Applied Digital ($APLD) Stock Hub 2026: 1.41 GW of AI Data Centers, $36 Billion of Lease Value—and the Funding Test
Verified through August 28, 2026: Applied Digital has long-term leases covering 1,410 MW of critical IT load across five campuses, but only 175 MW was live on June 30. The investment case is no longer about whether AI demand exists. It is whether APLD can build and commission the remaining 1,235 MW on time, turn fit-out-heavy reported growth into recurring base rent, and leave enough value for common shareholders after debt, project-level preferred capital and dilution.
Get every Merlintrader report in real time on Telegram: join @merlintraderpub_com.
At a glance
The contract book is already large enough. The next evidence is physical and financial: ready-for-service milestones, base-rent growth, construction cost control, and cash generation after interest and project-level claims.
It is not current backlog under a single accounting definition, it is not cash on the balance sheet, and it is not profit. Significant construction delays may create tenant termination rights, so the gap between contracted and commissioned capacity is the core risk.
01 The central question: can APLD convert contract value into common-equity value?
Applied Digital has already crossed the commercial threshold that stops many data-center developers: it has signed long-term leases for roughly 1.41 GW of critical IT load. The portfolio spans five named campuses and carries approximately $36.2 billion of nominal revenue over the initial lease terms. That is real contractual evidence, not a non-binding pipeline slide.
But the equity thesis starts where the headline ends. At June 30, only 175 MW was operational. Roughly 1,235 MW remained to be delivered, a derived figure equal to 1,410 MW less 175 MW. Building that capacity requires billions of dollars before most of the lease payments arrive. The company has funded the build with secured notes, convertible preferred stock, common equity, warrants, project-level preferred capital and noncontrolling interests. The company can therefore grow rapidly while value per common share grows more slowly—or not at all.
Merlintrader classification: APLD belongs in AI infrastructure and hyperscale data centers. It is not an AI software vendor, a semiconductor manufacturer or a conventional REIT. Its outcome depends on construction, power, tenant credit, financing and property-level returns.
Underwriting status: watchlist / preliminary. Evidence confidence is high for filed contracts, reported financials and capital instruments. It is lower for future delivery dates, the remaining unnamed tenant’s credit quality and eventual project returns. A precise price target would imply a level of fully diluted, project-by-project modelling that the public disclosures do not yet support.
02 What Applied Digital actually does
Applied Digital designs, develops, owns and operates large-scale data centers engineered for high-performance computing workloads, including artificial intelligence, machine learning and other accelerated computing. The company supplies the powered shell, cooling, electrical and physical infrastructure into which tenants install or operate dense compute equipment.
HPC Hosting Business
The strategic growth business. It develops and leases purpose-built AI/HPC campuses under long-duration agreements. FY2026 revenue was $385.3 million, but $270.6 million came from tenant fit-out services and $114.7 million from rent and tenant recoveries.
Data Center Hosting Business
The legacy crypto-mining hosting segment. It operates 106 MW in Jamestown and 180 MW in Ellendale for one customer. FY2026 revenue was $154.4 million and segment profit was $48.3 million.
ChronoScale complicates the picture. Applied Digital separated its historic cloud business into a separately traded company in May 2026, but still owned approximately 96% at the July results update and continued to consolidate it. Management excludes ChronoScale from its adjusted revenue and adjusted EBITDA, while GAAP results still include it.
The long-run corporate direction is clear: management wants the HPC and legacy hosting segments to represent the core platform and describes a transition toward a pure-play data-center model. The accounting has not fully reached that destination.
03 The five-campus contracted portfolio
| Campus | Tenant | Critical IT load | Base lease value | Expected delivery | What is verified |
|---|---|---|---|---|---|
| Polaris Forge 1 Ellendale, North Dakota | CoreWeave | 400 MW | ~$11.0B | 2H25–1H27 | 175 MW live at June 30; 225 MW remained to be delivered |
| Polaris Forge 2 near Harwood, North Dakota | Oracle | 200 MW | ~$5.0B | 2H26–1H27 | Under construction; two five-year renewal options |
| Polaris Forge 3 North Dakota | Unnamed high-investment-grade hyperscaler | 300 MW | ~$7.5B | 2H27–2H28 | Two buildings; initial operations guided for 2027 |
| Delta Forge 1 Boyce, Louisiana | Same unnamed hyperscaler | 300 MW | ~$7.5B | 1H27–1H28 | Two leases; construction and utility delivery remain critical |
| Delta Forge 2 unnamed southern state | Same unnamed hyperscaler | 210 MW | ~$5.2B | 1H28 | One building; lease signed June 5, 2026 as a subsequent event |
| Total | CoreWeave, Oracle and one unnamed hyperscaler | 1,410 MW | ~$36.2B | 2H25–2H28 | Initial terms approximately 15 years; renewal value excluded |
The table uses the portfolio presentation in the FY2026 Form 10-K and cross-checks each project against the related lease and financing filings. The delivery windows are management estimates, not completed milestones. The 10-K keeps Polaris Forge 2 anonymous, but the March 10 secured-note Form 8-K identifies its tenant as Oracle. “High investment-grade” remains the issuer’s description for the one tenant whose identity is still undisclosed; readers should not substitute Microsoft, Amazon, Google or Meta.
