Applied Digital Stock After FY2026: Revenue Exploded 407% — but the $44 Million Base-Rent Number Matters More

Two months ago, Applied Digital was mostly a promise with a construction schedule attached. The leases were enormous, the market narrative was even larger, and the stock traded as if every megawatt on a presentation slide were already humming inside a finished AI factory.

Now the accounting has begun to catch up with the architecture.

Applied Digital’s fiscal fourth quarter revenue jumped 407% year over year to $258.7 million. Full-year revenue reached $611.3 million, up 167%. The first 100 MW building at Polaris Forge 1 was operating during the quarter, a second 75 MW phase went live on June 30, and the company ended fiscal 2026 with roughly 1.41 GW of contracted critical IT load across five campuses.

Those numbers sound like the clean confirmation bulls were waiting for. They are not. At least not yet.

The most important number in Applied Digital’s quarter was not $258.7 million of revenue. It was the $44.1 million of base rent inside the HPC Hosting segment. Another $152.4 million came from tenant fit-out services. Fit-out revenue is real revenue, but it is economically different from long-duration rent. It carries costs, it does not represent the recurring annuity that ultimately makes the AI Factory thesis valuable, and it can make top-line growth look far more mature than the underlying rental stream actually is.

That distinction is the reason this update is different from our earlier Applied Digital deep dive. The debate is no longer whether Applied Digital can sign hyperscaler leases. It has. The new question is harder: how efficiently can the company turn contracted megawatts into recurring rent, NOI and ultimately cash attributable to common shareholders?

Data cut: August 31, 2026. Market-price references use late-August closes rather than an intraday quote.

The headline growth is real — but it mixes two different businesses

Applied Digital reported $258.7 million of fiscal Q4 revenue, compared with $51.1 million a year earlier. On the surface, that is the kind of growth rate usually associated with software, not concrete, substations and cooling systems.

But the composition matters.

• $203.0 million came from HPC Hosting.
• Within HPC Hosting, $44.1 million was base rent.
• $152.4 million came from tenant fit-out services.
• $6.5 million came from tenant recoveries.
• The legacy Data Center Hosting business generated $37.3 million of quarterly revenue.

The company’s own filing shows why investors should separate these categories. Tenant fit-out services produced $152.4 million of Q4 revenue but also generated roughly $145.6 million of associated service costs. The base-rental economics look very different. For fiscal 2026, Applied Digital reported $99.8 million of HPC base rental revenue and $90.4 million of Net Operating Income, a company-defined non-GAAP measure that subtracts property operating expenses, property taxes and insurance from base rent.

That is a reported NOI margin of roughly 91% on the rental stream.

This does not mean 91 cents of every rental dollar will become free cash flow to equity. NOI excludes corporate costs, depreciation, financing costs and the mountain of capital required to build the campuses. But it does show why the rental line matters so much more than the spectacular fit-out growth.

The point of the chart is not that fit-out revenue is unimportant. It is that investors should not value pass-through construction activity like stabilized rent.

Why $44.1 million matters more than 407%

The recurring-data-center thesis lives in base rent. Q4’s $44.1 million was generated before the second 75 MW phase at Polaris Forge 1 reached ready-for-service on June 30. That second phase raised live capacity at the campus to 175 MW.

If the first 100 MW building can generate meaningful base rent and very high property-level NOI margins, then every additional stabilized building has the potential to change the earnings profile dramatically. But there is a lag between four verbs that the market often compresses into one:

sign, finance, build, operate.

Applied Digital has become very good at the first verb. It has also demonstrated access to capital. The next phase of the story is whether execution remains as reliable when five campuses are being developed across multiple states at once.

The company ended fiscal 2026 with approximately 1,410 MW of contracted critical IT load and about $36 billion of contracted revenue over the initial 15-year lease terms, or about $86 billion if every renewal option is exercised. Mathematically, $36 billion spread across 15 years is roughly $2.4 billion of average annual contracted gross revenue once the portfolio is fully delivered and operating — though actual commencement dates and ramps differ by campus.

