The market is pricing Applied Digital as a company whose contracted book is deeply mispriced. Our underwriting says the contracted book is worth roughly what it costs today — and that everything above $30 is a bet on leases that do not yet exist.
The author holds no position in APLD as at the date of publication, and held none while the research was produced. No position will be taken for 72 hours after publication.
The author has received no compensation from Applied Digital or any affiliate, holds no position exceeding 0.5% of the issuer's share capital, and has no business relationship with the company. This report was not shown to the issuer before publication.
The beat was construction billing. Of $258.7m of Q4 revenue and a 171% headline beat, $152.4m was tenant fit-out work at roughly a 4–5% margin. Strip it out and Q4 revenue was about $106m.
The contracted book justifies today's price. Not a discount to it. Which means the entire upside case is pipeline conversion, and should be argued as such.
The $1bn NOI target looks early. Reaching it within a year needs roughly 775 MW energised. The disclosed build schedule delivers closer to 600 MW on that timeline.
Applied Digital is a data centre landlord, not a cloud company. It secures cheap power, builds the shell, and leases it whole to one tenant on 15-year take-or-pay terms. So the only question that matters is how much contracted capacity is actually energised and collecting rent — and how much is still a drawing.
Energisation is scheduled to reach roughly 600 MW by mid-2027 and about 1,000 MW during 2028. Roughly 76% of the contracted book is with investment-grade hyperscalers, which is the single strongest fact in the bull case and the reason the contracted revenue is worth underwriting at all.
Every research note ends with a number. Almost none of them show you how to disagree with it. Below is the actual valuation, with every assumption exposed. Move an input and the fair value moves with it. Our published $26–34 band is simply this model run across a reasonable range of cap rates and NOI assumptions — not a separate opinion layered on top.
Default NOI per MW is derived: $36bn base-term revenue ÷ 15 years ÷ 1,410 MW, at the ~91% NOI margin disclosed on the HPC rent line.
These are not separate models. Each is the same model above with different inputs — click any card to load it and watch which assumption is doing the work.
Applied Digital buys land next to cheap, abundant power, secures a utility interconnection, builds a purpose-built high-density building, and leases that building to a single large tenant on a 15-year take-or-pay contract. The tenant supplies the GPUs. Applied Digital supplies the shell, the power, the cooling and the operations.
That distinction matters more than any other fact in this report. A cloud provider bears hardware obsolescence risk, utilisation risk and pricing risk on compute. A landlord bears construction risk, tenant credit risk and interest-rate risk. Applied Digital has spent eighteen months deliberately moving itself from the first category into the second — most visibly by separating its cloud services business into ChronoScale Holdings (Nasdaq: CHRN) in May 2026, of which it still owns roughly 96%.
The company was founded in 2021 as Applied Blockchain, renamed Applied Digital in November 2022, and is chaired and led by Wes Cummins, a former asset manager. It is headquartered in Dallas.
Understanding the mix is the single most useful piece of analysis an investor can do on this name. The headline growth rate blends a 91%-margin annuity with cost-plus construction billing, and the two deserve nothing like the same multiple.
| Revenue stream | FY26 Q4 | FY26 year | Economics | Durability |
|---|---|---|---|---|
| HPC base rent15-yr take-or-pay leases | $44.1m | $99.8m | ~91% NOI margin. The real asset. | Contracted 15 years |
| Tenant fit-out servicesBuilding out tenant space | $152.4m | $270.6m | $145.6m of cost in Q4 → roughly 4–5% margin. Cost-plus construction billing. | Non-recurring; ends on delivery |
| Tenant recoveries | $6.5m | $14.9m | Pass-through, zero margin by design. | Follows the lease |
| Bitcoin-mining hosting286 MW, two North Dakota sites | $37.3m | ~$150m | $12.5m quarterly segment profit on $113.8m of assets. Highest return on assets in the group. | Paid on capacity, not bitcoin price |
| ChronoScale (cloud)96%-owned, consolidated | $18.4m | $71.6m | Loss-making ($37.0m FY26 operating loss). Excluded from all non-GAAP figures. | Separately listed; strategically detached |
Of the $258.7 million of fourth-quarter revenue that produced a "407% year-on-year increase" and a 171% headline beat versus consensus, $152.4 million — 59% — was tenant fit-out services carrying roughly a 4–5% margin. Strip it out and Q4 revenue was approximately $106 million. The number is not fabricated and it is not improper: it is real work, really billed. But it is construction pass-through recognised in a lumpy block, and analysts modelling ~$95 million had almost no way to forecast it.