Do not count five customers. Polaris Forge 1 is leased to CoreWeave; Polaris Forge 2 is leased to Oracle; Delta Forge 1, Polaris Forge 3 and Delta Forge 2 are leased to the same unnamed hyperscaler. The HPC portfolio therefore has three tenants.
04 175 MW live versus 1,410 MW contracted
The cleanest way to understand APLD is to separate commercial proof from operating proof. Executed leases demonstrate demand. Energized buildings demonstrate execution. Cash rent demonstrates economic conversion.
The first 100 MW data center at Polaris Forge 1 became operational in October 2025. Phase 1 of the second building added 75 MW on June 30, 2026, taking the campus to 175 MW live. The remaining capacity ranges from buildings already under construction to campuses with delivery windows extending into the second half of 2028.
This gap is why the stock cannot be evaluated from lease value alone. Construction delays, unavailable labor or equipment, permitting, utility interconnection, geological problems and site readiness can all move ready-for-service dates. The 10-K explicitly states that, under certain circumstances, lessees may terminate applicable leases if construction is significantly delayed.
The milestone that matters is not another press release about pipeline gigawatts. It is an accepted ready-for-service certificate, followed by lease commencement, base rent and cash collection.
05 What the $36 billion lease figure means—and what it does not
The Form 10-K presents approximately $36.2 billion of total contracted revenue over the initial terms. The lease note gives a more precise schedule of $35.845 billion of minimum contracted payments. The difference is largely the result of rounding and presentation. Neither number should be described as revenue already earned.
| Fiscal period | Minimum contracted payments | Share of total | Interpretation |
|---|---|---|---|
| FY2027 | $451.1M | 1.3% | Early ramp; still small relative to the total headline |
| FY2028 | $1.454B | 4.1% | More buildings must enter service |
| FY2029 | $2.252B | 6.3% | Portfolio moves closer to broad operation if schedules hold |
| FY2030 | $2.281B | 6.4% | Contractual payments approach a mature annual level |
| FY2031 | $2.300B | 6.4% | Coincides with a major debt maturity year |
| Thereafter | $27.108B | 75.6% | Most nominal value sits beyond the explicit five-year window |
| Total | $35.845B | 100% | Nominal minimum payments, not discounted value or profit |
The leases are described as take-or-pay and non-cancellable for convenience, with a convenience termination requiring payment of the remaining contractual value. That protection is meaningful. It is not absolute. The risk factors and CoreWeave lease filings identify circumstances in which failure to meet specified installation or delivery requirements can permit termination without a convenience termination fee.
The company also cites approximately $86 billion if every renewal option is exercised. That is optional future value dependent on tenant decisions many years from now. It belongs in an upside discussion, not in the base contracted-revenue figure.
06 Revenue quality: Q4 growth was mostly tenant fit-out
Fiscal fourth-quarter revenue of $258.7 million and adjusted EBITDA of $42.4 million look like a sharp operating inflection. The composition matters more than the headline.
| Q4 FY2026 HPC revenue component | Amount | Share of HPC revenue | Economic character |
|---|---|---|---|
| Tenant fit-out services | $152.4M | 75.1% | Construction-related services recognized as costs are incurred plus contractual markup; not recurring base rent |
| Base rent | $44.1M | 21.7% | The recurring lease line investors ultimately need to see scale |
| Tenant recoveries | $6.5M | 3.2% | Reimbursement of applicable expenses; revenue with corresponding costs |
| Total HPC revenue | $203.0M | 100% | Produced $26.2M of segment operating profit |
For the full fiscal year, HPC Hosting generated $385.3 million of revenue: $270.6 million of tenant fit-out services and $114.7 million of rental and other revenue. The rental line consisted of $99.8 million of base rent and $14.9 million of tenant recoveries. The cost of tenant fit-out services was approximately $258.1 million.
This does not make fit-out revenue illegitimate. It does mean total revenue can rise faster than the underlying rent-producing asset base. The clean recurring indicator is base rent per quarter, supported by live MW and reconciled with segment profit, interest and capital spending.
