The blunt verdict
Applied Digital has what most AI-infrastructure stories lack: the contracts are signed. Three hyperscaler leases, about 1.4 GW, about $36B of base-term lease revenue, take-or-pay, meaning the customer pays whether the capacity is used or not, with the GPU cloud split off into a separate listed company. The FY2026 results prove the revenue is real: $611.3M, up 167%. The risk is equally real: about $5B of debt on about $1.7B of equity, and a stock that has fallen 34% in three months. This is a builder trading at roughly 13x trailing sales with the lease book to grow into it. Positive verdict, high-risk grade, and the position size should reflect the balance sheet, not the story.
The business engine
Applied Digital designs, builds, owns, and operates hyperscale data centers purpose-built for AI compute, called AI Factories. Revenue comes from long-term, take-or-pay leases of critical IT load, which is the power delivered to the computing equipment, the only capacity customers can rent. The company was founded in 2021 by Wes Cummins (Chairman and CEO) and Jason Zhang (President) as a blockchain infrastructure company and rebranded in 2022 for the AI opportunity. In May 2026 the GPU cloud business separated as ChronoScale (Nasdaq: CHRN), with Applied Digital keeping about 96% ownership, a clean split between the lease-income factory business and the capital-hungry cloud business.
The contracts: 1.4 GW, $36B
The backlog is the moat. Selected disclosed leases: CoreWeave, 400 MW, roughly 15 years, about $11B, at Polaris Forge 1 in Ellendale, North Dakota (expanded August 2025); an investment-grade US hyperscaler, 200 MW, roughly 15 years, about $5B, at Polaris Forge 2 in Harwood, North Dakota, with a right of first refusal on 800 MW (October 2025); the same hyperscaler, 300 MW, roughly 15 years, about $7.5B, at Delta Forge 1 in Boyce, Louisiana, and another 300 MW, about $7.5B, at Polaris Forge 3 (July 2026). The three hyperscaler leases total 810 MW and about $20.2B. Unit economics: the full $36B over 1.4 GW over 15-year terms works out to roughly $1.7M of annual revenue per megawatt, and the first campus delivered at a power usage effectiveness of 1.18, near the best in the industry.
Delivery: the test the company keeps passing
Contracts mean nothing if campuses do not energize. Polaris Forge 1's first 100 MW went live on schedule: Phase I in October 2025, Phase II on November 24, 2025, and the campus is now a 400 MW fully contracted CoreWeave deployment with a path beyond 1 GW. Polaris Forge 2 broke ground in September 2025, with 200 MW phasing across two buildings energizing in 2026 and full capacity in 2027. That track record is why a second hyperscaler kept signing. Watch the same milestones through 2027: each energization converts backlog into cash flow, and any slip is the first warning.
The balance sheet: where the risk lives
This is the D in the grade card. Long-term debt reached about $4,959.5M in FY2026, up from $677.8M a year earlier, against about $1,717.5M of equity, a debt-to-equity ratio near 2.9x. The financings stack up fast: $2.15B of 6.75% senior secured notes due 2031, $1.59B more senior secured notes, a $430M revolving credit facility, and a $5B Macquarie preferred facility. The company raised $3,973M of cash and generated about $89.7M of operating cash flow in the latest quarter, but invested $2.94B, leaving free cash flow, operating cash flow minus capital spending, near negative $2.78B. The borrowings buy income-generating buildings; if deliveries slip, the interest clock does not. Also note: NVIDIA, which bought $160M of stock in September 2024, sold its entire 7.7M-share position in late 2025, disclosed in its February 2026 13F, and the stock fell 8.4% after hours. The exit looks like portfolio rebalancing, and Applied Digital remains an NVIDIA Cloud Partner, but it removed a marquee holder.
Technical setup: basing after a deep cut
From the $27.30 reference price the chart is a base, not a trend. The stock sits just above its 20-day average ($26.27) and below its 50-day ($27.66), with the 200-day ($32.27) overhead. Momentum is mixed but stabilizing: up 12.8% in five days against down 34.4% in three months, RSI 51.1, neutral. Support sits near $23.39 and then the $19 area; resistance stacks at $29.10, the 50-day, the 200-day, and the $50.73 high. With fiscal Q1 2027 earnings on October 8, 2026, the near-term path runs through that print.
The ruler: valuation and three-year targets
From $27.30, the scenario map is: bear $9.50 (-65%), base $41.50 (+52%), bull $88.00 (+222%). The formula is (FY2029 revenue times EV/Sales multiple, minus net debt) divided by diluted shares. Base-case logic: the contracted 1.4 GW is substantially online by FY2029 at $2.2B of revenue, which is $1.57M per MW per year, slightly below the contracted $1.7M rate, so the revenue carries a cushion; 8x EV/Sales sits between CoreWeave's roughly 5x and Nebius's roughly 17x forward multiples; net debt falls to $3.5B as cash flow funds construction. The bear case is delivery slips and a 6x multiple on a levered developer that missed schedule. The bull case is the pipeline converting beyond 1.4 GW at 10x sales. At 13x trailing sales today the stock is priced for backlog conversion, not hope, and there is no margin of safety on delays.
Risks and kill criteria
Principal risks: delivery delays pushing energization dates; leverage and refinancing on a 2.9x debt-to-equity base; customer concentration in CoreWeave and two hyperscalers; dilution, with shares rising from 224.9M to 287.9M in a year; and a broader AI capex pause. The kill criteria are contractual and financial: a lease cancellation, sustained multi-campus delivery slips, or an equity raise below book value would break the thesis outright.
Download the full report (PDF)
Research and opinion, not investment advice. Do your own due diligence before investing.