DeepValue | The New Way Investors Work

At first glance, Applied Digital looks like everything an AI‑infrastructure investor could want: massive long‑term leases, cutting‑edge data center design, and a share price that’s more than quadrupled in a year. Under the surface, though, the story is more nuanced—and far less forgiving—than the momentum narrative implies.

As of late January 2026, APLD trades around $33.88 per share, valuing the company at roughly $9.5B. That valuation rests on the assumption that its $16B+ backlog of long‑term AI data center leases will ramp largely on time, at high margins, and without major hiccups in funding or tenant credit quality. Yet according to the 10-Q (2026) and recent press releases, Applied Digital is still loss‑making, heavily levered, and generating sharply negative free cash flow as it builds out its flagship North Dakota campuses.

Our DeepValue team’s conclusion: APLD now behaves like a high‑beta AI infrastructure option more than a classic “picks and shovels” compounder. The upside remains meaningful if everything goes right, but the margin of safety at today’s price looks thin. We classify the stock as a POTENTIAL SELL, with a bias toward trimming into strength rather than adding new capital at current levels.

If you’re trying to keep up with fast‑moving AI infrastructure names like Applied Digital, it’s easy to drown in 10‑Ks, 10‑Qs, and industry commentary. We’ve found that using DeepValue to digest SEC filings and niche industry sources cuts our research time dramatically and standardizes how we compare complex, capital‑intensive businesses.

See how DeepValue’s AI agent turns dense SEC filings and industry sources into a fully cited, 3‑part institutional-grade stock dossier in minutes, not days.

Run Deep Research on APLD →

Let’s unpack the bull and bear case, walk through what the market is really pricing in, and outline what we’re watching through 2026.

What Exactly Does Applied Digital Do?

Applied Digital started life as a crypto hosting company and has spent the last few years pivoting into high‑density data centers for AI and high‑performance computing (HPC). The core of the story today is its North Dakota “Polaris Forge” campuses, where it’s building large, power‑dense facilities using proprietary waterless direct‑to‑chip cooling.

According to GlobeNewswire, Aug 18 2025, the plan is:

Polaris Forge 1 (Ellendale): 400MW high‑density campus for CoreWeave, with contracts that could total roughly $11B in revenue over about 15 years.
Polaris Forge 2: 280MW campus backed by a $3B capex plan, anchored by a $5B, ~15‑year “AI Factory” lease with a U.S. investment‑grade hyperscaler, as detailed in GlobeNewswire, Oct 22 2025.

Layered on top is a negotiated—but not yet signed—~900MW multi‑campus hyperscaler deal that management hopes to secure in early‑to‑mid 2026, per Barron’s, Jan 2026.

Alongside these AI campuses, Applied Digital retains a legacy Data Center Hosting segment serving a single crypto customer in Jamestown and Ellendale and is in the process of divesting its GPU‑based Cloud Services business (ChronoScale). The 10-K (2025) repositions the company firmly as an infrastructure landlord and capital allocator, not a vertically integrated AI cloud operator.

From a business‑model standpoint:

The company signs long‑term (≈15‑year) megawatt‑priced leases, often “take‑or‑pay” style.
It earns construction and fit‑out revenue as it builds out space.
Once energized, it collects recurring lease payments that should resemble net operating income (NOI) from a specialized data‑center REIT.

The catch? As the 10-Q (2026) and Form 10-Q, Aug 31 2025 make clear, Applied Digital is still in a heavy build phase. Construction revenues and costs dominate current results, obscuring steady‑state economics and leaving GAAP margins extremely thin.

Why Has APLD Stock Exploded Higher?

Over the past year, APLD’s share price has rocketed from about $6.48 to $33.88—an increase of roughly 423%, according to FMP pricing data. Mainstream financial media now group Applied Digital with much larger AI data‑center infrastructure plays and REITs.

Several forces have driven that move:

1. Huge lease signings and backlog headlines

Management has announced:

400MW of CoreWeave leases with roughly $11B in potential revenue at Ellendale, per GlobeNewswire, Jun 2 2025.
A $5B, 200MW “AI Factory” lease with an investment‑grade hyperscaler at Polaris Forge 2, per GlobeNewswire, Oct 22 2025.

Combined, these drive contracted revenue of about $16B, as highlighted in the report and in GlobeNewswire, Oct 9 2025.