07 Fiscal 2026 financial baseline
| Metric | FY2026 | FY2025 | How to read it |
|---|---|---|---|
| GAAP revenue | $611.3M | $228.6M | Consolidated and includes ChronoScale |
| Adjusted revenue | $539.7M | $144.2M | Management non-GAAP; excludes ChronoScale |
| Operating loss | $(236.5)M | $(72.2)M | GAAP consolidated |
| Net loss attributable to common | $(250.3)M | $(233.7)M | GAAP; $(0.91) per diluted share in FY2026 |
| Adjusted EBITDA | $107.2M | $19.6M | Management non-GAAP; not free cash flow |
| HPC Hosting revenue | $385.3M | $0 | First year of operations; fit-out was the majority |
| HPC segment profit | $39.1M | $(12.1)M | Segment measure, not consolidated operating income |
| Legacy hosting revenue | $154.4M | $144.2M | One crypto customer; stable but concentrated |
| Legacy hosting segment profit | $48.3M | $63.9M | Profit declined despite higher revenue |
| Stock-based compensation | $219.3M | $22.5M | Listed in the non-GAAP reconciliation; a real per-share cost |
Reported earnings contain unusually large moving pieces. Fiscal 2026 included a $59.7 million loss on classification as held for sale, $75.8 million of gains on derivatives and $10.8 million of gains on investments. Interest expense was partly offset by interest income on large construction cash balances. A clean reading therefore requires the GAAP statement, the segment table, the cash-flow statement and the non-GAAP reconciliation together.
Do not annualize Q4 adjusted EBITDA without adjustment. The period contained a large fit-out contribution, while the debt stack and construction assets were still ramping. A stable data-center model should ultimately be judged on recurring rent, property-level NOI, cash interest and maintenance plus development capital.
08 Cash flow and capex: the operating story is financed before it is harvested
| Cash-flow line | FY2026 | FY2025 | Interpretation |
|---|---|---|---|
| Operating cash flow | +$89.7M | $(115.4)M | Improved, but aided by non-cash compensation, interest, deferred revenue and working-capital changes |
| Purchases of property, equipment and other assets | $(2.866)B | $(681.6)M | Cash spending; excludes the full accrual-based segment capex figure |
| Total investing cash flow | $(2.936)B | $(667.7)M | Also includes related-party loans and investments |
| Financing cash flow | +$6.877B | +$874.7M | The build was funded primarily in capital markets |
| Segment capital expenditures | $3.042B | $986.8M | Of which $2.989B was assigned to HPC Hosting |
At May 31 the balance sheet showed $1.592 billion of unrestricted cash and approximately $2.561 billion of restricted cash, including long-term restricted amounts. The restricted balance is largely tied to construction, debt service or letters of credit. Subtracting all $4.153 billion from debt to claim a simple “net debt” figure would overstate liquidity available to common shareholders.
Applied Digital states that it expects to need substantial additional capital for growth. That statement is consistent with the scale of the build rather than evidence of imminent distress, but it creates a permanent underwriting question: what instrument funds each next dollar, and which claim receives the project economics first?
09 Debt and project-level capital: not one simple balance-sheet number
| Instrument | Principal / commitment | Rate | Maturity | Key point |
|---|---|---|---|---|
| Senior secured notes | $2.350B | 9.25% | December 2030 | Secured project financing tied to Polaris Forge 1 structure |
| Senior secured notes | $2.150B | 6.75% | March 2031 | Proceeds released after the post-year-end escrow condition was met |
| Convertible senior unsecured notes | $450M | 2.75% | June 2030 | Potential dilution if settled in shares under applicable terms |
| Bridge facility at May 31 | $300M | SOFR-based | April 2027 | Repaid after year-end from the next secured-note financing |
| Post-year-end senior secured notes | $1.590B | 7.00% | June 2031 | Closed June 16; refinanced the bridge and funds Building 4 |
| Secured revolver | $430M committed | SOFR + 2.25% or base + 1.25% | May 2029 | Includes letter-of-credit capacity; an additional $120M accordion remained |
| Texas Capital loan | $58.5M stated principal | Term SOFR + spread | June 2031 | Entered into June 30, after the May 31 balance-sheet date |
The May 31 contractual debt schedule showed $5.307 billion of principal and an estimated $1.785 billion of interest payments. It also concentrated $4.195 billion of principal in fiscal 2031. The June $1.59 billion issuance changed the structure after the balance-sheet date, so adding it mechanically to the May debt figure would be wrong: $300 million refinanced the bridge and the remaining proceeds are earmarked for construction.
A derived coupon check illustrates the scale. The four large note series—$2.35 billion at 9.25%, $2.15 billion at 6.75%, $1.59 billion at 7.0% and $450 million at 2.75%—carry roughly $486 million of annual stated coupon before smaller facilities, fees and amortization. That is not a forecast of consolidated net interest expense: project cash earns interest, some construction interest is capitalized, and legal borrowers differ. It is a reminder that recurring property cash flow must become very large.
