Applied Digital has 250 MW of AI data centres paying rent and 1,410 MW leased for about $36bn over fifteen years, and its first buildings have arrived when it said they would. Valued as a landlord would value it, the finished portfolio is worth about $36.5bn; debt, a Macquarie preferred stake compounding at 12.75% and the company’s own costs take most of that first. What is left is worth $20.68 to $34.02 a share, most likely $26.57. At $18.00 the price needs the finished campuses to sell at an 8.2% yield, 122 basis points above ours, or no expansion leases and a taxed exit together. We rate the shares undervalued.
The leases are signed and the buildings are arriving on time, but the rent does not rise. The second 75 MW phase of Building 2 at Polaris Forge 1 was ready for service on 1 October, three months after the first, taking live capacity to 250 MW. Leases cover 1,410 MW for about $36bn over fifteen-year base terms. The minimum payments in the Form 10-Q step from $468.0m for the rest of fiscal 2027 to $2,371.4m a year from fiscal 2030, and are identical in fiscal 2030 and 2031: the contracted rent carries no fixed annual increase. Base rent was $65.8m in the quarter at an 89% NOI margin; tenant fit-out work, billed at about a 4% margin, was 54% of revenue.
Most of what the buildings are worth is claimed before the common shareholder. When everything in our base case is paying rent, in November 2028, the 1,660 MW earn about $2.56bn of net operating income, worth $36.5bn at a 7.0% cap rate. Ahead of the shares by then: about $15.2bn of debt, $4.5bn of Macquarie preferred accreting at 12.75%, Macquarie’s 13.5% share of the campus equity ($2.1bn), and corporate costs and stock pay worth $3.7bn once capitalised. In today’s money that turns $76.08 a share of asset value into $22.83 of equity; half of ChronoScale’s market value and other assets add $3.73.
The price needs an exit yield above 8%; it sits below each of our cautious readings taken alone. At $18.00 the base case is reproduced only at an 8.22% exit cap rate instead of 7.0%. Two judgements carry the rating. The base case credits 250 MW of expansion leases that were in advanced negotiation in July and still unsigned in October; without them it is $20.19. The framework values the buildings as an untaxed owner would; deducting the $3.69bn of tax a sale would trigger gives $19.07. Each alone leaves the value above the price. Take away both and it falls to about $14–16.
Applied Digital is a landlord, not a cloud company. It finds land next to cheap power, builds high-density, liquid-cooled halls and leases each campus to one tenant for fifteen years on take-or-pay terms; the tenant brings the chips. Its value therefore lies in buildings that will exist and be leased in two years but are mostly under construction today, and its income statement — a quarter of fit-out billing and part-built rent — says little about that. We value it as the owner of a finished portfolio would be valued, then subtract every claim that will rank ahead of the common shares when it is finished.
On that basis the base case is $26.57 a share, 48% above the price. Run the model in Section 02 backwards and the price needs one of three things, each on its own with everything else at the base case: an exit cap rate of 8.22% instead of 7.0%; an NOI margin of 79% on base rent instead of 90%, against 89–91% reported so far; or no expansion leases at a 7.3% cap rate. The discount rate alone cannot do it: the price would need 45% a year on the equity. Put plainly, the market is valuing the signed leases at a yield 122 basis points above ours and giving nothing for the 250 MW of expansions or for anything beyond them. The price sits 13% below the bottom of our $20.68–34.02 band.
The signature measure for a levered landlord is the queue in front of the shareholder. The first bar is capacity, in megawatts of critical IT load; the second is our base case on the day the last lease starts paying rent, the $36.5bn of asset value at a 7.0% cap rate divided by claim, drawn to scale.
Capacity from the Form 10-Q and the 7 October call. The second bar is derived: debt includes $9.4bn of new project debt assumed at 80% of remaining build cost, and the Macquarie stake in the three campuses not yet financed is assumed on the same terms as its existing investment. The common shareholders’ slice is before the bitcoin-hosting business, cash, other claims and discounting, which Section 03 adds. Of every dollar the finished buildings are worth, about 30 cents is left for the common shareholder; that is why a small change in the yield moves the share price a long way. Each quarter-point on the exit cap rate is worth about $2.06 a share.
About 250 MW of expansion leases — roughly 100 MW at Polaris Forge 2 and 150 MW at a Delta Forge campus — were described as in advanced negotiation in July and are now expected “by calendar year-end” at pricing the chief executive put “north of 15%” above current rents. They are in our base case and worth $6.38 a share. Signed at that rent, they would be the first direct evidence of pricing power; still unsigned in January, the base case falls to $20.19 and the band to $15.53–26.11, which would put the shares inside it.
The rate backdrop is not friendly to a 7% yield. In late October the 10-year Treasury yielded about 5.0% and the 30-year about 5.3%; the upper bound of the federal funds target was 4.00%. A 7.0% exit yield is about 170 basis points over the long bond: thin compensation for single-tenant, single-purpose buildings, offset by fifteen-year take-or-pay leases with investment-grade tenants that no bond has. The band runs the yield from 6.25% to 7.75%.
Stabilised NOI $2.56bn a year at 7.00% · Macquarie common units $2.1bn, included above · shares with the converts · the exit cap rate does almost all the work, and it is ours, not a disclosure.
This is the model that produced every figure in this report, reduced to its ten moving parts. The finished portfolio’s net operating income is capitalised at the exit cap rate, as a buyer of stabilised data centres would value it. From that come, in order: the corporate cash costs and stock pay a listed owner carries for ever, capitalised at the same rate; debt and the other claims at stabilisation, net of the cash left and of the bitcoin-hosting business at five times its EBITDA; Macquarie’s preferred equity, accreted at 12.75%; and Macquarie’s common units, 13.5% of the equity of the campus vehicles it funds, which the model recomputes from the cap rate. What remains is the common equity on the day the last lease starts paying rent, discounted to 26 October 2026 at the equity rate. ChronoScale at half its market value, the Base Electron note, other investments and the cash from the warrants are added at today’s values. The share count is basic shares plus warrants; the convertible notes convert at the value being solved for, net of the capped call.
What the controls hold fixed. The scenario cards load whole published cases, including how much of the portfolio sits inside Macquarie-funded vehicles and the debt in them, which have no control. Moving a slider bends the loaded case; it does not re-run the month-by-month funding behind it. So changing the megawatts or the years here does not change the debt or the preferred, which in the full model depend on what is built and when: use the scenario cards for those. The exit cap rate and the discount rate are exact. The fair-value band in Section 03 is this model run across the cap rate and the discount rate, and every cell of that grid is reproducible from the two controls.
Why these inputs. 1,660 MW (assumed). The 1,410 MW leased, plus the 250 MW of expansion leases management expects by the end of the calendar year, live in late 2027 and 2028. NOI of $1.541m per MW a year (derived). Contracted rent of $1.68m per MW-year — contract totals divided by fifteen years, scaled so the five campuses add up to the $2,371.4m a year of minimum payments in the Form 10-Q — and expansion rent 15% above Polaris Forge 2’s, at a 90% NOI margin against 89% in the quarter and 91% in the one before. Exit cap rate 7.0% (assumed). Our yield for fifteen-year, take-or-pay, mostly investment-grade single-tenant leases with no fixed escalator. Corporate costs and stock pay, $260m a year (assumed). Cash overheads and development spending of $165m and stock pay of $95m in the first year after the build, against $69.2m of stock pay in the last quarter alone, $41.3m of it a one-time vesting. Equity rate 14.0% (assumed). For common equity behind about 80% project debt and a preferred tranche; it matters less than the cap rate because stabilisation is 2.0 years away.