2. AI infrastructure “pick‑and‑shovel” narrative

Publications like Investor’s Business Daily, Jan 2026 and Barron’s, Jan 2026 now present Applied Digital as a primary AI data‑center beneficiary, sitting at the intersection of explosive AI compute demand and power‑constrained grid infrastructure.

3. Technical and ESG edge

Applied Digital emphasizes waterless direct‑to‑chip cooling and very low power usage effectiveness (PUE). An AInvest, Jan 2026 analysis cites a targeted PUE of 1.18 at Ellendale and management’s claim of $2.7B in lifecycle cost savings over 30 years for a 100MW deployment versus conventional data centers. That’s catnip for hyperscalers trying to minimize both cost and water use.

4. Strong project milestones

The company has hit early “Ready for Service” milestones on schedule at Polaris Forge 1. GlobeNewswire, Oct 27 2025 and GlobeNewswire, Nov 24 2025 show the first 100MW building reaching full energy status, which supports the “they can actually build this” part of the story.

The market has taken these positives and extrapolated aggressively. As our base case scenario in the report notes, consensus now assumes that strong AI demand will keep utilization high and that management will execute the Polaris Forge ramp with only modest friction.

Under the Hood: Financial Reality vs Hype

When we move from headlines to numbers, the picture looks substantially less comfortable for a value‑oriented investor.

Income statement and cash flow pressures

From the 10-Q (2026) and FMP data:

Revenue for the period ended November 30, 2025 was $126.6M, up strongly year over year.
Yet trailing EPS remains about –$1.16.
Adjusted EBITDA margin in the earlier quarter ended August 31, 2025 was just ~1%, down from 18% a year prior, as construction and fit‑out costs offset revenue, per Form 10-Q, Aug 31 2025.
Operating cash flow for the period ended November 30, 2025 was –$15.8M, and free cash flow was an eye‑watering –$567.9M, largely driven by –$552.1M in capex, according to FMP.

These aren’t the numbers of a stable infrastructure landlord yet. They’re the numbers of a company mid‑construction, pulling forward costs and funding them with expensive capital.

Heavy leverage and high-cost debt

The balance sheet is where the risk really concentrates.

Applied Digital has issued $2.35B of 9.25% senior secured notes due 2030, as detailed in the Nov 13 2025 company press release.
It also has $450M of 2.75% convertible notes due 2030 and preferred equity obligations.
Net debt sits around $589M, with interest coverage of about –2.1x, per FMP ratios and the report.

Just the interest on the 9.25% notes alone runs over $217M annually, before considering other obligations. With negative operating cash flow and no steady‑state NOI yet, that’s a precarious place for common equity holders.

Minimal margin of safety

On the asset side, Applied Digital’s main support is the $16B+ in long‑dated contracted revenues and an undiscounted minimum lease payment schedule of about $15.6B (presented as $15,551,105 in thousands) across Polaris Forge 1 and 2, per the 10-K (2025). But those cash flows are:

Heavily back‑loaded beyond 2030.
Concentrated in a handful of tenants—namely CoreWeave and one hyperscaler.
Vulnerable to ramp timing, tenant financing conditions, and potential renegotiations.

Given that the company currently lacks robust earnings, has highly negative free cash flow, and relies on tenant solvency and execution rather than tangible asset liquidation value, we do not see a classic margin of safety at the current equity valuation. Position sizing here needs to reflect that you’re underwriting a leveraged, execution‑dependent build‑out, not clipping coupons on an established REIT.

If you regularly evaluate leverage‑heavy, capital‑intensive stories, this is where an AI research assistant really helps. Read our AI-powered value investing guide to see how tools like DeepValue can parse credit terms, lease schedules, and covenant language directly from filings so you’re not flying blind on balance‑sheet risk.

Use DeepValue to scan 10-Ks, 10-Qs, and bond covenants across multiple tickers in parallel so you can compare leverage, coverage, and cash-flow trends side by side.

See the Full Analysis →

Customer Concentration: A Feature and a Bug

From a commercial standpoint, winning CoreWeave and an investment‑grade hyperscaler is a major validation of Applied Digital’s technology and North Dakota strategy. But from a risk standpoint, this concentration is a double‑edged sword.

According to the 10-K (2025):

In FY 2025, four customers accounted for 100% of Data Center Hosting revenue, with one representing 93%.
The company expects a limited number of customers to continue to represent a large share of revenue going forward.

For crypto hosting, Applied Digital now effectively serves a single mining customer under a five‑year contract (with 2.5 years remaining as of May 31, 2025). That customer has options to reduce or discontinue service, subject to notice and conditions.

