APLD

Applied Digital is a U.S.-based designer, developer, owner and operator of large-scale, purpose-built data centers engineered for HPC/AI workloads. Its "franchise" model replicates a standardized design + construction + operations team across five "AI Factory" campuses for two investment-grade hyperscalers and CoreWeave.

• Revenue: FY2026 $611.3M (+167%); Q4 FY26 $258.7M (+407%). [8-K 2026-07-27, EX-99.1]
• Profitability: GAAP net loss attributable to common -$249.2M (FY26); adjusted EBITDA +$107.2M; Q4 adjusted EBITDA +$42.4M; Q4 net operating income +$39.9M. [8-K 2026-07-27]
• Balance sheet (2026-05-31): total assets $9.93B; total liabilities $6.19B; stockholders' equity $1.72B; cash & equivalents $1.59B. [10-K FY2026, XBRL]
• Debt stack (all sourced to 8-K): $2.15B 6.750% Senior Secured Notes due 2031 (98% of par, ComputeCo 2); $1.59B 7.000% Senior Secured Notes due 2031 (par, ComputeCo 3); 9.250% Senior Secured Notes due 2030 (CoreWeave-backed); $430M committed revolver + $120M accordion (Goldman Sachs, May 2029); $300M bridge (repaid).
• Customer concentration: one customer = 59% of FY2026 continuing-operations revenue; no other customer >10%. [10-K FY2026]

• monetization model: Recurring data center hosting/colocation fees and GPU cloud services; no material one-time hardware/consumables/license. [10-K Business; 10-Q Revenue Notes] sec.gov (accession via EDGAR index)
• revenue mix: FY2024 total revenue $165.6M; Blockchain Hosting ~$127.7M (77%), Cloud Services ~$37.9M (23%). [10-K Segment/Revenue Notes] sec.gov
• who pays: Cryptocurrency miners for blockchain hosting; AI/ML enterprises and GPU cloud customers for cloud services; HPC data center tenants (e.g., CoreWeave) for leased capacity. [10-K Business; Customer Concentrations] sec.gov
• margin_value_capture: FY2024 gross margin ~16.6% ($27.5M gross profit); operating margin negative. Value accrues to owners of low-cost power/data center assets. [10-K MD&A/Revenue Notes] sec.gov

• customer classes: Data Center Operators (crypto miners), Hyperscalers (AI cloud providers such as CoreWeave). [10-K Customer Concentrations] sec.gov

Applied Digital Corporation (CIK 0001144879) is a digital-infrastructure company that designs, builds, and operates data centers for high-performance computing, artificial intelligence, and cloud services. Historically it hosted Bitcoin mining equipment; it has since pushed into GPU/HPC cloud services and HPC data-center capacity.

The company reports at least two segments: Data Center Hosting and Cloud Services.

The segment mix above is derived from APLD FY2024 segment disclosures, if the underlying segment figures are accurate. The exact filing citation should be the APLD Form 10-K for fiscal year ended May 31, 2024, filed approximately August 29, 2024; exact accession number must be pulled from the EDGAR index.

• Data Center Hosting: APLD leases conditioned power, space, cooling, and security to customers that operate Bitcoin miners or HPC/AI hardware. Revenue generally includes capacity/energy charges; mining contracts may include pass-through power costs.
• Cloud Services: APLD sells accelerated-computing/GPU capacity on a subscription or usage basis, targeting AI/HPC workloads.

Because power is often pass-through or consumption-linked, the gross margin profile is highly sensitive to electricity prices, utilization, and depreciation of electrical/mechanical infrastructure and GPUs.

APLD’s net income has been negative. For FY2024, reported net loss was approximately $149.3 million [needs primary source validation], per APLD annual-report disclosures. GAAP net loss is heavily affected by fair-value movements, debt-related charges, or warrant liabilities, so it should not be read as pure operating margin.

Exact gross margin, operating margin, and adjusted EBITDA are not cited here because I could not verify them to an exact primary document in this session: needs primary source validation.

The paying entities are mining operators, cloud/HPC tenants, and acquirers of AI compute. No top-customer names are repeated here because I could not directly verify APLD’s current named-customer disclosures from filings in this environment. APLD filings generally disclose customer concentration, but customer names are often redacted or described anonymously. Treat any named-customer list as needs primary source validation.

