed and keeps paying rent on time[49], and if newly disclosed leases continue to come from investment-grade tenants.

Operations concentrated in North Dakota under leases that carry service-level commitments are the second risk. Every data center now operating is in North Dakota[27], and repeated failures to meet service levels can give tenants, beyond credits against rent, the right to terminate in certain cases[52]; all $269.1 million of fiscal 2026 rental and hosting revenue came from sites in that state[28], so a single regional power, weather or regulatory event would hit both segments together. The weight of this risk would fall if the company discloses no material outage or service-level credits and campuses in other states, such as Louisiana, begin operating in 2027 as planned.

Buildings delivered later than the leases require are the risk tied directly to the first core debate. Until the company completes the facilities its customer leases require, it will not realize the full projected revenue from those leases[53], and in certain cases of significant delay tenants have the right to terminate the affected leases[54]. The minimum contracted lease payments of $451.1 million for fiscal 2027 and $1,453.7 million for fiscal 2028 both assume on-time delivery[13]; a one-quarter delay on a 150 MW building means roughly $69 million less base rent at the Polaris Forge 1 contract rate. The delay concern would not hold if first-quarter base rent is above $46 million and the company keeps its schedule of completing Polaris Forge 1 by the first half of 2027 and reaching initial capacity at Polaris Forge 2 in the second half of 2026.

Expiry of the legacy hosting contract could remove the company's largest block of segment profit before HPC rent has filled the gap. The business serves a single crypto-mining customer under a contract with about one and a half years remaining[14]; hosting generated $154.4 million of revenue and $48.3 million of segment profit in fiscal 2026[40], about 25% of total revenue and about 55% of combined segment profit. The risk would be pushed back if quarterly hosting revenue stays at or above $35 million, both sites remain at full capacity, and the company discloses a renewal or a definite plan to convert the sites to HPC use.

Construction costs or schedules that overrun the budget would push the shortfall onto the parent and its common shareholders. If project notes and Macquarie's preferred equity fall short, the parent would need to issue Series G convertible preferred stock or common stock to cover the difference, and the company has provided completion guarantees on each of the project notes; construction cash spending was $2,865.8 million in fiscal 2026 with $556.4 million of unpaid construction costs at year-end[31], and cumulative Series G issuance of $1.04 billion has already converted into about 52.8 million common shares[47]. The concern would be falsified by two consecutive fiscal quarters with no new Series G or common stock issuance and unchanged delivery schedules across the campuses.

What to Watch Next

• HPC base rent: the reference is $99.8 million for fiscal 2026, covering about three quarters, or roughly $46 million per quarter for 100 MW[4]. A first-quarter figure clearly above $46 million confirms that the second building is paying rent; a figure stuck at the 100 MW level falsifies it.
• Operating critical IT load: 100 MW is operating, and the second, 150 MW building is partially operational with no megawatt figure disclosed[10]. Watch for a first specific megawatt number and ready-for-service dates; a later date for the second or third building is a falsifier.
• Net operating income margin on base rent: 91% in fiscal 2026[30]. Watch whether it holds near 91% as new space enters service; a drop below 88% is a falsifier.
• Rent plus hosting as a share of revenue: about 44.0%, on a numerator of $269.1 million[28]. The share should rise with the numerator growing; a higher share on a flat numerator is only a denominator effect.
• Fit-out services revenue and margin: $270.6 million at a gross margin of about 4.6%[11]. Watch the quarterly amount and which building it relates to; a negative margin is a falsifier.
• Selling, general and administrative expense: $332.1 million, including an increase of about $198.3 million in stock-based compensation[29]. If it does not fall, better segment profit will not reach the operating loss.
• Cash purchases of property and equipment: $2,865.8 million in fiscal 2026 plus $556.4 million of unpaid construction costs[31], and about $600 million for the coming quarter by management's figure[6]. Unrestricted cash clearly below $1.6 billion would mean the parent is covering project shortfalls.
• Net interest expense: $29.5 million, after $55.7 million of interest income and with $39.6 million of interest capitalized[48]. More than $15 million in a single quarter without a matching rise in base rent is a falsifier.
• Series G or common stock issuance: $1.04 billion issued to date, converted into about 52.8 million shares[47]. Two consecutive quarters with no new issuance and unchanged delivery schedules would be confirmation.