That is the enormous prize. But today only a fraction of those megawatts are live.

The operational progress since our June article is meaningful

The earlier Applied Digital article was written while the market was still trying to decide whether the company’s hyperscaler contracts justified the excitement. Since then, three facts have become harder to dismiss.

1. Polaris Forge 1 has moved from promise to production

The first 100 MW data center became operational in October 2025. Applied Digital then delivered Phase 1 of Building 2, another 75 MW, on June 30, 2026, bringing total live capacity at Polaris Forge 1 to 175 MW.

2. The contracted book has reached industrial scale

Across five campuses, Applied Digital now has roughly 1.41 GW of contracted critical IT load. Three of the newer leases — Delta Forge 1, Polaris Forge 3 and Delta Forge 2 — are with the same high investment-grade hyperscaler and represent about $20 billion of base-term contracted revenue.

3. The company has built a financing machine around the construction machine

During fiscal 2026 and shortly afterward, Applied Digital completed billions of dollars of project financing, including $2.15 billion of 6.75% senior secured notes for Polaris Forge 2 and $1.59 billion of 7.00% senior secured notes for the fourth building at Polaris Forge 1. Its revolving credit facility was also increased to $430 million of commitments, with a remaining accordion option.

That is not financial trivia. It is the business model.

The balance sheet is both the moat and the danger

At May 31, Applied Digital reported $4.2 billion of cash, cash equivalents and restricted cash against $5.0 billion of debt. A simplistic investor might call that only $800 million of net debt. That would be misleading because much of the cash is restricted or linked to project financing. It is not a general corporate piggy bank that can be freely netted against every obligation.

The better way to think about the capital structure is project by project. Long-term take-or-pay leases can support asset-level debt. If a project is delivered on time and the tenant pays for 15 years, the debt can be highly rational. It allows Applied Digital to transform a signed lease into an operating asset without funding the entire build with common equity.

The danger appears when the schedule slips. Interest does not care that construction is late. Debt amortization does not applaud a beautiful investor deck. A small timing problem at one campus can become expensive; simultaneous delays across several campuses could change the equity story very quickly.

This is why Applied Digital belongs in the same analytical family as our recent pieces on Nebius and AI-cloud financing and IREN’s transition from mining to AI infrastructure. In all three cases, the technology narrative is inseparable from the capital structure.

The accounting is still messy enough to punish lazy valuation

Applied Digital reported a Q4 net loss attributable to common stockholders of $110.6 million and a full-year loss of $249.2 million. Meanwhile, adjusted net income for the year was positive at $36.1 million, adjusted EBITDA reached $107.2 million, and NOI was $90.4 million.

Those figures can all be true at once because the company is going through a complicated transition. It separated its Cloud Services business into ChronoScale, still owns roughly 96% of that public entity, recorded very large stock-based compensation, and has derivative and investment fair-value movements flowing through the statements.

Stock-based compensation alone was approximately $219 million in fiscal 2026, according to the annual filing. That is not an item common shareholders should simply wave away because it is non-cash in the current period. Compensation paid in stock is compensation paid with ownership.

The cleanest operating metric is therefore increasingly the property-level rent and NOI from stabilized AI hosting assets. The messier consolidated income statement will matter, but it is not yet the best single lens through which to judge campus economics.

A new valuation lens: the stock is cheaper, but the denominator has changed

Applied Digital traded around $27 to $28 in late August, compared with roughly $46 when our earlier article was written in June. Using approximately 291 million shares outstanding, that places the equity value around $8 billion rather than the roughly $13 billion level embedded in the earlier analysis.

The business has not become smaller during those two months. Live capacity increased. Full-year results provided the first clean look at rental economics. Contracted capacity remains around 1.4 GW.