Street targets cluster at $70–77 against our $26–34. That is not a small difference of opinion, so it deserves an explicit explanation rather than a shrug.
Two things account for most of it. First, those targets were largely set before a 48% drawdown and have not been revisited. Second, and more substantively, the implied cap rates required to justify them sit below the prevailing 30-year Treasury yield — on single-tenant assets that have not yet been built. That is not a valuation we can reproduce under any assumption we are willing to defend in public.
Signed leases converting a material share of the 1.7 GW marketed pipeline at rates at or above the existing book; a financing package that reduces the preferred's claim ahead of the common; or energisation running ahead of the disclosed schedule. Any of the three moves the base case materially, and we will say so when it happens rather than quietly re-rating.
The natural comparable set is the power-and-shell cohort that pivoted from bitcoin mining to AI hosting: CoreWeave, IREN, TeraWulf, Cipher Mining and Core Scientific. On price to contracted base-term revenue, Applied Digital trades at roughly 0.21× — a $7.6bn market capitalisation against a $36bn contracted book. TeraWulf trades at approximately twice that ratio on a shorter book with weaker counterparty credit.
That is the single most favourable comparison available to the bulls, and it is genuinely favourable. It is also the comparison that most cleanly ignores the capital stack: the ratio charges nothing for the roughly $17bn of capital still to be spent, of which about $13bn is debt and $2.6bn is Macquarie project equity accreting toward a 1.8× return — all of it ranking ahead of the common shareholder.
Management has been consistently unwilling to issue stock to fund a campus at $27 a share, and consistently willing to issue it to themselves. The asymmetry is the tell.
The resolution most consistent with the evidence is that this is a team that thinks like owners about the business and like founders about the equity. That is a common and survivable combination — but it means the dilution line in the model should be treated as a permanent feature rather than a one-off, which is why share count is a slider above and not a constant.
BULLStart with what is contracted, because everything else is opinion. $36 billion of base-term revenue across 1,410 megawatts, roughly 76% of it with investment-grade hyperscalers, on 15-year take-or-pay terms. The market capitalisation is $7.6 billion. That is 0.21 times contracted revenue. TeraWulf trades at twice that ratio on a shorter, less investment-grade book.
BEARYou are pricing revenue you have not spent the money to earn. To collect that $36 billion you must first spend roughly $17 billion of capital that does not exist yet, of which about $13 billion will be debt sitting ahead of the common shareholder and $2.6 billion will be Macquarie project equity compounding to a 1.8× return ahead of the common shareholder. The ratio you just quoted deliberately ignores the capital stack. Run the bridge and you get $6 billion of residual equity at stabilisation, four and a half years out, discounted at a rate appropriate to a 5.7-beta pre-cash-flow equity. That is not obviously $27 a share.
BULLThe bridge undercounts. You are charging the full future debt against today's equity while giving no credit for four years of interim cash flow from capacity that energises progressively — 175 megawatts now, roughly 600 by mid-2027, a thousand by 2028. And you assign zero value to 1.7 gigawatts being actively marketed by a team that has signed three campuses in four months, at rising rates.
BEARThen be honest that the bull case is the pipeline, not the book. On the contracted book alone, at a 7% cap rate and a 14% discount rate, the stock is worth roughly what it costs today. Everything above $30 requires leases that do not exist. I am willing to underwrite a company for its pipeline. I am not willing to be told the contracted business is deeply mispriced when the arithmetic says it is fairly priced.
CHAIRThen the committee's position is neutral on price and constructive on the asset. The contracted book supports the current quote; the upside is real but unsigned; and the risk is that it is financed on the common shareholder's back. Rate it fairly valued, and revisit on the next signed lease.
Think of it as a landlord for computers. It finds places with a lot of cheap electricity, builds big specialised warehouses there, and rents each one to a single large technology company for fifteen years. The tenant brings the expensive chips. Applied Digital provides the building, the power and the cooling.
Revenue jumped enormously, and headlines said it beat expectations by 171%. But most of that jump was Applied Digital billing its tenants for construction work — money that passes almost straight through with only a few cents of profit on each dollar. The profitable part, the rent, is still small because most of the buildings are not finished.
The company has signed contracts worth about $36 billion. That sounds enormous next to a company worth $7.6 billion. The catch is that it must first spend around $17 billion building everything, and most of that money will be borrowed. Lenders get paid before shareholders do. Once you subtract what is owed, what is left for shareholders is worth roughly what the shares already cost.
None of this tells you whether to buy it. It tells you what you would be betting on if you did: not the contracts already signed, but the ones that have not been signed yet.