On the AI side, the situation is similar:

CoreWeave anchors 400MW of capacity at Polaris Forge 1.
A U.S. investment‑grade hyperscaler anchors 200MW at Polaris Forge 2.
The touted ~900MW multi‑campus deal, if signed, would further concentrate exposure in a small set of hyperscale counterparties.

This structure provides excellent visibility if tenants stay healthy and ramp on schedule. But the report notes that CoreWeave, for example, cut its 2025 capex plan by about $8.5B (~40%) due to third‑party capacity delays, despite rising RPO, as covered by Investor’s Business Daily, Nov 2025. That’s a real‑world demonstration that “contracted demand” doesn’t automatically translate into near‑term cash collection.

Meanwhile, The Verge, Dec 2025 details CoreWeave’s GPU‑backed loans and private credit dependence—reminders that neocloud tenants themselves can be financially fragile.

From our perspective, this combination—high leverage at the landlord level and concentrated, aggressively financed tenants—amplifies downside risk if funding conditions for AI workloads tighten.

Is APLD Stock a Buy in 2026?

At the heart of the investment decision is a simple question: Does the current price appropriately compensate you for execution, financing, and tenant risks over the next 6–24 months?

Our answer, based on the report’s scenarios and current trading levels, is no.

Scenario framework and implied values

Our DeepValue scenario work frames the risk/reward as follows (all from the one‑pager):

Base case (45% probability)

Implied value: $35
Assumes strong AI demand, broadly on‑time energization of Polaris Forge 1/2, EBITDA margin in the low teens, and operating cash flow approaching breakeven.

Bear case (35% probability)

Implied value: $20
Assumes the 9.25% debt burden and tighter credit conditions slow build‑out and force more dilutive funding. Polaris Forge 1/2 ramps materially lag, EBITDA margin stays below 10%, and interest coverage remains negative.

Bull case (20% probability)

Implied value: $50
Assumes AI capex stays very strong, the 900MW deal closes, North Dakota campuses ramp smoothly, EBITDA margin exceeds 20%, and free cash flow turns clearly positive.

At around $33.88, the market is essentially paying base‑case price for a base‑plus‑bull combination of outcomes. In other words, you’re not picking this up at a distressed “what if they just muddle through” valuation; you’re paying up for a story that still has a lot to prove.

What the market is implicitly assuming

Based on the “Market‑Implied Assumptions” section of the report, today’s valuation bakes in several optimistic beliefs:

AI/hyperscaler demand will remain strong for at least 12–24 months, keeping new megawatt leases and renewals flowing, as Barron’s, Jan 2026 coverage suggests.
APLD will energize and ramp its North Dakota campuses broadly on time by 2026–2027, as assumed by MarketMinute, Aug 2025.
High‑coupon debt will remain serviceable, with access to refinancing and project‑level financing, per the expectations around the 9.25% notes.
Revenue growth from AI leases will outpace cost growth, driving margin expansion and progress toward management’s aspirational $1B NOI target, a figure cited in Barron’s, Jan 2026.
Customer concentration will be managed by adding at least one more major hyperscaler beyond CoreWeave, observable through future MW signings.

These are not insane assumptions; they’re plausible. But they are stringent: they require multiple things to go right in parallel for a relatively young company with a leveraged balance sheet and a short operating history.

Our view: that’s not the setup we want to see when committing new capital at a 400%+ one‑year gain.

What Could Change the Call?

We’re not dogmatic. Our “POTENTIAL SELL” stance can move in either direction based on new data.

The one‑pager lays out clear triggers:

Call decreases (more bearish) if by mid‑2026:

Interest coverage remains below 0.5x, and
Net debt exceeds $1.2B, despite growth in energized megawatts.

Call increases (more constructive) if by mid‑2026:

Adjusted EBITDA margin exceeds 15%, and
Operating cash flow is positive for two consecutive quarters.

That outcome would indicate that contracted revenue is finally overcoming construction drag, and that the capital stack—while still heavy—is no longer existentially threatening.

We also monitor a set of 90‑day and 6‑ to 18‑month checkpoints:

By April 30, 2026: Evidence of additional Ready for Service milestones at Polaris Forge 1 or clear construction progress at Polaris Forge 2, per GlobeNewswire, Nov 24 2025 and GlobeNewswire, Aug 18 2025.
By July 31, 2026: Either the ~900MW hyperscaler negotiation is signed or it’s clearly replaced by a comparable deal, as flagged in Barron’s, Jan 2026.
Every quarter: Trend in adjusted EBITDA margin, interest coverage, and net debt/EBITDA using FMP data and SEC filings like the 10-Q (2026).