Derivations, base case. Stabilisation is 6 November 2028, when the last base-case lease starts paying rent. Net debt and other claims of $15.21bn: debt of $15.22bn ($6.27bn at 31 August 2026 excluding the convertible notes, less amortisation, plus $9.40bn of new project debt at 80% of remaining build cost), plus $267m of Series E-1 preferred, finance leases, customer deposits and accrued interest, plus a cash shortfall of $25m after the month-by-month funding run, less the bitcoin-hosting business at five times $60m of EBITDA. Macquarie preferred of $4.55bn: the $1.825bn invested in APLD HPC TopCo 2 from 2025, accreted to $2.55bn, and $2.00bn of new preferred assumed to fund the balance of the three unfinanced campuses, both at 12.75% compounding twice a year. Shares: 302.4m on the cover of the Form 10-Q (6 October 2026) plus 28.1m warrants at $8.61, exercised for cash; the $450m of 2.75% convertible notes convert at $9.75 with a capped call to $14.72, adding 37.5m net shares at the base-case value. Unvested awards are excluded because stock pay is charged as a cost. Other assets of $1.37bn: ChronoScale at half its 23 October 2026 market value ($1.03bn), the $58.5m Base Electron demand note, the Babcock & Wilcox shares and warrants and other stakes ($41.0m), and $242m of warrant cash. Amounts at stabilisation are discounted 2.03 years at 14.0%, a factor of 0.766.
The number above is not a fact about Applied Digital. It is what these assumptions produce, and the one that decides it is not in any filing: the yield at which a buyer would own the finished campuses. Because the common shareholder keeps only about 30 cents of each dollar the buildings are worth, each quarter-point on that yield moves the value by about $2.06 a share. The rating holds for an exit cap rate up to about 7.43% with the rest of our base case unchanged; above that, the band’s lower edge falls under the price and the reading turns fairly valued. The second judgement is the expansion leases: without them the band is $15.53–26.11 and the price sits inside it. Change the inputs until they are yours, and use the number that falls out.
Each card loads its published case into the model above and prints its value: $10.07, $26.57, $43.36 and $62.05. All four use the same 7.0% exit cap rate and 14.0% equity rate, so they differ only in what gets built, when, and at what cost; delivery, tenant and funding risk sit in the cases and the weights, not in the rates. Weighting the four at 30 / 45 / 20 / 5 gives $26.75.
| Input | Bear (30%) | Base (45%) | Bull (20%) | Super-bull (5%) |
|---|---|---|---|---|
| MW paying rent at stabilisation | 1,410 | 1,660 | 2,410 | 3,410 |
| Stabilisation (last rent start) | May 2029 | Nov 2028 | Oct 2030 | Oct 2031 |
| Delay to company dates | 3–9 months | 0–3 months | None | None |
| Contracted rent, $m per MW-year | 1.68 | 1.68 | 1.68 | 1.68 |
| Expansion and new-lease rent, $m per MW-year | — | 1.88 | 1.98 | 1.98–2.07 |
| NOI margin on base rent | 86% | 90% | 91% | 91% |
| Stabilised NOI | $2,039m | $2,558m | $3,958m | $5,843m |
| Remaining build cost against base | +15% | base | −5% | −5% |
| Cost of new project debt | 8.00% | 7.00% | 6.25% | 6.25% |
| Corporate costs and stock pay at stabilisation | $281m | $260m | $274m | $281m |
| Bitcoin hosting at stabilisation | 3× fading EBITDA | 5× $60m | 5× $60m | 5× $60m |
| Debt at stabilisation | $14,192m | $15,220m | $21,362m | $30,545m |
| Macquarie preferred, accreted | $5,315m | $4,548m | $5,780m | $6,541m |
| Common to all | Exit cap rate 7.0% and equity discount rate 14.0%; valued at 26 October 2026 from the 31 August 2026 balance sheet; stock pay charged as a cost; new project debt at 80% of remaining build cost; Macquarie funds the balance of the three unfinanced campuses on its TopCo 2 terms. Bear also carries 10% higher overheads and a fading hosting business. | |||
| Value per share | $10.07 | $26.57 | $43.36 | $62.05 |
Contracted rent is derived from contract totals aligned to the minimum payments in the Form 10-Q, flat, with no fixed escalator. Expansion rent is Polaris Forge 2’s × 1.15 (chief executive, 7 October). Remaining build cost at 31 August is derived: about $10.7bn for the leased campuses and $2.75bn for the expansions, roughly $12.4m per MW all-in for the leased 1,410 MW against the company’s $11–13m range.
To reach the day the last lease starts paying rent, the company must spend, borrow and raise money. We run its cash month by month from 1 September 2026: rent as each building starts, construction spending, interest, overheads, new project debt and new Macquarie preferred. Whatever cash is left, or missing, at stabilisation belongs to the shareholders.
| $ millions, base case | Sep–Dec 2026 | 2027 | Jan–6 Nov 2028 | Total |
|---|---|---|---|---|
| NOI from leases in service | 131 | 907 | 1,510 | 2,548 |
| Bitcoin-hosting EBITDA | 20 | 60 | 51 | 131 |
| Corporate cash costs | (45) | (140) | (132) | (317) |
| Stock pay, charged as a cost | (45) | (125) | (94) | (264) |
| Interest on existing and new debt | (175) | (713) | (846) | (1,734) |
| Capital spending, incl. $1,056m of construction bills at 31 Aug | (3,217) | (6,821) | (3,877) | (13,915) |
| Interest income, amortisation, preferred dividends | 21 | (161) | (248) | (388) |
| New project debt, 80% of new build cost | 984 | 4,682 | 3,023 | 8,689 |
| New Macquarie preferred, balance of the 810 MW build | 246 | 966 | 514 | 1,726 |
| Cash: 31 Aug 2026 → at stabilisation | 3,678 | (25) |
Sonde estimates. The cash line at stabilisation is after $891m of construction still owed then, 80% financed. Debt then: $15,220m, of which $9,402m new. Macquarie preferred $2,548m on the existing TopCo 2 investment and $2,000m on the new one, both accreted at 12.75%.
| Base case | $bn | $ per share, today |
|---|---|---|
| Stabilised NOI of $2.56bn, capitalised at 7.0% | 36.5 | +76.08 |
| − Corporate cash costs and stock pay of $260m a year, capitalised | −3.7 | −7.73 |
| + Bitcoin-hosting business at five times $60m of EBITDA | 0.3 | +0.62 |
| − Debt at stabilisation | −15.2 | −31.68 |
| − Macquarie preferred, accreted at 12.75% | −4.5 | −9.47 |
| − Macquarie common units, 13.5% of campus equity | −2.1 | −4.38 |
| − Other claims and cash shortfall | −0.3 | −0.61 |
| Equity at stabilisation, 6 November 2028 | 11.0 | 22.83 |
| + ChronoScale stake at half its market value, today | 1.0 | +2.81 |
| + Base Electron note, other investments and warrant cash, today | 0.3 | +0.93 |
| Value per share | $26.57 |
Amounts at stabilisation are discounted 2.03 years at 14.0%, a factor of 0.766; ChronoScale, investments and warrant cash are at today’s values. Per-share figures on 368.0m shares: 302.4m outstanding, 28.1m warrants and 37.5m net convertible shares. Rows may not add exactly because of rounding.