The Macquarie layer
At May 31, an affiliate of Macquarie Asset Management had funded $1.825 billion of preferred units in APLD HPC TopCo 2 and held 13.5% of its fully diluted common equity; APLD held 86.5%. The balance sheet reported $1.956 billion of redeemable noncontrolling interest, including accrued PIK economics. The preferred return begins at 12.75% a year and the contractual liquidation framework includes a minimum 1.80x multiple of invested capital, with step-ups and redemption or sale-process rights under specified future conditions. On $1.825 billion funded, 12.75% is roughly $232.7 million of annualized preferred return by arithmetic. It is not GAAP interest expense, but it is an economically senior project-level claim before the residual common value.
Capital-structure rule: enterprise growth is not the same as growth attributable to the listed common share. Debt, project preferred units, noncontrolling interests, convertibles and warrants must all be considered before assigning value to APLD common.
10 Dilution: the share count has already moved materially
Applied Digital had 287.884 million common shares outstanding at May 31, 2026, up from 224.910 million one year earlier—an increase of about 28%. The Form 10-K cover reported 291.469 million shares on July 28. The company also increased authorized common shares to 600 million in November 2025.
| Potential source of dilution | Filed amount | Why it matters |
|---|---|---|
| Series G conversions to date | Approximately 52.8M common shares | Generated roughly $1.04B of gross preferred financing through the filing date |
| Series G remaining after June conversions | 73,154 preferred shares | Conversion outcome depends on contractual pricing mechanics |
| Warrants originally issued with CoreWeave leases | 21.456M shares before the disclosed exercise | 13.063M at $7.19 and 8.394M at $10.75, before adjustments; later assignments changed the holders |
| Total warrants outstanding at May 31 | 28.091M shares | Includes CoreWeave and other counterparties |
| Unvested restricted and performance units | 22.352M shares | Potential future dilution and compensation expense |
| $450M convertible notes | Conversion-dependent | Exact common-share outcome depends on note terms and settlement |
The PEPA commitment was increased to $2.0 billion, giving the company additional funding flexibility through Series G preferred issuance. Flexibility has value for a capital-intensive developer; it also means investors should track the fully diluted share count, not only basic EPS or the headline market capitalization.
Post-10-K ownership update. Four Form 4 filings made on August 4 disclosed the July 31 vesting of 3.045 million performance stock units across the CEO, president, COO and CFO. A combined 1.223 million shares were withheld for taxes, leaving approximately 1.822 million net shares delivered to those officers—a derived figure, not a newly reported company-wide share count. The filings describe the withheld shares as tax settlement rather than open-market sales. Consequently, the 291.469 million shares outstanding reported for July 28 remain the latest issuer-wide baseline used here, but should not be interpreted as a live fully diluted count.
At the August 28 market snapshot, the warrants originally issued with the CoreWeave leases were above their stated exercise prices, but a single “fully diluted” total would still require assumptions about exercise method, the other warrants, capped-call protection on the convertible notes, award vesting and Series G conversion pricing. The filed 291.5 million basic shares should therefore be read alongside the 28.1 million warrants, 22.352 million unvested restricted and performance units, convertible notes and remaining preferred stock—not replaced by a falsely precise sum.
This is an exposure map, not an instruction to add every instrument one-for-one. Treasury-stock treatment can reduce warrant dilution, performance awards may not vest, capped calls can offset some convertible dilution and preferred conversion pricing can vary. Ignoring the instruments would be misleading; treating all of them as immediately outstanding would be misleading too.
11 Customer concentration and CoreWeave-specific exposure
During FY2026, one HPC customer accounted for 59% of continuing-operations revenue. The legacy hosting customer accounted for 25%. The first figure largely reflects the early Polaris Forge 1 ramp, before Oracle and the remaining hyperscaler begin contributing, but it shows how concentrated the current income statement remains.
CoreWeave is one of the two named HPC tenants. Its three Polaris Forge 1 leases cover 400 MW and approximately $11 billion of initial-term value. Warrants originally issued in connection with those leases covered up to 21.456 million APLD shares before the disclosed 300-share cashless exercise, but subsequent assignments moved the warrant interests to financial counterparties. During March 2026, portions of the lease structure moved to a CoreWeave subsidiary, supported by springing guarantees and, for one arrangement, a $50 million letter-of-credit requirement. A memorandum concerning the Building 4 lease contemplated assignment to a subsidiary if it obtains an investment-grade rating; a conditional memorandum is not a completed credit upgrade.
Oracle is identified as the Polaris Forge 2 tenant in the March 10 financing filing. The third HPC tenant, covering Delta Forge 1, Polaris Forge 3 and Delta Forge 2, remains unnamed. Applied Digital says approximately 70% of contracted revenue is backed by U.S. investment-grade hyperscalers. That credit description is useful but remains issuer-reported, and investors cannot independently analyze the undisclosed tenant’s financial statements, lease obligations or strategic commitment.