So is the stock suddenly cheap?

Possibly. But only if we compare the price with the right future earnings stream.

One useful bridge starts with the lease portfolio. The roughly $36 billion of base-term contracted revenue, divided by 15 years, equates to about $2.4 billion of average annual gross contracted revenue when fully online. If mature property-level NOI margins ultimately settle somewhere between 75% and the current reported 91%, the portfolio could theoretically support roughly $1.8 billion to $2.2 billion of property-level NOI before corporate costs, financing and other items.

At an $8 billion market cap, that sounds astonishingly cheap. And it is exactly where investors can fool themselves.

The market cap is equity value today. The future NOI requires billions more of construction capital, years of execution, debt service, tenant performance and no catastrophic loss of demand. The correct denominator is not “future NOI if everything is built.” It is the present value of the future cash that belongs to common shareholders after the capital required to create that NOI.

The two mature-NOI bars are scenarios derived from approximately $2.4 billion of average annual base-term contracted revenue. They are not company guidance.

Three valuation scenarios at the current share count

I would not use current-year P/E for Applied Digital. The company is still building the assets that define the eventual earning power. A scenario approach around stabilized NOI is more informative, provided the assumptions are stated openly.

These are not price targets. They are a way to make the debate explicit. The bear case assumes the campus portfolio produces much less NOI than the gross lease book suggests and receives a lower infrastructure multiple. The base case assumes Applied Digital becomes a genuine scaled AI-infrastructure landlord. The bull case assumes both execution and capital markets remain unusually favorable.

The biggest weakness in this table is the shorthand net-debt adjustment. Applied Digital’s debt is heavily project-related and the cash balance includes restricted cash, so a single corporate net-debt number oversimplifies the structure. That is intentional: the table is a sensitivity map, not an appraisal.

The hidden catalyst: every successful handoff lowers the risk premium

Applied Digital does not need another spectacular press release as badly as it needs boring construction updates.

The most valuable announcement over the next year may be one that says a building was delivered on schedule, power was energized, rent commenced and the financing remained inside budget. Each successful handoff reduces the probability that the five-campus franchise model is simply a collection of individually heroic projects.

This is the same reason Nvidia’s margins matter more than another AI slogan; our latest Nvidia analysis focuses on what happens when an extraordinary narrative finally has to be measured against economics. Applied Digital is entering that phase now.

ChronoScale: valuable option or source of noise?

The cloud-services separation adds another layer. Applied Digital owned approximately 96% of ChronoScale at the end of fiscal 2026. That stake can have real value, but it makes consolidated reporting less intuitive. The company excludes ChronoScale from certain non-GAAP measures because management views Data Center Hosting and HPC Hosting as the core operations going forward.

For investors trying to value APLD as an AI data-center platform, that is sensible. But the ownership stake remains part of shareholder value and part of consolidated accounting. If ChronoScale succeeds, APLD shareholders retain substantial indirect exposure. If it needs more capital or performs poorly, the stake can become a distraction.

I would therefore treat ChronoScale as an option on top of the core data-center thesis, not as the reason to own Applied Digital.

The bull case is stronger than it was in June

The bull case has improved in three concrete ways.

• Execution evidence: 175 MW of Polaris Forge 1 capacity is now live.
• Economic evidence: the first meaningful base-rent stream produced a reported 91% property-level NOI margin for fiscal 2026.
• Financing evidence: the company has repeatedly funded large projects in the debt markets despite their scale.

If that pattern repeats across Polaris Forge 2, Polaris Forge 3 and the Delta Forge campuses, the current $8 billion-ish equity value could look modest relative to the stabilized asset base.

The bear case has also become clearer

Ironically, more disclosure makes the risks easier to see.

Fit-out revenue can flatter headline growth

Q4’s 407% revenue growth is not the same thing as 407% growth in recurring rent. Investors who value every revenue dollar equally will overstate maturity.