The grid holds the base case fixed and varies the exit cap rate against the equity discount rate. Its corners are the published fair-value band, $20.68–34.02: an exit cap rate of 6.25–7.75% by an equity rate of 13–15%.
| Exit cap rate ↓ / equity rate → | 13.0% | 14.0% | 15.0% |
|---|---|---|---|
| 6.25% | $34.02 | $33.50 | $32.98 |
| 6.75% | $29.15 | $28.71 | $28.27 |
| 7.00% base | $26.97 | $26.57 | $26.17 |
| 7.25% | $24.94 | $24.58 | $24.22 |
| 7.75% | $21.28 | $20.98 | $20.68 |
The $18.00 price lies below every cell, 13% under the lowest: undervalued. The discount rate matters little because stabilisation is two years away; the cap rate carries the band. Run across the same grid, the probability-weighted value spans $20.45–34.76. The second grid holds the rates at the base case and varies what the buildings cost against what they earn:
| Remaining build cost ↓ / NOI margin → | 86% | 88% | 90% | 92% |
|---|---|---|---|---|
| −10% | $25.62 | $27.19 | $28.76 | $30.33 |
| Base | $23.43 | $25.00 | $26.57 | $28.14 |
| +10% | $21.25 | $22.82 | $24.38 | $25.95 |
| +20% | $19.08 | $20.64 | $22.21 | $23.78 |
Base case at a 7.0% cap rate and a 14.0% equity rate. A 10% overrun on the remaining build costs $2.18 a share.
Modal base case $26.57, 48% above the price. Fair-value band $20.68–34.02, the exit cap rate 6.25–7.75% by equity rate 13–15% grid, which anchors the rating; the price is below it: undervalued. Probability-weighted value $26.75 across the four cases at our own weights, 49% above the price and almost on the base case. There is no material skew to name: the bear sits $16.50 below the base and the bull $16.79 above it; the bear carries more weight and the super-bull adds a tail built on the company’s own plan, and the two almost cancel.
| Single input changed, all else at base | Price implies | Our base |
|---|---|---|
| Exit cap rate | 8.22% | 7.00% |
| NOI margin on base rent | 79% | 90% |
| No expansion leases, and an exit cap rate of | 7.35% | 7.00%, with 250 MW |
| Equity discount rate | 45% | 14.0% |
The market is valuing even the signed leases well below how we value them, at a yield 122 basis points higher, and giving nothing for the 250 MW of expansions or anything beyond them. An NOI margin of 79% would be 10 points below the lowest the company has reported. A cap rate above 8% on fifteen-year leases with investment-grade tenants is not impossible with the long bond at 5.3%, and it is the reading the price asks a buyer to accept.
| Scenario | Value | Against price | Weight | Contribution |
|---|---|---|---|---|
| Bear | $10.07 | −44% | 30% | $3.02 |
| Base | $26.57 | +48% | 45% | $11.96 |
| Bull | $43.36 | +141% | 20% | $8.67 |
| Super-bull | $62.05 | +245% | 5% | $3.10 |
| Probability-weighted | $26.75 | +49% | 100% | $26.75 |
We weight the bear at 30% because two of the three things it needs — delays and an overrun — are ordinary in construction at this scale, and four campuses must be delivered at once from 2027, two in states where the company has never built. We weight it no higher because every delivery the company has dated so far has arrived on time. Weights are our judgement, not a model output.
| Risk | Priced in | Not also in |
|---|---|---|
| Construction delay | Rent start dates: the southern campuses and Polaris Forge 3 three months behind the company’s dates in the base case; in the bear, three months more in North Dakota and nine at Delta Forge 1 and 2 and Polaris Forge 3 | The discount rate |
| Build-cost overrun | Remaining build cost: within the company’s $11–13m per MW range in the base case, 15% over in the bear | The cap rate |
| Expansion leases not signed | Volumes: 250 MW in the base case, none in the bear; and the weights | The cap rate |
| Tenant credit and renewal | The 7.0% exit cap rate, set for single-tenant fifteen-year leases | The discount rate or the cash flows |
| New campuses, the 3.5–4 GW ambition | Volumes: 750 MW in the bull, 1,750 MW in the super-bull; and their weights | The base case |
| Flat rent, no escalator | The exit cap rate; listed below as a known weakness | — |
| Cost of new project debt | Interest during construction: 8.0% bear, 7.0% base, 6.25% bull | The discount rate |
| Operating margin | NOI margin on base rent: 86% bear, 90% base, 91% bull | — |
| Macquarie preferred and common units | Claims at stabilisation, with the common units assumed on the three new campuses in every case | The discount rate |
| Dilution: warrants, convertibles, stock pay, Series G | The share count at the value solved for; stock pay charged as a cost; any funding gap deducted from the equity | — |
| Corporate tax on an exit | Not priced in the framework: an unpriced item, disclosed here and in the known weaknesses ($19.07 with it) | — |
| Related-party power contract | Not priced; price terms undisclosed; Sections 07 and 08 | — |
The usual multiples say little about a landlord whose rent roll is about a ninth built. The useful question is what the market pays per signed megawatt, and what is still to be spent to collect it.
| $ millions | Value | Note |
|---|---|---|
| Market value | 5,443 | 302.4m shares (Form 10-Q cover, 6 Oct 2026) × $18.00 |
| + Debt at face | 6,722 | Secured notes $6,090m, convertible notes $450m, revolver and other $182m |
| + Macquarie preferred, accreted | 1,982 | $1,825m invested, accreted at 12.75% to 26 Oct 2026; carried at $2,011m |
| + Series E-1 preferred and finance leases | 108 | |
| − Cash and restricted cash | (3,678) | 31 Aug 2026; $3.4bn of it inside TopCo 2 |
| Enterprise value | 10,576 | Before the capital still to be spent and raised |
| Company | Mkt value | EV | Leased MW | Contracted rent | EV per leased MW | Mkt value / contracted rent |
|---|---|---|---|---|---|---|
| Applied Digital $18.00 | $5.4bn | $10.6bn | 1,410 | ≈$36bn | $7.5m | 0.15× |
| TeraWulf $11.60 | $5.8bn | $8.5bn | 839 | ≈$27.3bn | $10.1m | 0.21× |
| Core Scientific $12.89 | $4.1bn | $5.9bn | 1,120 | >$24bn | $5.3m | 0.17× |
Applied Digital at the last Nasdaq pre-market trade at 9:04 a.m. Eastern on 26 October 2026; TeraWulf and Core Scientific at their last pre-market trades within two minutes of it. Market value is price × the share count on the cover of each company’s latest Form 10-Q (APLD 6 October, WULF 31 July, CORZ 23 July 2026). EV adds debt at face, convertibles included, and deducts cash and restricted cash from the latest 10-Q (APLD 31 August; WULF and CORZ 30 June 2026; CORZ also its bitcoin); APLD also adds the Macquarie preferred. Leased MW and contracted rent are company disclosures. Tenants: Applied Digital, an investment-grade hyperscaler (810 MW), CoreWeave (400 MW) and a second hyperscaler; TeraWulf, Fluidstack (Google-backed), Core42 and Anthropic, with mining about a quarter of revenue; Core Scientific, CoreWeave (590 MW) and AMD and an AMD-backed neocloud (530 MW). Not like for like: Core Scientific’s CoreWeave sites were largely paid for by the tenant, and TeraWulf’s Anthropic lease runs twenty years.
| Measure | APLD | Why it misleads here |
|---|---|---|
| P/E, trailing and forward | n/m | Losses; a landlord in its building phase should not have earnings yet |
| EV / Q1 adjusted EBITDA, annualised | 41× | One quarter with 100–175 MW live, against 1.4 GW signed |
| EV / FY2029 contracted rent | 4.5× | Minimum payments of $2,360.6m in FY2029; before the capex still to spend and the claims raised to build it |
| Price / book | 3.4× | Book equity $1,602.5m at 31 Aug 2026 |
| Market value / contracted rent | 0.15× | $36bn over the base terms |
Cheap, fair or expensive? In the middle of its peers on what they pay per signed megawatt. Applied Digital at about $7.5m of enterprise value per leased megawatt sits between Core Scientific ($5.3m, where the tenant paid for much of the building) and TeraWulf ($10.1m). All three trade at 0.15–0.21 times their contracted rent. None of these figures counts the build cost still to be spent, which is the main difference between them: Applied Digital has about $14.5bn of construction spending ahead of it on the leased portfolio and the expansions, $1.1bn of it bills already incurred, to be funded mainly with project debt and preferred equity. The discount to replacement cost that the per-megawatt figure seems to show is therefore not a discount at all; it is the claim that the next round of capital will take. That claim is deducted explicitly in our valuation, so we do not charge the peer-group risks again in the cap rate or the discount rate.