Take-or-pay reduces utilization risk; it does not eliminate counterparty risk. In bankruptcy a tenant may seek to reject contracts and legal recovery can be limited. In a construction default, contractual termination rights may apply before the economic stream begins.
12 ChronoScale: legally separate, still consolidated
On May 5, 2026, Applied Digital completed a transaction that combined its cloud business with Ekso Bionics, renamed ChronoScale and listed the resulting company on Nasdaq under ticker CHRN. Applied Digital owned approximately 97% immediately after closing and approximately 96% in the July results release.
ChronoScale is therefore an independent publicly traded company in legal and market terms, but APLD remains its controlling shareholder and consolidates it under GAAP. Management excludes ChronoScale from the adjusted revenue and adjusted EBITDA used to describe core data-center performance.
What the separation improves
It gives the cloud assets a separate security, isolates operating narratives and allows management to present core data-center non-GAAP metrics without the cloud business.
What remains
APLD GAAP statements, controls, earnings volatility and valuation still reflect a roughly 96% consolidated holding. Calling APLD a completed pure-play data-center company would be premature.
The appropriate monitoring questions are whether APLD reduces its stake, whether ChronoScale can fund itself without APLD support, and how the holding is treated in any sum-of-the-parts valuation.
13 The legacy crypto-hosting business still matters
The Data Center Hosting Business operates 106 MW at Jamestown and 180 MW at Ellendale for one crypto-mining customer. Both facilities were at full capacity on May 31, and the remaining contractual term was approximately one and a half years.
FY2026 revenue rose to $154.4 million from $144.2 million, while segment profit fell to $48.3 million from $63.9 million. The segment provided operating income while the HPC platform was being built, but its customer concentration and approaching contract horizon matter. If the customer departs or economics weaken, APLD must renew, replace, repurpose or accept a declining contribution.
This exposure also makes the sector label important. APLD is increasingly an AI infrastructure company, but current consolidated economics are not entirely insulated from digital-asset hosting.
14 The possible moat: power, sites, design and execution—not the word “AI”
The scarce inputs in large-scale AI infrastructure are not only GPUs. They include secured land, utility interconnection, substations, transmission, cooling, fiber, construction capacity and a design that can support high rack density. Applied Digital’s proposition is that its North Dakota and southern U.S. sites can combine these inputs at hyperscale.
The company promotes waterless cooling and a standardized 150 MW building design. Those features may shorten replication and lower water constraints, but they should be evaluated through delivered buildings, tenant acceptance, power usage effectiveness, uptime and construction cost per MW. Marketing descriptions are not substitutes for operational data.
Power figures also require discipline. Campus announcements sometimes cite gross utility power greater than net critical IT load. The contracted table uses critical IT MW; it should not be combined with gross utility pipeline figures. Similarly, the company’s 3 GW active and more than 5 GW extended pipeline are development optionality, not executed leases.
Durable advantage test: if APLD repeatedly delivers standardized buildings on schedule and tenants accept them without material remediation, the power-and-execution platform begins to look like a moat. If costs and dates slip, the same fixed infrastructure becomes a source of leverage risk.
15 Peer map: useful references, not clean valuation comparables
| Reference group | Examples | What is comparable | Why the comparison breaks |
|---|---|---|---|
| AI compute and neocloud | CoreWeave, Nebius | AI demand, customer concentration, accelerated-compute ecosystem | They sell compute services; APLD primarily leases physical data-center infrastructure |
| Power-to-AI transition developers | IREN, TeraWulf | Large power portfolios, former or continuing mining exposure, construction funding | Ownership models, GPU exposure, tenant structures and balance sheets differ |
| Established data-center landlords | Digital Realty, Equinix | Lease economics, utilization, recurring rent and cost of capital | Mature diversified portfolios, investment-grade funding and REIT structures are not comparable to APLD’s development stage |
| Large hyperscalers | Microsoft, Amazon, Alphabet, Meta | Demand source and data-center capex cycle | They are potential customers or ecosystem drivers, not like-for-like landlords |
This is why a single peer revenue multiple is weak. APLD combines a developer, a landlord, a legacy crypto host, a controlling stake in a cloud company and multiple project-level financing structures. The right endpoint is stabilized property cash flow attributable to common equity, but the portfolio is still too early for that figure to be observed across most campuses.
16 Management, governance and execution
Wes Cummins is co-founder, chairman and chief executive; Jason Zhang became president and co-founder in January 2026; Saidal Mohmand is chief financial officer; Laura Laltrello is chief operating officer. The team has executed an unusually large sequence of leases and financings in a short period. That is evidence of transaction capability. The next phase demands construction discipline, controls and capital allocation rather than more signed headlines.