If APLD hits these operational and financial milestones while the stock pulls back or trades sideways, we’d be more inclined to look at dislocations as entries rather than exits.

Will Applied Digital Deliver Long-Term Growth?

Beyond the near‑term trading setup, investors need to decide whether Applied Digital can evolve into a durable, high‑return infrastructure franchise—or whether it’s a cycle‑peak AI build‑out story.

On the positive side:

The technical moat looks real. Low PUE, waterless cooling, and rapid deployment times (12–14 months from concept to power) differentiate APLD from traditional colocation players, per AInvest, Jan 2026 and GlobeNewswire, Oct 27 2025.
The North Dakota power advantage positions it well in a world of national grid constraints, as highlighted by GlobeNewswire, Aug 18 2025.
Third‑party validation from NVIDIA‑backed CoreWeave and institutional capital partners like Macquarie (see DataCentreMagazine/GlobeNewswire, Dec 18 2025) supports the idea that Applied Digital is building assets the ecosystem actually wants and is willing to finance.

On the negative side:

The balance sheet remains fragile, with high coupon debt and convertible overhang.
Customer concentration and neocloud credit risk are serious; CoreWeave’s financing model, described in The Verge, Dec 2025, shows how quickly lender sentiment could shift.
Industry competition from large data‑center REITs and hyperscaler self‑builds is intensifying, squeezing the space where APLD can earn super‑normal returns.

We think the path to long‑term value creation is narrow but not nonexistent. It likely requires:

1. Disciplined capital allocation from here—more Macquarie‑style project equity, less reliance on high‑coupon corporate debt, and measured use of equity as the stock price allows.

2. Successful ramp of Polaris Forge 2 with visible, high‑margin lease economics from the investment‑grade hyperscaler by 2026–2027.

3. Tenant diversification beyond CoreWeave and the initial hyperscaler, so that no single counterparty dominates revenue and bargaining power.

If management threads that needle, Applied Digital could justify and even exceed current valuations over a multi‑year horizon. But until we see clear evidence on margins, cash flow, and leverage trending in the right direction, we think the stock should be treated as a high‑risk, high‑beta position, not a core compounder.

How We’d Approach APLD as Investors

Putting it all together, here is how we’d approach Applied Digital from a portfolio‑construction standpoint.

1. Respect the run and the risk/reward skew

After a 400%+ gain in a year, and with the stock trading around our $35 base‑case value, the asymmetry is no longer your friend. Upside from here depends on fast, clean execution across multiple fronts; downside only needs normal project hiccups, funding friction, or a tenant‑level surprise.

For existing holders with sizable gains, we lean toward trimming on strength, especially above the report’s $45 “Trim Above” marker. That’s where expectations move from “optimistic but plausible” into “priced for perfection.”

2. Size it like an option, not a bond proxy

Despite the data‑center and lease language, APLD should not be sized like a stable infrastructure REIT. The right mental model is a leveraged call option on AI infrastructure build‑out, with real execution and balance‑sheet risk.

In practice, that means:

Keeping position size modest relative to diversified, cash‑generative holdings.
Avoiding leverage or derivatives that multiply exposure.
Being willing to cut if early‑warning indicators start flashing red (e.g., delayed energization, worsening interest coverage, tenant financing stress).

3. Wait for either a reset or real de‑risking

For new money, we see two attractive entry conditions:

Valuation reset: A drawdown toward or below our $24 Attractive Entry zone, possibly triggered by sentiment shifts, minor delays, or equity issuance, could create a more favorable skew—provided the fundamental thesis isn’t broken.
Fundamental de‑risking: Evidence of sustained mid‑teens EBITDA margins, positive operating cash flow, and stabilizing leverage may justify paying closer to today’s price, especially if the 900MW deal is signed and tenant diversification improves.

Until one of those two scenarios materializes, we’d rather watch and track than chase.

If you want to keep APLD and its peers on your radar without manually combing through every 10‑Q and press release, using DeepValue as a standing “research agent” can help you update your view every quarter in a consistent, structured way.

Let DeepValue pull the latest SEC filings, leases, and capital-structure details so you can re-underwrite your APLD thesis every quarter with fresh, citation-backed data.

Research APLD in Minutes →

Applied Digital (APLD) Deep Research Report: High-Beta AI Infrastructure Play With More Risk Than Margin of Safety?