| What would expand the multiple | What would compress it |
|---|---|
| Expansion leases signed at rents 15% or more above current | A delay at Delta Forge 1 or Polaris Forge 3 |
| Investment-grade financing for the next three campuses | Macquarie terms for the next campuses at or above 12.75%, with a larger common share |
| A sale of a stabilised campus, or a REIT or infrastructure buyer, that sets a yield below 7% | Tenant credit: a downgrade or restructuring at CoreWeave |
| Stock pay falling toward a single-digit share of NOI | Continued Series G issuance below the value per share |
No corporate tax on the exit value. The framework values the buildings as an untaxed owner, such as a REIT, would. A C-corporation that sold them would owe about $3.69bn of tax on the gain over their tax basis and its losses; deducting it gives $19.07. An unlevered cash-flow valuation to stabilisation at 10%, exiting at the same 7.0% less that tax, gives about $21.23. Both are above the price; without the expansion leases as well, the two give about $14 and $16, below it. Unsigned expansion leases in the base case, worth $6.38 a share. ChronoScale at half its market value: only about 4% of its shares trade; at full market value the base case is $29.31, at nil $23.83. Macquarie terms for the next three campuses are assumed: the same 13.5% common share as in TopCo 2; without it the base case is $28.72. The 1.80× minimum return and the 120% early-redemption terms are not modelled, which flatters the base case. Flat rent capitalised at 7%: a buyer paying 7% for flat rent accepts a lower return than on escalating leases; the grid’s 7.75% edge, $20.98, covers that view. Build cost and rent dates are derived: the company gives an $11–13m per MW range but no remaining budget, and rent dates are calibrated to the minimum-payment schedule; physical delivery could lag it. Unmodelled: the 3.5–4 GW ambition beyond our bull cases, Finland, lease renewals, a conversion of the Ellendale hosting site, and the price terms of the Base Electron power contract.
Applied Digital reported its fiscal first quarter of 2027, the three months to 31 August 2026, on 7 October after the close, and filed its Form 10-Q the same day. Wes Cummins, chairman and chief executive, and Saidal Mohmand, chief financial officer, took questions. The company gives no revenue or earnings guidance, so there is no beat or miss to score; the tests are the build, the rent line and the leases.
| Q1 FY2027 | Value | Change y/y | Why it matters |
|---|---|---|---|
| Total revenue | $341.9m | +322% | Fit-out work is 54% of it |
| HPC hosting revenue | $262.6m | n/m | Base rent $65.8m, fit-out $183.5m, recoveries $13.3m |
| Net operating income | $58.8m | n/m | 89% of base rent; 91% in Q4 FY2026 |
| Adjusted EBITDA | $64.4m | from $0.5m | Excludes ChronoScale; adds back $59.4m of stock pay |
| Data Center Hosting | $37.8m | 0% | Segment profit $13.3m on $111.9m of assets |
| ChronoScale revenue | $41.5m | +148% | $23.0m of it GPU hardware resale; operating loss $25.4m |
| SG&A | $114.7m | +289% | Stock pay $65.4m, of which $41.3m a one-time vesting of performance units |
| Non-cash losses on Babcock & Wilcox | $67.5m | n/m | Warrants $56.1m, shares $11.4m |
| Net loss to common, continuing | $(221.0)m | from $(18.5)m | $(0.76) a share; adjusted $(0.01) |
| Cash and restricted cash | $3.7bn | $2.9bn unrestricted | |
| Debt | $6.7bn at face | More than 80% due in fiscal 2031 or later (CFO) |
Sources: Q1 FY2027 results release and Form 10-Q, 7 October 2026; call remarks from a published transcript of the 7 October call. Adjusted figures are the company’s and exclude ChronoScale.
One: the build is on track, and the next year is the big one. The second 75 MW phase of Building 2 went live on 1 October, taking Polaris Forge 1 to 250 MW. Management expects more than 600 MW in service over the next twelve months, against 250 MW over the last twelve, and about 300 MW in North Dakota by the end of 2026.
Two: the expansion leases are still pending. About 250 MW of expansions are now expected “by calendar year-end” at “materially higher pricing”, which the chief executive put at “north of 15% plus increases on the lease rate”, possibly with longer terms. In July the same expansions — about 100 MW at Polaris Forge 2 and 150 MW at a Delta Forge campus — were described as in advanced negotiation. None was signed by 7 October.
Three: the next three campuses will be financed like the first two. Polaris Forge 1 and 2 are “fully funded”. Polaris Forge 3, Delta Forge 1 and Delta Forge 2, all leased to the same tier-one hyperscaler, are next, through Macquarie preferred equity followed by project debt, “so APLD shareholders keep majority ownership”. The chief financial officer cited investment-grade bond products and a preference for structures that can be refinanced in one to two years.
Four: the ambition is 3.5–4 GW by the end of 2030. The chief executive first said 4 GW and later corrected it. He named construction, not power, as the binding constraint, and said key components were locked up two years ago, including the full factory output of some suppliers for four years.
Five: a first step abroad. Up to about 1 GW of power in Finland, the first 100 MW available in 2028. Management called it measured and opportunistic, with an exit before further payments. ChronoScale, it said, is on track for $1bn of annual recurring revenue in 2027, helped by a 50 MW Microsoft deployment.
The operating markers: about 300 MW live in North Dakota by the end of 2026; more than 600 MW placed in service over the next twelve months; revenue from the southern campuses beginning in the first half of 2027 (Delta Forge 1; the 10-Q dates Delta Forge 2 to the first half of 2028); about 250 MW of expansion leases by year-end; and a 3.5–4 GW operating portfolio by the end of 2030. The goal of a $1bn annual NOI run-rate, set out on the July call, was not restated in the remarks we reviewed; at our rent and margin assumptions it needs about 640 MW paying rent, all of Polaris Forge 1 and 2 plus some 40 MW more.
The tone was confident on demand and delivery and careful on numbers. The vocabulary has shifted from capacity and speed to longevity: “30-plus-year” campuses, “terminal values”, the “scarcity value” of powered, community-supported sites. The hedges were consistent, and specific figures were absent on the topics an investor most needs: build cost per megawatt, capital spending, the size of the next Macquarie investment, and the price terms of the Base Electron power contract.