The construction program is enormous

Applied Digital says another 1.7 GW of capacity is being marketed beyond the 1.4 GW already contracted. Ambition can be a competitive advantage, but development pipelines consume management attention and capital long before they produce rent.

Debt costs are real

Recent secured notes carry coupons of 6.75% and 7.00%. Those rates are manageable if assets stabilize on time, but they create a high hurdle for projects that slip or underperform.

Customer concentration remains material

Three major new leases are with the same investment-grade hyperscaler. High credit quality reduces default risk; it does not eliminate concentration risk.

The AI capacity cycle can still overshoot

Every infrastructure shortage eventually attracts capital. The question is whether AI demand grows faster than supply for long enough to protect economics across the entire development period.

What I would watch next

For the next several quarters, I would ignore the loudest headline and track six quieter numbers:

1. Live MW: how quickly contracted capacity turns operational.
2. Base rental revenue: the recurring core of the HPC thesis.
3. NOI and NOI margin: the cleanest current view of property economics.
4. Construction cost per MW: whether the franchise model actually produces repeatable economics.
5. Project debt and interest burden: the price paid to create the rental stream.
6. Common-share count: whether shareholders keep the upside or finance it through dilution.

Verdict: the stock is more interesting at $27 than it was at $46 — but not because revenue grew 407%

Applied Digital is in a better fundamental position than it was when we first wrote about it in June. It has more live capacity, more evidence that base rent can produce strong property-level NOI, and a full-year report that gives investors a clearer view of how the platform actually earns money.

The stock is also dramatically cheaper than the level used in that earlier analysis.

That combination makes APLD more interesting.

But the reason is not the 407% headline revenue growth. The reason is that the recurring rental business has finally become visible — and the market price has fallen before most of the contracted megawatts have begun producing rent.

At roughly $27–28, I would describe Applied Digital as a high-upside infrastructure equity with a more favorable risk/reward than it had near the mid-$40s, but still far too leveraged to execution and financing to call “cheap” in the conventional sense. The bull case can justify a much higher valuation if the campus handoffs keep arriving on time. The bear case can still take the stock into the low $20s or below if the buildout stalls.

The central question has changed.

In June it was: Are the leases real enough to justify the story?

Now it is: How much of the $36 billion lease book can Applied Digital convert into recurring NOI without spending away the economics before common shareholders receive them?

That is a much better question. It is also the question that will decide the stock.

Sources

• Applied Digital: Fiscal Fourth Quarter and Full Year 2026 Results
• Applied Digital 2026 Form 10-K
• Applied Digital investor-relations press releases and financing updates
• Late-August APLD historical market-price reference

This article is educational financial analysis, not investment advice. Scenario values are illustrative and depend on assumptions that may prove wrong.

Related company analysis: Applied Digital is one way to own AI infrastructure risk. Palo Alto Networks is the opposite side of the stack: a software platform monetizing the security complexity that AI creates. Our latest Palo Alto Networks stock analysis examines that platformization thesis and its demanding valuation.

Where Applied Digital sits in the AI-infrastructure capital chain

Applied Digital is only one part of a much larger physical buildout. Our CoreWeave deep dive shows the more leveraged GPU-cloud version of the same power-and-financing problem, where depreciation and interest can absorb a large share of headline growth. Our Dell analysis looks at the systems layer, where huge AI orders have to translate into margin and cash conversion. At the hyperscaler level, our Amazon AWS analysis explains the capex cycle that ultimately supports demand for data-center landlords and compute providers.

IREN adds a financing-heavy AI-cloud comparison

Applied Digital is not the only infrastructure company trying to convert scarce power and data-center capacity into higher-value AI revenue. Our IREN analysis examines a more vertically integrated model in which power sites, GPU financing, customer prepayments and cloud contracts all sit inside the same capital structure. Comparing the two helps separate genuine infrastructure scarcity from the financing and dilution risks that can determine per-share returns.