An August 6 Form 4 also reported that director Richard Nottenburg sold 75,000 shares on August 4 at a $31.15 weighted-average price. This was an actual open-market sale, unlike the tax-withholding transactions reported by the four officers, and is included as a governance datapoint rather than evidence about near-term operating performance.
Governance deserves attention because capital-market execution is central to the model. The CEO founded 272 Capital, where the CFO also serves as director of research. Several officers and directors have individual interests in Base Electron, an independent power developer. APLD reported a $58.6 million related-party loan receivable from the Base Electron structure at May 31 and said it was not the primary beneficiary and was not contractually required to provide support.
The company remediated its previously reported material weakness relating to complex financial instruments and concluded that internal control over financial reporting was effective at May 31, 2026; the auditor also expressed an effective-control opinion. Newly acquired Ekso/ChronoScale controls were excluded from management’s assessment under permitted guidance. The correct status is therefore remediated, with continuing complexity—not an unresolved material weakness.
17 The result must be measured on explicit time horizons
The primary research horizon for APLD is 18 to 24 months. Shorter windows can validate construction and financing; they cannot prove mature economics. Each horizon needs an observable scorecard.
3–6 months: delivery discipline
Pass: announced ready-for-service milestones occur within the disclosed windows; no material termination notice; construction cost and financing disclosures remain controlled. Fail: meaningful delays, new expensive rescue capital or a reduction in contracted capacity.
12 months: rent replaces fit-out
Pass: live MW and quarterly base rent rise materially, fit-out becomes a smaller share of HPC revenue, and property-level profit covers more of the cash interest burden. Fail: revenue remains dominated by fit-out while delivery dates move right.
18–24 months: common-equity economics
Pass: multiple campuses are in service, recurring NOI and cash flow scale, and funding per new MW no longer requires outsized dilution. Fail: debt, preferred claims and new shares absorb most of the asset growth.
| Calendar window | Company delivery framework | Evidence to require |
|---|---|---|
| 2H 2026–1H 2027 | Polaris Forge 2 initial through full capacity; remaining Polaris Forge 1 phases through 1H27 | RFS acceptance, lease commencement, base rent and no material cost reset |
| 1H 2027–1H 2028 | Delta Forge 1 delivery phases | Utility interconnection, building delivery and tenant acceptance |
| 2H 2027–2H 2028 | Polaris Forge 3 phased delivery | Initial operation followed by both buildings reaching contracted load |
| 1H 2028 | Delta Forge 2 delivery | Single-building execution and first rent from the fifth campus |
No exact next earnings date had been announced in the primary-source set used for this Hub. The next quarterly release should be treated as an update point when the company schedules it, not guessed from a third-party calendar.
18 Valuation: why false precision is especially dangerous here
At the August 28 intraday snapshot, APLD’s market-data-provider capitalization was approximately $7.14 billion. The filed July 28 share count multiplied by the same $25.31 price implies roughly $7.38 billion, illustrating that provider share-count conventions and recent conversions can create differences. Neither figure includes the full economic effect of debt, project preferred capital, noncontrolling interests, convertibles or warrants.
A superficial comparison divides the equity value by $611.3 million of FY2026 GAAP revenue and obtains a high revenue multiple. That denominator is not clean: it includes ChronoScale and substantial fit-out revenue. Dividing by $36.2 billion is worse, because the numerator is current equity value while the denominator is fifteen years of nominal lease payments before construction cost, operating cost, interest, partner preferences and taxes.
| Valuation method | Use today | Main limitation |
|---|---|---|
| Current revenue multiple | Context only | Revenue mix is transitional and fit-out-heavy |
| Contracted lease value multiple | Avoid as a headline | Undiscounted fifteen-year payments are not enterprise value or backlog cash |
| Stabilized NOI / property value | Best eventual framework | Most campuses are not operating, so NOI and final project cost are unobserved |
| Sum of the parts | Useful but preliminary | Requires allocation of debt, preferred claims, minority interests and ChronoScale value |
| Discounted cash flow | Scenario tool only | Small changes in delivery, cost of capital and terminal value dominate the answer |
The underwriting gate is therefore operational: value each building only after a credible cost, financing package, lease commencement and attributable cash-flow path exist. Until then, the disciplined conclusion is a scenario range, not a single target price.
19 The constructive and skeptical cases
The constructive case
APLD has secured power-rich sites and signed long leases before most competing developers can assemble the same inputs. CoreWeave, Oracle and an issuer-described high-investment-grade hyperscaler validate demand. The first 175 MW is live, showing the model can move from lease to operation. Minimum payments rise from $451 million in FY2027 to more than $2.25 billion annually by FY2029–FY2031. If standardized buildings come online on schedule, recurring base rent and property-level NOI can scale much faster than corporate overhead. Project financing also matches long-lived assets with long-lived contracts rather than funding everything through common equity.