| Firm | Question | Answer, in short |
|---|---|---|
| Northland Securities | Why hyperscalers want North Dakota; bond-market options | Several tier-one hyperscalers in talks; another hyperscaler is self-building about 1 GW in the state. Project finance and investment-grade bond products are open as alternatives to high yield. |
| B. Riley Securities | Finland; the new emphasis on campus life | Site bought, power held under options with an exit before further payments. Buildings are a “fourth generation” design meant to outlive a 15-year lease. |
| Texas Capital | Demand in the US; cost in Finland | Demand “extremely robust”; 2027 “fairly contracted”; expansions at “materially higher rates”. Finland costs similar in dollars, with a longer build. |
| Lake Street | Role of Base Electron; supply chain | Base Electron is about 2030 onward: about 1.2 GW at Center and more in the east of the state, about 2.4–2.5 GW of generation equipment procured. Key components locked two years ago, including the full factory output of some suppliers for four years. |
| Craig-Hallum | What “selective” and “premium pricing” mean | Selective on tenant credit. Pricing target “north of 15%” above current rates, possibly with longer terms; declined to “pin that” down. |
| Roth Capital | Finland timing | First 100 MW available in 2028, the full gigawatt through 2031; the site is about 30 km from a hyperscaler self-build. |
| Needham | What the 3.5–4 GW goal requires; ChronoScale | Power is available; construction is the bigger constraint. The high end implies about 2 GW by end-2028 and about 1 GW a year in 2029–30. ChronoScale could reach “multiple hundreds of megawatts” on 50–100 MW grid-tied sites in 2027–28. |
| Compass Point | Macquarie outside North America; fit-out; modular builds | The Macquarie agreement covers North America only. Fit-out is becoming larger in dollar value. Modular builds are used for ChronoScale capacity. |
| Positive surprises | Negative surprises |
|---|---|
| Building 2 completed on 1 October, three months after its first phase; the third building “progressing on schedule” | The expansion leases slipped from “advanced negotiation” in July to “by year-end” |
| Pricing on new leases guided more than 15% higher | A fifteen-year power contract with a company the chief executive runs, price terms undisclosed |
| NOI margin held at 89% | A $221.0m GAAP loss to common, with $69.2m of stock pay |
| Operating cash flow positive at $63.9m; $35.8m of interest income on the cash | No cost-per-megawatt or capital-spending figures |
What to watch next. The signing of the expansion leases, with their rent and term; the size and price of the Macquarie preferred and the project debt for the three unfinanced campuses; first rent at Polaris Forge 2 by December; first rent at Delta Forge 1 in the first half of 2027; Base Electron reaching financial close by 31 March 2027; and the quarterly NOI margin as more buildings start.
Applied Digital builds very large data centres for artificial-intelligence computing and rents them out. It finds land next to cheap, plentiful power, secures a grid connection, builds high-density, liquid-cooled buildings and leases each campus to one large tenant for fifteen years. The tenant brings the chips; Applied Digital provides the building, the power and cooling plant and the site operations, and is paid a fixed monthly rent per megawatt. The leases are take-or-pay and non-cancellable: a tenant that walked away would owe the remaining rent. The company calls these campuses AI factories.
It has two smaller businesses. Data Center Hosting rents 286 MW of power and space to bitcoin miners in North Dakota, billed on capacity rather than on bitcoin mined. ChronoScale (Nasdaq: CHRN), about 96% owned, is the former cloud-computing arm: it rents GPU capacity, resells GPU hardware and has announced a 50 MW deployment for Microsoft. It is consolidated in the accounts but excluded from the company’s adjusted figures, and we value it separately. The chairman and chief executive is Wes Cummins, a former asset manager; the company was founded as Applied Blockchain and renamed Applied Digital in 2022.
| Revenue stream | Q1 FY27 | Q1 FY26 | Economics | Durability |
|---|---|---|---|---|
| Base rent15-year take-or-pay leases | $65.8m | — | NOI $58.8m, an 89% margin. The asset being valued | Contracted |
| Tenant fit-out services | $183.5m | $26.3m* | Cost $176.1m: about a 4% margin, billed as each building is fitted out | Ends at delivery |
| Tenant recoveries | $13.3m | — | Pass-through of operating costs, no margin | Follows the lease |
| Data Center HostingBitcoin miners, Jamestown and Ellendale, ND | $37.8m | $37.9m | 286 MW; segment profit $13.3m | Billed on capacity |
| ChronoScaleAbout 96% owned, consolidated | $41.5m | $16.7m | $23.0m of GPU hardware resale; operating loss $25.4m | Listed separately |
Source: Form 10-Q and results release, 7 October 2026. *Q1 FY2026 HPC services revenue in total; the rent line began in Q2 FY2026. Fit-out margin derived from HPC segment services revenue less services cost of revenue.
Only one of these lines is the business being valued. Base rent is the long-dated, high-margin income a buyer of finished data centres would pay for. Fit-out revenue is real work, billed to the tenant as it is done, but at a 4% margin it is close to pass-through, and it stops when a building is handed over. It was 54% of revenue in the quarter; a reader comparing revenue growth of 322% with a rent line that has not yet had a full quarter of 250 MW should keep the two apart.
| Campus | Location | Tenant | MW | Base-term rent | Rent, $m per MW-yr | Rent starts (company) |
|---|---|---|---|---|---|---|
| Polaris Forge 1 | Ellendale, ND | CoreWeave | 400 | ≈$11.0bn | 1.80 | B1 Oct 2025; B2 Jul and Oct 2026 (250 MW live); B3 in 2027 |
| Polaris Forge 2 | Harwood, ND | IG hyperscaler | 200 | ≈$5.0bn | 1.64 | Initial 2H 2026; full early 2027 |
| Polaris Forge 3 | Center, ND | Tier-one IG hyperscaler | 300 | ≈$7.5bn | 1.64 | 2H 2027 to 2H 2028 |
| Delta Forge 1 | Louisiana | Same hyperscaler | 300 | ≈$7.5bn | 1.64 | 1H 2027 to early 2028 |
| Delta Forge 2 | Alabama | Same hyperscaler | 210 | ≈$5.2bn | 1.62 | 1H 2028 |
| Total | 1,410 | ≈$36bn | 1.68 | Minimum payments $2,371.4m a year from FY2030 |
Sources: FY2026 Form 10-K (campus table), Delta Forge 2 release (8 June 2026), Form 10-Q. Rent per MW is derived: contract total ÷ 15 years ÷ MW, scaled by 0.983 so the five campuses add up to the full-year minimum payment in the 10-Q. Three of the leases are with one tier-one investment-grade hyperscaler, which the company does not name.
The 10-Q’s schedule of minimum lease payments is the cleanest statement of what is signed: $468.0m for the nine months to May 2027, $1,579.5m in fiscal 2028, $2,360.6m in fiscal 2029, then $2,371.4m in each of fiscal 2030 and 2031 and $26.6bn thereafter, $35.7bn in all. Two things follow. The ramp is steep, so almost all of the rent roll is still to come: the quarter’s base rent of $65.8m is about 11% of the eventual run-rate. And the fiscal 2030 and 2031 figures are the same to the thousand dollars, so the fixed rent does not escalate. Any inflation-linked increases are excluded from the table as variable rent, and the company does not disclose whether there are any. We model contracted rent as flat.
The binding constraint on AI data-centre supply is electricity and grid connections, not capital or land. The company says the 1,410 MW portfolio is backed by about 2.15 GW of grid-connected utility power. North Dakota offers cheap power, a cold climate that reduces cooling load, and little competition for construction labour; the company has added two southern campuses in Louisiana and Alabama for the same hyperscaler. For growth after 2029 it relies partly on Base Electron, an independent power producer developing a 1,200 MW gas-fired plant next to Polaris Forge 3, in which it owns about 10% and from which it agreed on 4 October to buy power for fifteen years. Base Electron is owned and run in part by Applied Digital’s own chief executive, president and a director (Section 07).
The closest listed comparisons are the former bitcoin miners that now lease AI capacity, TeraWulf and Core Scientific among them, and the tenant CoreWeave, which owns the chips and rents out compute. IREN, which owns its GPUs and sells cloud capacity, is the contrast. Applied Digital is the purest landlord of the group: it owns no GPUs in its main business, and its hyperscaler tenants carry the hardware and utilisation risk. Section 03 compares the valuations.
| $ millions | Q1 FY27 | Q1 FY26 | Q4 FY26 |
|---|---|---|---|
| Total revenue | 341.9 | 80.9 | 258.7 |
| Adjusted revenue, excluding ChronoScale | 300.4 | 64.2 | 240.4 |
| Base rent | 65.8 | — | 44.1 |
| Net operating income | 58.8 | — | 39.9 |
| Adjusted EBITDA | 64.4 | 0.5 | 42.4 |
| Stock-based compensation | 69.2 | 15.5 | 127.8 |
| Operating loss, GAAP | (62.4) | (9.8) | (124.8) |
| Net loss to common, continuing operations | (221.0) | (18.5) | n/a |
| Cash from operations | 63.9 | (81.5) | n/a |
| Capital expenditure | (2,074.7) | (249.9) | n/a |
| Interest expensed / capitalised / paid in cash | 77.4 / 35.7 / 242.9 | 8.0 / 8.7 / 9.0 | n/a |
Sources: Q1 FY2027 Form 10-Q and release (7 October 2026); Q4 FY2026 release (27 July 2026). NOI is the company’s measure on base rent. Adjusted figures exclude ChronoScale.
