Esta infraestructura de IA está contribuyendo silenciosamente al aumento de precios de las acciones relacionadas con ella. Además, podría estar previendo un enorme esquema de short selling que podría llevar los precios por encima de los 31 dólares.

Applied Digital is a pretty clean way to play the “picks-and-shovels” side of AI: instead of designing the chips or training the models, it builds and operates the powered data-center capacity those chips sit inside. In the AI infrastructure stack, APLD is essentially selling the hardest-to-source ingredient right now—ready-to-go megawatts with high-density cooling, grid access, and construction execution—so hyperscalers and AI cloud players can deploy GPU clusters without spending years fighting interconnection queues and supply chains. That positioning matters because the AI boom is increasingly constrained by physical infrastructure (power, cooling, and time-to-deliver), not just silicon.

On results , APLD delivered a clear beat versus Street expectations. Fiscal Q2 adjusted EPS was $0.00 versus an expected loss of approximately -$0.16, and revenue came in at $126.6 million versus roughly $89.8 million expected. Revenue was up 250% year over year, and adjusted EBITDA jumped to $20.2 million from $6.1 million a year ago. The key driver was the ramp of its HPC Hosting / “AI Factory” business at Polaris Forge 1, including a large amount of tenant fit-out revenue ($73 million) and partial-quarter lease revenue ($12 million) as the first building (ELN-02) became fully energized and “ready for service.” A second, steadier contributor was the legacy data center hosting business (historically more tied to crypto/bitcoin workloads), which generated $41.6 million of revenue and about $16 million of segment operating profit.

Operationally, the quarter was all about execution milestones and lease momentum. APLD delivered 100 MW at Polaris Forge 1 on schedule, completing the first of three contracted buildings supporting a 400 MW campus leased to CoreWeave. Management framed that CoreWeave relationship as roughly $11 billion of prospective lease revenue over the term, and the tenant’s payment activity was meaningful in the quarter—about $85 million received, largely tied to tenant fit-out. The bigger strategic headline, though, was the announcement of an approximately 15-year lease with a U.S.-based, investment-grade hyperscaler for 200 MW at the under-construction Polaris Forge 2 campus, with phased delivery beginning in 2026 and management estimating roughly $5 billion of prospective revenue over the lease term. Put together, the company says it now has two hyperscaler-grade customers across two North Dakota campuses, totaling 600 MW of leased capacity and about $16 billion of prospective lease revenue before renewals.

On the “who is the hyperscaler” question: management did not name the U.S. investment-grade hyperscaler, but it’s explicitly positioned as a large, creditworthy cloud provider (think the typical short list of big cloud platforms). Separately, Citizens noted the company is in “advanced negotiations” with yet another hyperscaler, and management said it’s in advanced discussions across multiple regions (more Dakotas and select southern markets), including commentary suggesting a pipeline of roughly 900 MW across three sites. The bullish interpretation is that leasing activity is accelerating at a time when parts of the market are still questioning AI data-center capex; the cautious interpretation—also flagged by Citizens—is that APLD can inadvertently “talk expectations” up faster than contracts can close, and long enterprise procurement cycles can create timing risk.

Balance sheet and funding were a central part of the story, because this is an infrastructure buildout that only works if capital is both available and repeatable. APLD disclosed a $2.35 billion private offering of 9.25% senior secured notes due 2030 (issued at 97% of par), with proceeds earmarked for construction at Polaris Forge 1, repayment of the SMBC loan, and required debt service reserves. It also leaned heavily on a preferred equity financing facility with Macquarie Asset Management (up to $5.0 billion), drawing an additional $562.5 million during the quarter and another $337.5 million after quarter-end; management said total draws to date are about $900 million. The company also entered a development loan facility with Macquarie Equipment Capital to fund pre-lease sourcing and early development activities (including an initial $100 million). Management’s pitch is that this “repeatable financing framework” lets APLD retain over 85% common equity ownership at the site level while minimizing the need to tap public markets repeatedly—important if you’re trying to scale multiple campuses on a tight timeline.

Liquidity looked robust on the face of it. As of period-end, APLD reported about $2.3 billion in total cash, cash equivalents, and restricted cash, with $2.6 billion of debt and the bulk of maturities pushed out to 2030. Total assets were cited at $5.2 billion, liabilities $3.2 billion, and roughly $2.1 billion of temporary and stockholders’ equity, plus additional financing proceeds closing after the quarter. The bears will focus on the cost of capital (9.25% secured notes are not cheap), the ramp in interest expense (up to $11.5 million), and the reality that this model demands continuous construction spending before the full revenue base is online; the bulls will counter that the long-duration leases and investment-grade counterparties are exactly what makes the financing viable.

So why is short interest so high (roughly 34% of a ~224 million float, per your notes)? The short thesis typically clusters around three buckets. First, execution and timing risk: this is a construction-heavy story where delays, interconnection issues, equipment lead times, or customer procurement cycles can push revenue recognition out while costs arrive on schedule. Second, financing and dilution risk: rapid expansion often forces companies into expensive debt, preferred equity, or equity issuance, and skeptics worry the “capital-efficient framework” still ends up diluting common shareholders or compressing returns if terms worsen. Third, legacy perception: APLD’s roots in blockchain/crypto hosting still color investor views, and some shorts treat it as a “story stock” that can overshoot on hype. The counter-argument—supported by the signed leases and “ready for service” milestones—is that the business is evolving into a hyperscaler-grade AI infrastructure landlord rather than a cyclical hosting operator.

From a trading setup standpoint, your range framing is the right lens. The stock has churned between roughly $21 and $31 for months and is now near the top end, with the kind of short interest that can turn a clean catalyst into an air pocket for shorts if price pushes through obvious levels. If the tape starts triggering stops above ~$31 and there’s follow-through volume—especially on any incremental lease confirmation—the ingredients for a squeeze are there. The risk, of course, is that “advanced negotiations” headlines can inflate expectations, and if conversion takes longer than investors want, you can get the same squeeze dynamics in reverse. In AI infrastructure, time-to-power is everything—and the market tends to price that like it’s either immediate or never, with very little patience in between.