The skeptical case
Only 12.4% of contracted load was live, and most Q4 HPC revenue was tenant fit-out rather than rent. The build requires enormous capital, while the debt stack, Macquarie preferred units, noncontrolling interests, convertibles, warrants and equity awards sit ahead of or dilute common economics. One tenant produced 59% of FY2026 continuing revenue. Significant delays can create termination rights. The share count rose sharply in one year. APLD may build a valuable platform while the common share receives a lower return than the campus headlines imply.
20 Scenario framework for the next 18–24 months
These are analytical paths, not forecasts, targets or recommendations.
| Scenario | What has to happen | How it becomes visible | Common-equity implication |
|---|---|---|---|
| Execution compounds | RFS dates broadly hold; base rent scales; capex stays within financed budgets; tenant credit remains intact | Live MW rises in step with the filed schedule, rent becomes the dominant HPC line and cash interest coverage improves | The market can begin valuing stabilized property cash flow rather than optional capacity |
| Assets grow, per-share value lags | Most leases survive but costs and dates slip; more project capital and equity are required | Contracted MW remains high while basic and diluted shares, preferred claims or debt rise faster than attributable cash flow | Enterprise value can grow without a matching common-share return |
| Execution breaks | Material delays trigger remedies or termination rights; power or construction costs reset; refinancing conditions tighten | Capacity reductions, delayed rent, impairment, covenant pressure or financing on distressed terms | Fixed claims absorb value and common equity becomes the shock absorber |
21 Red flags and thesis-breakers
- A missed ready-for-service date that gives a tenant a contractual termination right.
- A reduction or cancellation of contracted critical IT load, especially at Polaris Forge 1 or across the three-campus second-hyperscaler relationship.
- Quarterly revenue growth without comparable base-rent growth, indicating continued dependence on fit-out accounting.
- Construction cost per MW rising faster than lease economics or material unbudgeted retrofit requirements.
- New capital that is materially more expensive or more senior than the existing structure, or large common issuance without a clearly accretive project return.
- Restricted cash presented as though it were freely available, or a liquidity analysis that ignores project covenants.
- Further concentration in one tenant or deterioration in a counterparty’s credit support.
- ChronoScale requiring support from APLD or creating control and reporting complications without a clear path to value realization.
- Legacy crypto-hosting economics weakening before replacement capacity is secured.
- Fully diluted share growth outpacing recurring cash flow per share.
Immediate thesis failure: a material lease termination caused by delivery default, combined with an inability to refinance or redeploy the affected asset on economic terms.
22 Evergreen monitoring checklist
| Every quarter | Why it matters | Preferred evidence |
|---|---|---|
| Live critical IT MW by building and campus | Turns development claims into operating capacity | 10-Q/10-K, 8-K and tenant-accepted RFS disclosure |
| Base rent, fit-out and tenant recoveries separately | Measures recurring revenue quality | Segment note and earnings release reconciliation |
| HPC segment profit and property NOI | Shows operating economics before corporate items | Filed segment table; company non-GAAP definition checked each period |
| Cash purchases of PPE and total segment capex | Tracks cash versus accrual construction intensity | Cash-flow statement and segment footnote |
| Unrestricted and restricted cash separately | Prevents false net-debt conclusions | Balance sheet and restricted-cash note |
| Debt principal, coupon, maturity and project borrower | Maps the claims on future rent | Debt footnote and new 8-K indentures |
| Basic shares, warrants, preferred conversions and awards | Measures per-share participation | Filing cover, EPS note, equity and warrant notes |
| Tenant concentration and credit enhancements | Captures counterparty risk | Risk factors, customer footnote and lease amendments |
| ChronoScale ownership and funding | Determines whether APLD is becoming a true pure play | APLD and CHRN filings |
23 Merlintrader bottom line
Applied Digital is one of the clearest public-market ways to express a high-risk view on the physical build-out behind artificial intelligence. The company has done the hard commercial work: five campuses, 1.41 GW under long-term lease and a payment schedule that could exceed $2.25 billion a year by FY2029 if delivery proceeds. The first 175 MW shows that the model is more than a presentation.
The common share is not a simple claim on $36 billion. It is the residual claim after construction spending, secured debt, interest, project preferred capital, noncontrolling interests, convertible securities, warrants and future operating costs. Only 12.4% of contracted load was live, and 75% of Q4 HPC revenue came from fit-out services. Those two numbers are the reality check on the headline.