The balance sheet at 31 August held $3,677.9m of cash and restricted cash, $3.4bn of it inside APLD HPC TopCo 2, the Macquarie-funded vehicle that owns Polaris Forge 1 and 2. Debt at face was $6,721.6m: $2,350m of 9.25% notes due 2030, $2,150m of 6.75% notes and $1,590m of 7.00% notes due 2031, all secured on Polaris Forge campuses; $450m of 2.75% convertible notes; $82.4m drawn on the revolver and $99.2m of other loans. Macquarie’s preferred stake in TopCo 2 was carried at $2,010.9m. Common shares outstanding were 302.4m on 6 October, up from 287.9m at 31 May, mainly from Series G preferred stock converted into common.
Disclosed drivers: more than 600 MW in service over the next twelve months, and everything leased paying rent during 2028; about 250 MW of expansion leases expected by the end of the calendar year at pricing “north of 15%” above current rates; new campuses toward a 3.5–4 GW portfolio by the end of 2030 if demand holds; and cheaper capital, with secured coupons down from 9.25% in November 2025 to 6.75% and 7.00% in 2026, investment-grade structures in view, and CoreWeave’s agreement to move the Building 3 lease to a subsidiary if that subsidiary is rated investment grade. Outside the model: up to 1 GW of power in Finland from 2028, the Base Electron stake and ChronoScale. Assumed in our base case: the 1,410 MW delivered about on the company’s dates and the 250 MW of expansions; nothing beyond.
Risks, in order of weight: the capital stack, with about $15bn of debt and $4.5bn of 12.75% preferred ahead of the shares at stabilisation, so that a 0.75-point move in the exit cap rate moves the base case by $6–7; tenant concentration, with three tenants and one customer 56% of the quarter’s revenue; execution, with 1,160 MW to deliver across four campuses by late 2028, two of them in new states; funding, with three campuses still to finance; dilution and pay, with $69.2m of stock pay in the quarter and a $2.0bn preferred equity line that converts into common; and related parties, with Base Electron owned in part by the chief executive, the president and a director. Section 07 scores each.
| Source | Score / 5 | Assessment |
|---|---|---|
| Power and interconnection | 4 | About 2.15 GW of grid power behind 1.4 GW; utility relationships in North Dakota; Base Electron for 2030 onward. Others are buying power too |
| Switching costs | 4 | Tenants install billions of dollars of GPUs in a liquid-cooled hall; moving is costly and slow — until the leases expire, from 2040 |
| Contracts | 5 | Fifteen-year take-or-pay, non-cancellable, mostly investment-grade; two counterparties dominate |
| Cost position | 3 | Cheap power and a cold climate; build cost in the industry range of $11–13m per MW, with no disclosed advantage |
| Execution and scale | 4 | On-time deliveries at Polaris Forge 1; a repeatable design across five campuses; four campuses at once from 2027 |
| Cost of capital | 2 | Falling, but still a sub-investment-grade borrower using 12.75% preferred; REITs and hyperscalers fund far cheaper |
| Brand, network effects, data, IP | 1 | Not relevant to a landlord |
| Regulation and community | 3 | Community support, tax base, grid credits for local ratepayers; permitting risk in new states |
Strongest advantages. A fifteen-year take-or-pay lease with an investment-grade tenant is a bond-like stream a competitor cannot take away, and the grid connections behind it take years to replicate. Execution is the third leg: the record at Polaris Forge 1 is the evidence a hyperscaler can check, and one has chosen Applied Digital repeatedly — the same tier-one hyperscaler signed three campuses in about seven weeks between April and June 2026.
Biggest threats. Cost of capital is the weak flank. Hyperscalers can build for themselves, and one is self-building about 1 GW in North Dakota, by the chief executive’s account. REITs such as Digital Realty and Equinix fund at investment-grade rates. A landlord that funds with 7% project debt and 12.75% preferred keeps less of the spread between rent and cost than those competitors. The second threat is technology: the leases run to the early 2040s, and the residual value of a purpose-built AI hall after that is unknown.
Direction: stable to widening. Each new lease with an investment-grade tenant and each step down in borrowing cost strengthens the position; the expansion leases at higher rents, if signed, would be the first direct evidence of pricing power.
| Concern | Severity | Evidence and why it matters |
|---|---|---|
| Capital stack: Macquarie preferred | HIGH | Macquarie funded $1.825bn into TopCo 2 (Polaris Forge 1 and 2) at 12.75% a year, compounding twice a year and paid in kind; a minimum return of 1.80× its money, counting its common units; 13.5% of TopCo 2’s common equity; rate step-ups from October 2030 to as much as 16.75%; a forced sale process if still outstanding in October 2032. The same structure is planned for the next three campuses |
| Related party: Base Electron | HIGH | Owned and managed in part by the chief executive (its CEO), the president (its president and secretary) and a director. Applied Digital owns about 10%, has lent it $58.5m on demand (facility up to $100m) and on 4 October signed a fifteen-year contract to buy its power with fixed monthly capacity payments, plus a credit support agreement. Price terms undisclosed; approved by a board related-party committee |
| Dilution | HIGH | Shares outstanding rose from 287.9m (31 May) to 302.4m (6 October). The Series G line, now $2.0bn and open to August 2029, sells preferred stock that converts into common: $815m in fiscal 2026 became 51.0m shares. A further 28.1m warrants, 46.1m convertible-note shares (about 38m after the capped call) and 15.2m unvested awards sit behind them |
| Tenant concentration | HIGH | One customer was 56% of the quarter’s revenue. One hyperscaler holds 810 MW; CoreWeave, a leveraged GPU cloud, holds 400 MW |
| Pay | MED-HIGH | Chief executive total compensation of $189.2m in fiscal 2026, almost all stock: 4.5m performance units vest at average share prices of $50, $75 and $100. Stock pay was $69.2m in the quarter, 107% of adjusted EBITDA |
| Revenue quality and non-GAAP | MEDIUM | Fit-out at about a 4% margin is 54% of revenue. Adjusted revenue excludes ChronoScale. “Other non-recurring expenses” are $1.3m in one adjusted reconciliation and $2.6m in the other |
| Cash flow | MEDIUM | Operating cash flow of $63.9m against capital spending of $2,074.7m. $242.9m of interest paid in cash in the quarter, against $77.4m expensed and $35.7m capitalised |
| Pledged shares | MEDIUM | The proxy shows 17.6m of the chief executive’s 23.4m shares held by Cummins Family Ltd and pledged |
| Litigation | LOW | A 2023 securities class action is stayed pending a motion to dismiss; a derivative suit was dismissed without prejudice |
Sources: FY2026 Form 10-K (Macquarie terms, Series G), Q1 FY2027 Form 10-Q (related parties, debt, shares, customers), Q1 FY2027 release (reconciliations), 2026 proxy statement (pay, pledge), Forms 4 to 19 October 2026.