Conclusion

Applied Digital's business is driven by megawatts delivered and paying rent, and the company is still at the stage where signed capacity far exceeds rent-paying capacity. As of May 31, 2026, it had signed about 1,410 MW and roughly $36.2 billion of take-or-pay leases[1], but base rent recognized in fiscal 2026 was only $99.8 million[4]; that rent carries a 91% net operating income margin[30], yet the consolidated operating loss was $236.5 million[5], construction cash spending in the same year was $2,865.8 million[31], and net long-term debt reached $4,959.5 million[34]. The central unresolved relationship is therefore a race between the speed of the rent ramp and construction spending with its financing cost: the later rent scales, the more note interest, preferred distributions and Series G dilution accumulate ahead of common shareholders.

Two independent commentaries published after the annual report look at the same timing gap from opposite directions, and both are outside interpretations, not facts. Daniel Sparks of The Motley Fool uses the company's own schedule of minimum lease payments to point out that less than $2 billion of the roughly $36 billion due over 15 years, about 5%, arrives before fiscal 2029, while the construction bill is due now and lenders and Macquarie's preferred equity rank ahead of common stock; he acknowledges that management has delivered on schedule so far, but he argues that big builds can slip and prefers to wait until the company keeps delivering campuses on time[55]. B. Riley analyst Nick Giles, as relayed by Blockspace, reported after visiting Polaris Forge 2 that the shell of Building 1 was complete and that electrical and mechanical contractors were installing interior systems, and that the company plans to deliver capacity in blocks of about 75 MW; the firm estimated a yield on cost of about 11.5% for the campus using annual revenue of $1.61 million per MW, an 86% net operating income margin and development costs of $12 million per MW, relayed management's expectation that later agreements would reach 15% to 20%, and disclosed that Applied Digital is or recently was an investment-banking client[51]. The two agree in not treating tenant demand as the main risk, and they differ on how much the timing of cash recovery matters: the first matches the cautious side of the first and third debates, while the second offers site evidence that construction is on plan and unit economics are positive, although those yields and unit costs are sell-side estimates and cannot be used as company figures. B. Riley's timing for the first data hall, the fourth quarter of 2026, is also more specific and later than the annual report's language about reaching initial capacity in the second half of calendar 2026[21], so the first rent from Polaris Forge 2 is unlikely to appear in the quarter being reported.

The combination that would materially strengthen the current understanding is first-quarter base rent clearly above $46 million, a first quantified operating megawatt figure for the second building and a net operating income margin holding near 91%, together with a higher dollar amount of rent plus hosting, lower selling, general and administrative expense and no new Series G or common stock issuance in the quarter. The combination that would materially weaken it is base rent stuck at the 100 MW level, a later ready-for-service date for any building, net interest expense above $15 million in a quarter without rent keeping pace, or a clear decline in unrestricted cash accompanied by new convertible preferred issuance. A decline in fit-out services revenue alone is neither of these outcomes and has to be judged together with the delivery milestones and the dollar amount of recurring revenue.

Sources

[1] APLD 10-K filed 2026-07-29 · contracted lease portfolio · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[2] APLD earnings calendar · 2026-10-08 fiscal first quarter 2027 earnings call (calendar last updated 2026-09-18) · 2026-09-18 · Drillr earnings calendar

[3] APLD 10-K filed 2026-07-29 · consolidated results of operations · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[4] APLD 10-K filed 2026-07-29 · base rent and tenant recoveries · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[5] APLD 10-K filed 2026-07-29 · operating loss and net loss · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[6] APLD earnings call 2026-07-27 · guidance (Drillr call summary) · 2026-07-27 · Applied Digital Corporation · https://ir.applieddigital.com/news-events/ir-calendar

[7] APLD earnings call 2026-07-27 · NOI run-rate target (Drillr call summary) · 2026-07-27 · Applied Digital Corporation · https://ir.applieddigital.com/news-events/ir-calendar

[8] FXEmpire APLD earnings page (accessed 2026-09-18) · 2026-09-18 · FXEmpire(6 位分析师) · https://www.fxempire.com/stocks/apld/earnings

[9] stockanalysis.com APLD analyst forecast (accessed 2026-09-18) · 2026-09-18 · stockanalysis.com(5 位分析师,数据来自 S&P Global Market Intelligence) · https://stockanalysis.com/stocks/apld/forecast/