That makes APLD promising but not yet cleanly underwritten. The appropriate posture is watchlist/preliminary, with an 18–24 month horizon and hard checkpoints at 3–6 and 12 months. A bullish conclusion should be earned by on-time MW, rising base rent, improving cash interest coverage and stable fully diluted economics—not by another increase in nominal pipeline value.
24 Related Merlintrader research
- Space, Defense & AI Stock Hubs 2026 — the broader infrastructure and strategic-technology directory.
- C3.ai ($AI) Stock Hub — a useful contrast between AI software economics and physical AI infrastructure.
- Dilution, ATMs and PIPEs — how equity and convertible funding affect existing shareholders.
- Free Catalyst Calendar — earnings, delivery windows and market events in one place.
Primary sources and reference links
- Applied Digital Form 10-K for the fiscal year ended May 31, 2026, filed July 29, 2026: business, portfolio table, lease-payment schedule, financial statements, segment data, cash flows, debt, equity, warrants, customer concentration, risk factors, controls and subsequent events.
- Fiscal Q4 and full-year 2026 results exhibit, July 27, 2026: 175 MW live, Q4 revenue composition, full-year adjusted measures, ChronoScale ownership and post-year-end developments.
- Initial CoreWeave lease Form 8-K, June 2, 2025: 250 MW, base economics, renewal options, termination conditions and the first CoreWeave warrant.
- Third CoreWeave lease Form 8-K, August 29, 2025: the additional 150 MW Building 4 lease and related warrant.
- Polaris Forge 2 lease filing, October 22, 2025: 200 MW and approximately $5 billion; the tenant was unnamed in this release and later identified as Oracle in the March 10 financing filing.
- Delta Forge 1 lease filing, April 23, 2026: 300 MW, approximately $7.5 billion and the remaining unnamed hyperscaler.
- Polaris Forge 3 announcement, May 20, 2026: 300 MW, base and renewal value, delivery window and power description.
- Delta Forge 2 lease announcement, June 8, 2026: 210 MW, approximately $5.2 billion base value and approximately $12.7 billion with all renewals.
- Polaris Forge 1 second-building delivery, July 1, 2026: 75 MW added and 175 MW total live capacity.
- $1.59 billion 7.0% senior secured notes Form 8-K, June 16, 2026.
- Secured revolving credit facility Form 8-K, June 26, 2026: $430 million commitments and $120 million remaining accordion.
- Form 10-Q for the six months ended November 30, 2025: second filing used to cross-check the $2.35 billion 9.25% secured-note issuance and debt table.
- $2.15 billion 6.75% senior secured notes Form 8-K, March 10, 2026.
- ChronoScale separation Form 8-K, May 5, 2026.
- August ownership filings: CEO Form 4, president Form 4, COO Form 4, CFO Form 4 and director-sale Form 4.
- Applied Digital leadership page, read August 28, 2026, cross-checked to the Form 10-K signatures and January president appointment filing.
Source method. Every material fundamental claim was checked first against the July 29 Form 10-K and then against the corresponding SEC 8-K, filed exhibit or official issuer release where available. This is a double primary-source check, but both layers are controlled by the issuer; it does not independently verify future tenant behavior, construction completion, the rating of the unnamed tenant or project returns. Market price and market capitalization are delayed provider data as of the timestamp shown and are not company fundamentals.
Get these reports in real time
Every Merlintrader stock hub, catalyst update and market brief is published to Telegram the moment it goes live. No paywall, no spam, just the research.
Disclaimer. This content is published by Merlintrader for educational and informational purposes only. It is independent journalism and research. It does not constitute investment advice, an investment recommendation, an offer or a solicitation to buy or sell any security, and it is not a research report within the meaning of applicable United States securities regulation. Nothing here should be read as a recommendation to buy, sell or hold $APLD or any other security.
Figures are taken from public filings with the U.S. Securities and Exchange Commission, company releases and clearly identified market-data providers, and are stated with their reference dates. Data can change without notice. Forward delivery windows, lease values including renewal options, pipeline capacity and tenant credit descriptions are management statements and may not be achieved. Readers should verify every figure against the primary source before acting.
AI infrastructure and data-center development are capital-intensive and carry substantial risk. Construction delays, cost overruns, utility and interconnection problems, tenant defaults, lease termination, refinancing, high-yield debt, preferred capital, convertible instruments and warrant or equity issuance can materially reduce or eliminate the value available to common shareholders. Contracted value is not revenue already earned, cash or profit. Loss of principal is possible.
Merlintrader may hold positions in securities mentioned. Some links on this page are affiliate or referral links, including those to Finviz, which may generate a commission at no cost to the reader. Full legal information is available on the disclaimer and terms of use and privacy pages.
Earnings dates, construction milestones, launch windows, regulatory deadlines and market catalysts in one free, filterable calendar.
Open the calendar →