Overall red-flag score: 6 out of 10, elevated. The concerns are an expensive preferred investor that ranks ahead of the common at the campus level, an equity line that converts into common shares continuously, pay set at levels that assume the share price quadruples, and a power contract with a company run by the same executives. Each is disclosed. None is cheap for the common shareholder.
| Strength | Score / 10 | Evidence and why it matters |
|---|---|---|
| Delivery record | 8 | Building 1 (October 2025) and the first phase of Building 2 (1 July 2026) delivered on schedule, by the company’s account; the second phase followed on 1 October. 250 MW of Polaris Forge 1’s 400 MW is live |
| Lease quality | 8 | Fifteen-year, take-or-pay, non-cancellable leases; 1,010 MW with investment-grade hyperscalers. Termination for convenience would require the full remaining rent |
| Cost of capital falling | 7 | Secured coupons from 9.25% to 6.75% and 7.00%; investment-grade structures in view; CoreWeave’s agreement to move the Building 3 lease to an investment-grade subsidiary if one is rated |
| Insider buying | 6 | On 16 October, after the shares fell 22% in four sessions, the chief executive bought 55,803 shares at $18.04 and the president 300,000 at $18.86 in the open market (Forms 4). Earlier sales were a director’s: 75,000 shares in August and 31,250 on 13 October |
| Power and supply chain | 6 | About 2.15 GW of grid-connected power behind the contracted portfolio; Base Electron generation from 2030; key equipment reserved for four years |
| Margins | 6 | NOI margins of 89–91% on base rent in the first quarters of operation |
| Pricing power | 6 | New leases guided “north of 15%” above current rents, possibly longer. Not yet signed |
| Optionality at little or no value | 6 | Nothing in the base case for the 3.5–4 GW ambition, Finland, the renewal options (the company puts contracted revenue at $86bn if all are exercised) or a conversion of the 180 MW Ellendale hosting site, and only half of ChronoScale’s market value |
Overall green-flag score: 7 out of 10, strong. The strongest fact in the company’s favour is falsifiable and has held so far: every delivery the company has dated has arrived on time. Leases of this length with tenants of this quality are the asset infrastructure buyers pay low yields for, and the company has more power and more land than its signed leases use.
| Dimension | Score / 5 | Assessment |
|---|---|---|
| CEO track record | 4 | Wes Cummins turned a bitcoin-hosting company into a 1.4 GW hyperscale landlord in about two years; on-time deliveries at Polaris Forge 1 |
| CFO credibility | 4 | Saidal Mohmand: secured coupons from 9.25% to 6.75–7.00%; revolver and bridge refinanced; clear on financing plans |
| Guidance accuracy | 3 | Delivery dates met; the July expansion-lease timing slipped to year-end; the $1bn NOI run-rate goal not restated |
| Transparency | 2 | No build cost per MW or capex outlook; Base Electron price terms undisclosed; adjusted metrics exclude ChronoScale |
| Capital allocation | 3 | Sound site and tenant choices; expensive preferred equity; Finland and ChronoScale add scope |
| Dilution | 2 | Shares up 5% in four months; a $2.0bn convertible preferred line open to 2029 |
| Insider ownership | 4 | Officers and directors own 9.5%, the chief executive 7.8%, and the chief executive and the president bought shares on 16 October; most of the chief executive’s stake is pledged |
| Compensation | 1 | Chief executive pay of $189.2m in fiscal 2026; stock pay larger than adjusted EBITDA in the quarter |
| Board quality and related parties | 2 | The chief executive, the president and a director are owners and officers of a power supplier the company lends to and buys from |
| Communication | 3 | Consistent, measured; specifics withheld on the costs that matter most |
| Average | 2.8 | Strong operators; weaker stewards of the common shareholder’s claim |
Sources: FY2026 Form 10-K, Q1 FY2027 Form 10-Q, 2026 proxy statement (25 September 2026), Forms 4, call transcripts.
Do they act like long-term owners? Of the campuses, yes; of the common stock, only lately. Operationally this is a strong team: it chose its sites and tenants well and has delivered what it promised on the ground. As stewards of the common shareholder’s claim it is weaker. Pay is set at levels that only make sense if the share price quadruples, the equity line converts into common shares continuously, the costs investors most need to model are not disclosed, and the Base Electron arrangement puts the chief executive on both sides of a fifteen-year power contract. The board approved it through a related-party committee, as it should, but the terms are not public. The chief executive’s and the president’s open-market purchases on 16 October, about $6.7m between them, are the best evidence yet that they think of the common stock the way they think of the campuses.
| Window | Event | What it decides | Impact · source |
|---|---|---|---|
| 27–28 Oct 2026 | FOMC meeting | Rates set the exit yield and the cost of refinancing | Medium · Federal Reserve calendar |
| Late Oct – early Nov 2026 | Hyperscaler third-quarter results and capex guidance | The demand signal for the expansion leases and new campuses | High · company calendars |
| 4 Nov 2026 | Annual meeting | Votes on pay and directors after fiscal 2026 chief executive pay of $189.2m | Low · 2026 proxy |
| Nov 2026 | CoreWeave third-quarter results | Tenant credit for 400 MW | Medium · expected |
| By 31 Dec 2026 | About 250 MW of expansion leases | Worth about $6 a share in our base case; the rent is the test | High · CEO, 7 Oct call |
| By 31 Dec 2026 | Polaris Forge 2 initial operations | About 300 MW live in North Dakota | Medium · Q1 FY2027 release |
| Next 3–6 months | Financing for Polaris Forge 3 and Delta Forge 1 and 2 | Size and price of the Macquarie preferred and the project debt | High · CFO, 7 Oct call |
| 8–9 Dec 2026 | FOMC meeting | As above | Medium · Federal Reserve calendar |
| Early–mid Jan 2027 | Q2 FY2027 results | Expansion leases, Polaris Forge 2 rent, financing terms, NOI margin | High · timing typical |
| 1H 2027 | Delta Forge 1 first rent | The first building outside North Dakota | High · Form 10-Q |
| 31 Mar 2027 | Base Electron financial-close deadline | Base Electron may end the power contract if not reached | Low · Form 10-Q |
| Apr 2027 | Q3 FY2027 results | Delta Forge 1 delivery; capex run-rate | Medium · timing typical |
| 2027 | Polaris Forge 1 Building 3, 150 MW | Completes Polaris Forge 1; a possible move of the lease to an investment-grade CoreWeave subsidiary | Medium · 10-Q; June 2026 MoU |
| Jul 2027 | Q4 FY2027 results | More than 600 MW placed in service over the year, against the target | High · CEO |
| 2H 2027 | Polaris Forge 3 first rent | Second phase in 2H 2028 | Medium · Form 10-Q |
| From Apr 2028 | Macquarie redemption window opens | Refinancing the 12.75% preferred | Medium · FY2026 10-K |
Dated where the company or a counterparty has given a date; “timing typical” where we infer it from the company’s reporting pattern. Applied Digital’s fiscal year ends on 31 May. Upside in each case is the event arriving as described or better; downside is a delay, a weaker rent or financing term, or tenant stress. The single most informative item is the expansion-lease signing, because it tests pricing power and is already in our base case. Impact reflects how much the event can move the value, not the direction of its outcome.
BULLGrowth. More than 600 MW goes live in the next twelve months, more than doubling capacity, and every megawatt is already leased. Rent goes from $66m a quarter to $2.4bn a year in under three years. Management has more power than leases and says 3.5–4 GW is achievable by 2030; even 750 MW of that is worth $16.79 a share.
BEARThe growth is contracted, so it is also already known. The only new growth on offer, the 250 MW of expansions, has been “imminent” since July. Beyond it, 3.5–4 GW is an ambition, and the chief executive himself says construction is the constraint.
BULLValuation. Fifteen-year leases to investment-grade tenants are what REITs and infrastructure funds, with a lower cost of capital, buy at the lowest yields. At 6.5% the base case is $31.01; at 6.25%, $33.50. The market is pricing the shares at an 8.2% yield, as if the tenants were weak.
BEARThe 30-year Treasury yields about 5.3%. Flat rent with no escalator does not deserve a premium yield, and the framework ignores the corporate tax a sale would trigger. Deduct it and drop the unsigned expansions and the value is about $14–16, below the price. Half of ChronoScale’s paper value is in the number too.
