Applied Digital

The contracted book got dramatically bigger and dramatically better-rated: three new 15-year take-or-pay leases (Delta Forge 1, Polaris Forge 3, Delta Forge 2) totalling ~$20B were all signed with the SAME high investment-grade hyperscaler, taking the portfolio to 1,410 MW and ~$36B of base-term contracted revenue. And unlike most AI-infrastructure promises, the delivery track record is on schedule.

• Repeat business from a single high-IG hyperscaler is the strongest available third-party validation: the same customer chose Applied Digital three consecutive times, in three different states, for 810 MW and ~$20B of base-term revenue. Customers with alternatives do not re-up three times with an operator they doubt.
• The backlog is contractually take-or-pay over 15-year base terms: ~$36.2B across 1,410 MW, ~$86B if renewal options exercise (three five-year options per campus, two at Polaris Forge 2). Backlog of that duration against a ~$8.6B market cap is the core of the thesis.
• Delivery is actually happening on schedule - the differentiator management claims and, in this cohort, the rarest one. Polaris Forge 1's first 100 MW energized in October 2025, and Building 2 Phase 1 (75 MW) hit Ready-for-Service on time post-quarter, taking live capacity to 175 MW.
• The underlying rental economics are genuinely good once the pass-through is stripped out: FY26 rental revenue $114.7M against $56.8M of cost, and Net Operating Income $90.4M on $99.8M of base rent - a ~91% NOI margin. Adjusted EBITDA scaled 5.5x to $107.2M and Q4 adjusted EPS was positive at $0.04.
• Operating cash flow turned positive for the first time (+$89.7M in FY26 vs -$115.4M in FY25), so the operating business is no longer a cash drain - the cash burn is entirely discretionary construction capex.
• Capital access has been proven at scale and at improving terms: $6.88B of financing in FY26, including $2.15B of 6.750% secured notes and, post-quarter, $1.59B of 7.000% notes issued AT PAR (vs 98% of par for the prior tranche), plus a revolver upsized to $430M committed. Counterparty credit was also hardened - CoreWeave lease security was restructured with springing guarantees from CoreWeave, Inc. and a $50M letter of credit following CoreWeave's A3-rated refinancing.
• A ~1.7 GW additional pipeline is being actively marketed, plus roughly 3 GW of near-term gross utility power in diligence and a ~1.2 GW front-of-meter gas generation project via Base Electron (APLD holds ~10%) - power access, not demand, is the industry bottleneck, and this is an attempt to own that constraint.
• The Cloud Services distraction is gone: separated into ChronoScale (Nasdaq: CHRN, ~96% owned) and excluded from non-GAAP measures, leaving a cleaner landlord story - plus a stable 286 MW bitcoin-hosting business paid on capacity, not BTC price.

The 167% / 407% revenue growth is largely a zero-margin construction pass-through: fit-out services generated $270.6M of FY26 revenue against $258.1M of cost. Strip that out and the real business is ~$115M of rental revenue against $5.3B of debt (already ~$6.9B pro forma), $1.96B of preferred sitting above common, and a $249M GAAP loss. The market saw it - the stock fell ~12.8% two days after the beat.

• The headline growth is an accounting artifact of construction, not operating scale: FY26 tenant fit-out revenue $270.6M at $258.1M of cost = ~4.6% margin; Q4 $152.4M at $145.6M = ~4.5%. Management itself flags fit-out will not repeat at this scale. Recurring rental revenue is only $114.7M - roughly 19% of the headline.
• Leverage is the dominant risk and it is still climbing: $5,306.7M of scheduled debt principal at May 31, 2026 (from ~$700M a year earlier), plus $1.59B of new 7.000% notes closed after year-end - roughly $6.9B pro forma. Contractual interest on existing debt alone totals $1,784.7M, with $377.5M due in FY2027 against $611M of FY26 revenue.
• The capital structure buries common equity: $1,956.3M of redeemable noncontrolling interest (preferred units at APLD HPC TopCo 2, with embedded redemption and contingent-dividend-rate-increase features valued by a third-party specialist using a binomial lattice) plus $62.8M of Series E/E-1 preferred sit in temporary equity ahead of common. Book equity attributable to APLD is $1,717.5M against $9,929.3M of total assets.
• Customer concentration is now extreme in a new way: ~$20B of the ~$36B backlog is with ONE unnamed high-investment-grade hyperscaler and ~$11.0B with CoreWeave. Two counterparties are effectively the entire book, and the CoreWeave exposure required a restructured SPV, springing guarantees and a $50M letter of credit to make it financeable - a tell that the original credit was thin.
• Nearly all of the contracted revenue is unbuilt: of 1,410 MW contracted, only 175 MW is live. Delta Forge 2 does not deliver until 1H2028. Every dollar of that $36B depends on multi-year construction, power interconnection and equipment lead-times (management explicitly flags generator lead-times and steel/copper commodity risk) executing on schedule.
• Stock-based comp is a very large real cost being adjusted away: SG&A rose $224.2M to $332.1M, of which ~$198.3M was the increase in SBC. Management's non-GAAP measures exclude it, which is how a -$236.5M GAAP operating loss becomes +$57.8M 'adjusted operating income'.
• Reported results carry heavy non-operating noise that flattered the loss: a $75.8M gain on change in fair value of derivatives (mostly the Babcock & Wilcox warrant) and a $10.8M gain on investments offset a $59.7M loss on the Cloud/ChronoScale held-for-sale write-down. None of it is operating cash.
• The REIT conversion is NOT mentioned in the FY2026 10-K or the Q4 earnings release. Third-party coverage still describes it as under consideration, so the re-rating catalyst is at best unconfirmed at 4 August 2026.
• Related-party and off-balance-sheet complexity is increasing: Base Electron is a non-consolidated VIE ~10% owned by APLD but managed partly by APLD officers and directors in their individual capacities, funded by an APLD demand note that grew to $58.6M with a $100M advance ceiling, and APLD guaranteed (then was released from) B&W's 1.2 GW design-build obligations. That is a lot of structure around the power story.
• No formal FY2027 guidance was given - investors are asked to underwrite a $36B backlog with no management-provided revenue or EBITDA bridge to it.