A-Portfolio

The company in one breath

Applied Digital, ticker APLD on Nasdaq, builds giant data centers and rents the power, cooling, and floor space to companies running artificial intelligence systems. CoreWeave (CRWV) and one unnamed hyperscaler pay through long leases because graphics chips are useless without buildings that can feed and cool them.

This is closer to a landlord than a chipmaker. Nvidia (NVDA) sells the shovels. Applied Digital builds the mine, wires it, cools it, and hands over the keys. Good would look like each finished campus producing predictable rent for years while the company funds new construction without repeatedly watering down shareholders.

Why now

Applied Digital reports fiscal first-quarter results after today's close. Wall Street expects roughly $134.1 million of revenue, up 109% from one year earlier, and a loss of $0.26 per share. Revenue is the money earned during the quarter, while loss per share spreads the accounting loss across each share. Good would look like revenue above $135 million, a loss smaller than $0.26, and clean proof that recurring rent, not one-time construction work, drove the beat.

The operating story just moved. Polaris Forge 1 reached 250 megawatts of live computing capacity on October 2, up from 175 megawatts before the latest halls opened. A megawatt measures power capacity, and more live megawatts create more space that can earn rent. Good would look like the campus reaching its contracted 400 megawatts on schedule, with each new block producing rent soon after the customer accepts it.

The company also announced access to as much as 1 gigawatt of potential power in Finland, its first international project. One gigawatt equals 1,000 megawatts, enough power for a very large AI campus. Good would look like a signed tenant, firm utility milestones, and project financing before management spends heavily. Today Finland is an option, not revenue. A shiny map pin does not pay interest.

The stock closed Tuesday at $25.34 and traded at $24.78 at 7:06 a.m. Eastern today. The close values the common shares near $7.66 billion. That means buyers already pay for years of successful construction, not the business sitting here today. Good entry quality would look like a lower price or a much larger stream of recurring rent.

The numbers, translated

• Fiscal 2026 revenue reached $611.3 million, up 167%. Revenue is the money the company earned from rent, tenant construction work, and legacy hosting. Good growth for an early infrastructure build would look above 30%, so the headline crushed the bar.
• Only $114.7 million of that annual revenue came from data-center rent and related recoveries. That means less than 19 cents of each revenue dollar came from the landlord stream investors really want. Good would look like recurring rent becoming the majority of revenue as finished capacity opens.
• Fourth-quarter revenue reached $258.7 million, but $152.4 million came from tenant fit-out work. Fit-out work means Applied Digital built equipment and space for a tenant, which can create lumpy sales that do not repeat every quarter. Good would look like base rent rising every quarter and one-time fit-out revenue becoming less important.
• The company signed roughly $36.2 billion of contracted revenue across 1,410 megawatts for initial terms near 15 years. Contracted revenue means customers agreed to pay if Applied Digital delivers the buildings under the lease terms. Good would look like operating megawatts, accepted buildings, and collected rent climbing together. A contract is a loaded order book, not cash in the register.
• Only 100 of those 1,410 contracted megawatts produced revenue at the May 31 year-end, or 7.1%. That means more than nine-tenths of the contracted story still needed construction. Good would look like the operating share moving above 25% during fiscal 2027 without major delays.
• The company ended May with $1.59 billion of unrestricted cash, $2.38 billion of restricted cash, and about $4.98 billion of debt. Restricted cash has a job already, usually construction or debt service, so management cannot spend it like ordinary cash. Good would look like rent covering interest and construction needs while net debt, debt minus usable cash, falls from roughly $3.38 billion.
• Fiscal 2026 operating cash flow reached about $89.7 million, but property and equipment spending approached $2.94 billion. Operating cash flow is cash generated by normal business activity, while property spending builds the campuses. Applied Digital spent about $33 on construction for every $1 of operating cash. Good would look like operating cash rising fast enough that outside financing supplies less of each new project.
• The common-stockholder loss reached $249.2 million, equal to 41 cents lost for every revenue dollar. This is the accounting bill after operating costs, interest, depreciation, and other items. Good would look like that loss margin moving below 10 cents per dollar as rent starts.
• Stock-based compensation reached $220.1 million, equal to 36% of revenue. Stock-based compensation pays workers with ownership, which can increase the share count and shrink each existing owner's slice. Good would look below 10% of revenue and falling.
• At the $25.34 close, the common shares cost about 12.5 times trailing revenue, and the operating business costs about 18.1 times after adding debt and subtracting unrestricted cash. These ratios tell us how many dollars investors pay for each dollar of last year's sales. Good for a capital-heavy, loss-making landlord would look below 10 times sales, or higher only after recurring rent and cash flow explode.