[10] APLD 10-K filed 2026-07-29 · Polaris Forge 1 operating status · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[11] APLD 10-K filed 2026-07-29 · services revenue commentary · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[12] APLD 10-K filed 2026-07-29 · tenant fit-out cost of revenue · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[13] APLD 10-K filed 2026-07-29 · minimum contracted lease payments · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[14] APLD 10-K filed 2026-07-29 · legacy data center hosting business · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[15] APLD 10-K filed 2026-07-29 · competitive strengths · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[16] APLD 10-K filed 2026-07-29 · employees · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[17] APLD 10-K filed 2026-07-29 · ChronoScale separation · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[18] APLD 10-K filed 2026-07-29 · ChronoScale consolidation · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[19] APLD 10-K filed 2026-07-29 · Polaris Forge 3 and Delta Forge campuses · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[20] APLD 10-K filed 2026-07-29 · subsequent lease and notes offering · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[21] APLD 10-K filed 2026-07-29 · Polaris Forge 2 delivery timing · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[22] APLD 10-K filed 2026-07-29 · HPC lease revenue recognition · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[23] APLD 10-K filed 2026-07-29 · tenant fit-out revenue recognition · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[24] APLD 10-K filed 2026-07-29 · hosting sites at full capacity · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[25] APLD 10-K filed 2026-07-29 · adjusted EBITDA reconciliation · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[26] APLD 10-K filed 2026-07-29 · customer concentration · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[27] APLD 10-K filed 2026-07-29 · North Dakota concentration · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[28] APLD 10-K filed 2026-07-29 · HPC and hosting segment revenue · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[29] APLD 10-K filed 2026-07-29 · SG&A and stock-based compensation · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[30] APLD 10-K filed 2026-07-29 · net operating income reconciliation · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[31] APLD 10-K filed 2026-07-29 · cash flow statement · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[32] APLD 10-K filed 2026-07-29 · financing cash flow and ending cash · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[33] APLD 10-K filed 2026-07-29 · liquidity and capital resources · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[34] APLD 10-K filed 2026-07-29 · long-term debt table · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[35] APLD 10-K filed 2026-07-29 · property and equipment · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[36] APLD 10-K filed 2026-07-29 · tenant recoveries policy · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[37] APLD 10-K filed 2026-07-29 · ChronoScale revenue decline · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[38] APLD 10-K filed 2026-07-29 · rental cost of revenue components · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[39] APLD 10-K filed 2026-07-29 · segment profit three-year comparison · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[40] APLD 10-K filed 2026-07-29 · FY2026 segment revenue and profit · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[41] APLD 10-K filed 2026-07-29 · cash flow commentary · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[42] APLD 10-K filed 2026-07-29 · industry trends · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[43] APLD 10-K filed 2026-07-29 · competition · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[44] APLD 10-K filed 2026-07-29 · take-or-pay lease structure · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[45] APLD 10-K filed 2026-07-29 · FY2025 segment detail · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[46] APLD 10-K filed 2026-07-29 · Macquarie preferred equity funding · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[47] APLD 10-K filed 2026-07-29 · Series G preferred issuance and dilution · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[48] APLD 10-K filed 2026-07-29 · interest expense and capitalized interest · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[49] APLD 10-K filed 2026-07-29 · CoreWeave lease credit enhancement · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[50] APLD 10-K filed 2026-07-29 · funding requirements · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[51] Blockspace via Yahoo Finance 2026-08-21 · B. Riley Polaris Forge 2 site visit · 2026-08-21 · Blockspace(经 Yahoo Finance 转载) · https://finance.yahoo.com/markets/stocks/articles/b-riley-keeps-75-applied-152118518.html

[52] APLD 10-K filed 2026-07-29 · service level commitments · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[53] APLD 10-K filed 2026-07-29 · construction completion risk · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[54] APLD 10-K filed 2026-07-29 · lease commencement delay risk · 2026-07-29 · 10-K · https://www.sec.gov/Archives/edgar/data/1144879/000114487926000048/

[55] The Motley Fool 2026-09-07 · Applied Digital rent timing · 2026-09-07 · The Motley Fool · https://www.fool.com/investing/2026/09/07/applied-digital-won-t-collect-most-of-its-rent-until-2029-i-d-wait-to-buy-the-stock/