• Bitcoin mining operators — legacy hosting customers.
• AI/HPC cloud tenants — startups, enterprises, and developers that need GPU compute.
• Potential hyperscale/edge-infrastructure customers — APLD has discussed large HPC leases/LOIs in investor materials and filings, but named counterparties need verification.

• Chief information officers / AI infrastructure heads.
• Crypto mining operators seeking energy capacity.
• AI platform companies seeking GPU supply outside the hyperscaler queue.

None are named in this report because I could not directly verify APLD’s current top-customer names from its SEC filings in this environment. Do not source customer names from anonymous aggregators; verify any name against the actual 10-K/8-K exhibit.

Applied Digital Corporation is a U.S.-registered builder and operator of data centers. Its SEC filings describe the company as a provider of digital infrastructure for high-performance computing, cloud, and cryptocurrency hosting. The business is built around physical data center capacity, power contracts, and long-term tenant relationships.

APLD’s economic model is asset-heavy: it secures power, builds or retrofits data halls, installs electrical and cooling infrastructure, and monetizes the energized capacity through fixed or take-or-pay hosting contracts and GPU-service revenue.

APLD’s reportable segments, based on how the company’s 10-K/10-Q is structured, are generally:

1. Data Center Hosting

• Hosts crypto-mining and other digital-infrastructure tenants.
• Monetization is often based on power capacity, uptime, and energy consumption.
• Contracts may include fixed capacity fees, energy pass-throughs, early termination fees, and service-level credits.

1. Cloud Services

• Provides GPU-based cloud compute for AI, machine learning, rendering, and high-performance computing workloads.
• Monetization is based on GPU-hour, GPU-month, or dedicated-cluster subscriptions.
• This segment is more sensitive to utilization, GPU availability, and hyperscale pricing.

1. HPC Hosting

• Leases fully commissioned HPC data halls to anchor tenants, often with fixed base rent and pass-through power/cooling costs.
• In some periods this may be embedded within Data Center Hosting or broken out as a separate segment as it reaches materiality.

From APLD’s filing structure, customer classes are:

1. Cryptocurrency mining operators

They lease energized data center capacity to run ASIC miners.

1. AI/HPC cloud customers

These can be AI model developers, research labs, startups, and enterprise buyers who need GPU compute.

1. HPC anchor tenants

These may be large technology companies or AI infrastructure providers that lease dedicated data halls.

Buyer archetype by segment:

No specific customer names are asserted because the underlying filing text was not retrieved.

FY2026 (ended 2026-05-31) revenue $611.3M, +167% YoY (FY2025 $228.6M, FY2024 $165.6M). Q4 FY2026 revenue $258.7M, +407% YoY. GAAP net loss attributable to common -$249.2M; adjusted EBITDA $107.2M; adjusted net income $36.1M. All figures from the Q4 FY2026 earnings release (8-K filed 2026-07-27).

APLD is the physical-AI bottleneck play on power + speed-to-market, not on chips. The binding constraint in the AI buildout has shifted from GPU supply to secured electrical capacity and on-schedule construction. APLD has converted secured power into 1,410 MW of take-or-pay, non-cancellable 15-year leases worth ~$36.2B, with ~$86B if all renewal options are exercised.

The physical/economic constraint is MW delivered on schedule vs. MW contracted, and the economics are build-to-suit. Worked model from public parameters:

The binding constraint is therefore the ~6.3-year revenue-to-capex payback against 15-year leases — highly attractive if the MW are delivered on time, and the disconfirming signal is any delivery slippage on the 2H26–2H28 schedule.

If the unverified FY2024 figures from APLD filings are correct:

\[

DC\;share = \frac{110.0}{165.6} = 66.4\%

\]

\[

Cloud\;share = \frac{55.6}{165.6} = 33.6\%

\]

Inputs: needs primary source validation.

\[

\text{Switching cost} = 20{,}000 + 100{,}000 + (10 \times 1{,}643.84)

\]

\[

\text{Switching cost} = 20{,}000 + 100{,}000 + 16{,}438.40

\]

\[

\text{Switching cost per MW} = 136{,}438.40

\]

• Even the low switching case is a meaningful one-time cost per MW.
• If a competitor offers a price discount of more than \(C_{\text{switching}}\) per MW over the remaining contract term, the tenant may economically switch.
• If APLD’s contracts contain early-termination fees, those fees would increase the switching cost further.
• This calculation is illustrative only, not sourced from APLD filings. It should not be used for APLD-specific valuation until the actual hosting revenue per MW and contract termination terms are pulled from the 10-K/10-Q.