BULLBusiness quality and risks. On-time delivery at Polaris Forge 1, 89–91% NOI margins, take-or-pay leases. CoreWeave has agreed to move its Building 3 lease to a subsidiary if that subsidiary is rated investment grade, and 810 MW are with a tier-one hyperscaler.
BEARThree tenants, one of them a leveraged GPU cloud. Four campuses must be delivered at once from 2027, two in states where the company has never built. A three-to-nine-month slip, a 15% overrun and no expansions take the value to $10.07.
BULLFinancials and management. Borrowing costs have fallen by more than two points in a year, and investment-grade structures are next. Cash flow from operations has turned positive. And the chief executive and the president have just bought shares in the open market.
BEARMacquarie is paid 12.75%, compounding, with a share of the equity on top, and the same terms are planned for the next three campuses. Pay, dilution and a related-party power contract all point value away from the common shares.
BULLCatalysts. Signed expansion leases at rents 15% or more higher would confirm pricing power within weeks; investment-grade financing would lower the equity’s risk.
BEAREach new campus financing adds a claim ahead of the shares, and the Series G line can sell stock into any rally.
The bull has the better of the argument on the asset, and at this price on the valuation as well: the leases, the tenants and the delivery record are as good as this sector offers, and $18.00 values the finished campuses at an 8.2% yield. The bear has the better of it on the claim: a great deal of the value goes to lenders, to Macquarie and to the cost of running a listed company, and the framework’s untaxed exit flatters the common. But to justify the price the bear needs two things at once, no expansion leases and a taxed exit. Undervalued is the evidence-based conclusion, with that dependence stated. Most uncertain: the yield a buyer would pay for flat rent, and the terms of the next Macquarie money. What would change it: the expansion leases failing to appear, or signing at flat rents; financing for the next three campuses on terms worse than Macquarie’s current ones; or a delay at Delta Forge 1. Verify next: the expansion terms, the Macquarie terms for the new campuses, and the cost per megawatt.
Applied Digital is a landlord for artificial-intelligence computers. Big technology companies need huge buildings full of powerful chips, with enormous amounts of electricity and cooling. Applied Digital finds land where power is cheap and plentiful, mostly in North Dakota, builds those buildings and rents each one to a single large tenant for fifteen years. The tenant brings and owns the chips.
Rent. Each tenant pays a fixed amount every month for every megawatt of power its space can use, whether it uses it or not. After the costs of running the building, about nine-tenths of the rent is left over. The company also earns a small margin fitting out the inside of buildings for tenants, rents space to bitcoin miners, and owns most of a separately listed cloud company, ChronoScale.
Almost all of its space is already rented, to some of the largest and most creditworthy companies in the world, and the buildings are opening on time. Rent should rise about ninefold from the latest quarter’s level as the leased buildings open over the next two years.
The buildings cost about $12m per megawatt, so the company borrows most of the money and has sold an expensive share of its projects to Macquarie. Lenders and Macquarie get paid before shareholders. Of every dollar the finished buildings are worth, only about 30 cents is left for the shareholders, so a small change in what the buildings are worth makes a big difference to the shares.
We think the shares are worth about $21 to $34 each, with $26.57 the most likely value, so at $18.00 we call them undervalued. That depends on the company signing the new leases it expects by December, or on valuing the buildings the way a tax-exempt owner would. Without both, the value is a little below the price.
Right: the new leases are signed at higher rents, the next buildings are financed more cheaply, and everything keeps opening on time.
Wrong: a delay, a cost overrun, higher interest rates or a tenant in trouble would hit the shareholders’ slice hardest. Executive pay is very high, and the company keeps issuing new shares.
Whether the new leases are signed by the end of December, and at what rent. How the next three campuses are financed, over the next three to six months. And the results in January, which will show the second North Dakota campus starting to pay rent.
| Status | What it covers |
|---|---|
| Verified from Applied Digital filings and releases | Revenue by stream and segment, NOI and its margin, adjusted EBITDA, stock-based compensation, losses, cash flow, capital spending, interest, cash, debt and its terms, the Macquarie preferred and its terms, the Series G line, the convertible notes, warrants and awards, the campus table and the schedule of minimum lease payments, customer concentration, related-party arrangements with Base Electron, executive pay and the pledge — from the Q1 FY2027 results release and Form 10-Q (7 October 2026), the FY2026 Form 10-K (29 July 2026), the Q4 FY2026 release (27 July 2026), the 2026 proxy statement (25 September 2026), Forms 8-K of 9 June 2026, the convertible-note pricing release (31 October 2024) and Forms 4 filed to 19 October 2026, including the purchases by the chief executive and the president on 16 October. |
| Verified from market data and peer filings | Applied Digital at the last Nasdaq pre-market trade at 9:04 a.m. Eastern on 26 October 2026; TeraWulf and Core Scientific at their last pre-market trades within two minutes of it; ChronoScale and Babcock & Wilcox at their 23 October 2026 closes. Peer share counts, debt and cash from the TeraWulf and Core Scientific Forms 10-Q for the quarter to 30 June 2026; ChronoScale’s share count from its Form 10-K (19 August 2026). Treasury yields and the federal funds target for late October 2026 from FRED data. |
| Third-party, attributed | Call remarks and analysts’ questions from a published transcript of the 7 October 2026 call (Benzinga), checked against the Form 10-Q for every figure. |
| Our estimate or judgement | Rent per MW by campus, aligned to the minimum payments; the base-case delivery dates; remaining build cost; the expansion rent; NOI margins; corporate costs and stock pay; the month-by-month funding run, new project debt at 80% of cost and new Macquarie preferred on its TopCo 2 terms; the bitcoin-hosting multiple; ChronoScale at half its market value; the sale-tax and unlevered cross-checks; the 7.0% exit cap rate and 14.0% equity rate; the 30 / 45 / 20 / 5 weights; the grid, the reverse valuation, the $20.68–34.02 band and the $26.57 central estimate; and all scores in Sections 06, 07 and 08. The model on this page is the model that produced those figures — the scenario cards load the published cases, and every cell of the cap-rate grid is reproducible from the controls. |
| Still open | The rent and term of the expansion leases; the size and price of the financing for the three unfinanced campuses; build cost per megawatt and the remaining budget; the price terms of the Base Electron power contract; whether any of the leases carries inflation-linked rent. |
Method. Stabilised net operating income, capitalised and de-levered, for contracted asset-backed infrastructure, methodology version 1.3. The portfolio’s NOI on the date the last base-case lease starts paying rent is capitalised at a 7.0% exit cap rate, less corporate cash costs and stock pay capitalised at the same rate; every claim ranking ahead of the common at that date is deducted, after a month-by-month funding run from the 31 August 2026 balance sheet; the residual is discounted to 26 October 2026 at a 14.0% equity rate. One exit cap rate and one equity rate are used across all scenarios, and each material risk is priced once. Stock-based compensation is charged as a cost and unvested awards are excluded; warrants are exercised for cash; the convertible notes convert at the implied value net of the capped call. Scenario weights are our judgement, the fair-value band is the corners of the exit cap rate 6.25–7.75% by equity rate 13–15% grid, and the reverse valuation is run against the market price. The rating changes to fairly valued if the price rises to $20.68 or above with no change in the evidence, and to overvalued above $34.02; the conclusion is most sensitive to the exit cap rate, the expansion leases and the tax on an exit, in that order. Ratings are undervalued, fairly valued or overvalued according to where the price sits relative to the band; Sonde does not issue buy, sell or hold recommendations or price targets. The reference price used throughout is $18.00, the last Nasdaq pre-market trade at 9:04 a.m. Eastern on 26 October 2026, and every multiple in this report scales directly with it. Ratings are reviewed each quarter on results and immediately on a material lease, a financing event or a share-price move of more than 25%.