What I like

Applied Digital has real contracts, not handshake hype. Its 1,410 megawatts of signed capacity carry about $36.2 billion of base-term revenue. Take-or-pay means a tenant owes rent once the facility meets the lease conditions even if it uses less capacity. Good would look like those contracts surviving construction, acceptance, and financing with no ugly amendments.

Execution has improved. Polaris Forge 1 now runs 250 megawatts, and the latest 75-megawatt block arrived before today's earnings report. Good would look like management repeating that delivery rhythm at Polaris Forge 2, Polaris Forge 3, and Delta Forge 1 instead of treating one campus as a victory lap.

The company owns a scarce ingredient, power. AI customers can buy chips, but they cannot instantly create utility connections, land, cooling, and high-density electrical systems. Digital Realty (DLR) and Equinix (EQIX) prove that good data-center property can become durable infrastructure. Good would look like Applied Digital earning landlord-like margins without carrying startup-like financing risk forever.

Management owns about 9.5% of the common stock, including directors and named executives. That means leadership feels part of the stock's gain or pain. Good would look like the ownership stake staying meaningful while compensation shifts away from giant stock grants.

The public debate has the right argument. Recent community research found five YouTube videos with 17,745 views, and the strongest analysis kept returning to the same split: signed rent versus debt and dilution. NikRayaniTrades put it cleanly, "Applied Digital has 36 billion in signed rent from some of the biggest companies in the world." Good would look like the next year turning that sentence from pitch into collected cash.

What worries me

Debt does not care how exciting the keynote sounds. Applied Digital carries about $4.98 billion of debt against $1.59 billion of unrestricted cash. Good would look like each project supporting its own interest bill and principal without forcing the parent company to sell common shares at bad prices.

Customer concentration is brutal. One unnamed hyperscaler represents about 56% of contracted revenue, while CoreWeave (CRWV) anchors all 400 megawatts at Polaris Forge 1. Good would look like a third large customer and no single tenant above 35% of the contract book. Two customers make a relationship. They do not make diversification.

The share count keeps moving. Applied Digital reported 291.5 million shares on July 28, then 302.3 million by September 23, an increase near 3.7% in eight weeks. More shares mean each old share owns less of the company. Good would look like diluted share growth below 2% a year once the current financing package settles.

The latest registration statement covers as many as 53.1 million shares for resale or possible issuance from preferred-stock conversions and warrants. That does not mean every share hits the market tomorrow, but it creates a dilution overhang, which means potential new supply hangs above the stock. Good would look like project cash replacing equity issuance before that ceiling gets tested.

The accounting profit needs a flashlight. Fiscal 2026 adjusted net income reached $36.1 million, but management removed $219.3 million of stock-based compensation from that adjusted figure. Good would look like standard-accounting profit and adjusted profit moving closer together. When the adjustment is six times the profit, the asterisk owns the sentence.

Last-thirty-day discussion looked loud but sloppy. The broad scrape found 26 Reddit threads and 158,149 upvotes, yet the highest-engagement threads mostly missed Applied Digital and earned no weight here. Good would look like customers, engineers, and creditors discussing uptime, acceptance, cooling performance, and financing terms, not ticker chants before earnings.

What would make this a buy

My position size is zero. This is research and commentary, never a buy signal.

The price door starts at $18 to $20. At $18, common equity would cost about $5.44 billion, or 8.9 times trailing revenue. At $20, it would cost about $6.05 billion, or 9.9 times trailing revenue. Those prices still assume growth, but they leave more room for construction delays, debt, and dilution than $25.34 does.

The proof door matters more. I want data-center base rent above $75 million in one quarter, at least 300 operating megawatts, operating cash flow above $150 million over twelve months, and diluted share growth below 5%. Those numbers would mean the buildings are turning into rent, the rent is turning into cash, and common owners are keeping enough of the upside. Good would look like all four arriving together.

I also want management to give firm ready-for-service dates for the remaining 150 megawatts at Polaris Forge 1 and explain the financing for the next campuses without hiding behind total contract value. Good would look like project-level funding matched to signed leases, fixed costs, and visible tenant acceptance tests.

What makes me walk away

I walk if a major campus slips by more than two quarters. That would mean the construction machine cannot match the promises that support the valuation. Good would look like every delay measured in weeks, not seasons.

I walk if gross debt grows faster than operating megawatts for another full year. Debt growth without live capacity means the company borrows for tomorrow while interest arrives today. Good would look like rent-bearing capacity outrunning the debt balance.

I walk if the diluted share count climbs above 330 million before recurring rent covers interest and overhead. That would mark roughly 9% dilution from the September count, enough to eat a meaningful chunk of business progress. Good would look like management funding projects without passing the hat to common shareholders again.

I walk if either big tenant cuts capacity, renegotiates economics, or invokes delay rights. With two customers carrying most of the contract book, one broken relationship can punch through the whole model. Good would look like new tenants reducing concentration before trouble arrives.

IPO Watch

• Retension Pharmaceuticals, private, proposed ticker RTSN, expected to trade October 9 at $11 to $13. It develops a once-daily pill for uncontrolled and resistant high blood pressure. I would write it up when the final price and cash runway show whether the offering can fund the Phase 2b readout expected in the first half of 2027 without another quick stock sale.
• TRex Bio, private, proposed ticker TRXB, expected to trade October 9 at $14 to $16. It develops medicines that boost regulatory T cells, immune cells that keep inflammation from attacking the body. I would write it up when final pricing shows the valuation and when the filing proves the roughly $111 million of expected net proceeds can reach meaningful human trial data.

Earnings On Deck

• APLD, October 7 after the close. Watch revenue against roughly $134.1 million, loss per share against $0.26, base rent, live megawatts, capital spending, and the share count. A good print means recurring rent beats expectations, campus dates hold, and management funds growth without another common-stock wave. A bad print means fit-out work props up sales, losses widen, or another project slips.

The verdict

Score: 69 out of 100. Bucket: watch. Applied Digital earned points for scarce power, purpose-built cooling, 250 live megawatts, about $36.2 billion of signed base-term revenue, long leases, aligned insider ownership, and a near-term earnings catalyst.

It lost points for about $4.98 billion of debt, only 7.1% of contracted capacity operating at year-end, two-customer concentration, a $249.2 million common-stockholder loss, stock pay equal to 36% of revenue, rapid share growth, and a price above 12 times trailing sales. Candidate status starts at 80. Applied Digital misses by eleven because the construction proof improved, but the stock already charges rent on buildings that do not exist yet.

The DBTech read

Take-or-pay is a lease rule that makes the customer pay once the facility meets its agreed conditions, even if the customer uses less capacity. Applied Digital says its long leases cover about $36.2 billion, which gives the future revenue more muscle than a sales forecast. The catch sits before payment starts, the company still has to finish the building, connect the power, and pass acceptance tests. Good take-or-pay revenue comes from a strong customer inside a completed asset. The contract protects the dining room after opening night, it does not build the restaurant.

Sources and as-of date

• Price: $25.34 at 4:00 p.m. Eastern on October 6, 2026, the latest completed regular-session close, and $24.78 at 7:06:32 a.m. Eastern on October 7, 2026, the latest available premarket print, Yahoo Finance. Retrieved at 7:07 a.m. Eastern on October 7, 2026. Valuation uses the liquid $25.34 close.
• Financial results, cash flow, debt, cash restrictions, stock compensation, shares, lease terms, customer concentration, and risk factors: Applied Digital fiscal 2026 Form 10-K and fiscal fourth-quarter release, filed July 2026.
• Current share count, preferred-stock conversion exposure, and resale registration: Applied Digital September 25, 2026 prospectus. The filing listed 302,321,687 shares outstanding on September 23 and covered up to 53,087,689 shares tied to resale, conversions, and warrants.
• Management ownership and board structure: Applied Digital 2026 proxy statement, filed September 25, 2026.
• Operating capacity: Applied Digital Polaris Forge 1 update, published October 2, 2026. Finland power option: Applied Digital release, published October 6, 2026.
• Current business description and positioning: Applied Digital, retrieved October 7, 2026.
• Earnings date and consensus: Applied Digital conference-call notice and Zacks earnings preview via Yahoo Finance, retrieved October 7, 2026.
• IPO dates and ranges: Stock Analysis IPO calendar, Retension Pharmaceuticals amended filing, and TRex Bio amended filing, retrieved October 7, 2026. IPO dates can move until the deals price.
• Community research covered September 7 through October 7, 2026. The usable evidence included five YouTube videos with 17,745 views. The engine also returned Reddit, Hacker News, and GitHub results, but most high-engagement items missed the company and received no weight. Raw research: workspace/newsletters/a-portfolio/research/applied-digital-apld-raw-v3.md.
• Valuation calculations use Decimal arithmetic, the $25.34 closing price, 302.3 million shares outstanding, $4.98 billion of debt, $1.59 billion of unrestricted cash, and $611.3 million of fiscal 2026 revenue. Research and commentary only, never a buy signal.