The global race to scale artificial intelligence is no longer just about who can secure the most advanced GPUs. Increasingly, it’s about who can secure the hardest inputs of all: grid-ready power, purpose-built data centers, and financing that can survive multi-year construction timelines.
That shift is pushing Applied Digital Corporation (NASDAQ: APLD) into the spotlight. The company has positioned itself as a builder and operator of high-density, liquid-cooled “AI factories” — specialized data centers designed to handle modern AI training and inference workloads — and it’s doing it with long-term leases that are starting to look like a prized form of infrastructure real estate. [1]
On December 15, 2025, the broader AI infrastructure story became even clearer: power providers and grid assets are now being treated as strategic choke points for AI growth, while new mega-projects in Europe show the buildout is far from slowing. [2]
Why Applied Digital keeps showing up in “AI infrastructure winner” conversations
Applied Digital’s recent investor attention comes down to one number: contracted backlog.
Market commentary published this week puts Applied Digital’s contracted backlog for AI data center-related revenue at nearly $16 billion, anchored by two large, long-duration lease arrangements:
- CoreWeave: an expanded deal valued at $11 billion over 15 years, tied to Applied Digital’s Polaris Forge 1 campus in Ellendale, North Dakota. The lease expanded from 250 MW to the full 400 MW under construction. [3]
- An unnamed U.S. hyperscaler: a separate 15-year lease valued at about $5 billion for 200 MW of capacity at Polaris Forge 2, also in North Dakota. [4]
These aren’t short cloud contracts that reset every few years. The strategic appeal (and the investor thesis) is that long-term leases can turn AI data centers into something closer to infrastructure-style cash flow — if the facilities are delivered on time and the tenants remain strong.
The real bottleneck: “energized” capacity — not empty land
The debate around AI infrastructure has increasingly focused on a blunt reality: it can take years to bring a new data center from concept to “energized” (connected to power, permitted, and ready to run high-density compute).
One recent market note highlighted that lead times for new data centers can run at least five years, in large part because securing land and electricity isn’t trivial — and hyperscalers can’t always wait. [5]
That’s where Applied Digital claims an edge: it has been building a pipeline of sites and supply chain relationships so that it can bring capacity online faster than greenfield competitors.
In recent commentary, Applied Digital is described as having:
- Two North Dakota sites with 286 MW of operational capacity, plus major expansion underway. [6]
- A 4 GW active development pipeline, with some projects positioned to enter construction in the next 6–12 months. [7]
- A construction timeline that has been reduced from roughly 24 months to about 12–14 months via earlier equipment and supply chain planning. [8]
In a world where power-ready data halls are scarce, speed-to-energization is increasingly treated like a competitive moat.
The “one catalyst” bulls keep pointing to: more hyperscaler leases
A recurring theme in recent coverage is straightforward: the next lease announcement.
Applied Digital’s stock has been a high-volatility proxy for AI data center demand — and investor commentary this week suggested the market may reprice the stock higher if the company continues stacking long-term lease agreements with additional hyperscalers, especially given its 4 GW pipeline. [9]
The logic is simple:
- AI demand keeps rising.
- Hyperscalers and “neo-cloud” providers need capacity sooner than typical construction schedules allow.
- A builder with power, sites, and faster build cycles can lock in long-term contracts — and each new deal can expand the company’s visible revenue base.
That’s also why some market observers have framed APLD as a long-duration “buy-and-hold” AI infrastructure play, arguing that the stock trades at a relatively modest multiple of projected operating profit if the company executes and reaches the scale implied by its pipeline. [10]
The financing question: building AI factories is capital-intensive
No matter how attractive the leases look, AI data centers are expensive to build — and the sector’s financial engineering matters.
Applied Digital’s current growth plan is built around multiple layers of funding, including:
- A private offering of $2.35 billion in senior secured notes. [11]
- A preferred equity financing arrangement that can reach up to $5 billion from Macquarie Asset Management, with commentary noting the company has already drawn funding from that facility. [12]
The upside of that structure is clear: it can accelerate buildouts without immediately relying on large-scale equity issuance. The downside is also clear: debt, preferred equity costs, and construction timing can quickly become the story if schedules slip.
What’s new on Dec. 15, 2025: power infrastructure is becoming the “AI trade”
Today’s biggest AI infrastructure headlines aren’t limited to data center developers. They’re increasingly about who owns the grid and generation capacity that makes AI possible.
Utilities are suddenly “AI infrastructure” targets
In a Reuters Breakingviews column published December 15, 2025, the AI-driven surge in electricity demand is framed as one of the forces turning utilities — historically viewed as dull income stocks — into strategic assets. The piece points to BlackRock’s reported $38 billion bid for AES as a symbol of this shift. [13]
The column ties the new wave of interest to three drivers:
- the green transition,
- higher grid investment needs,
- and the spike in power demand from AI.
It also cites estimates suggesting global data center power consumption could grow about 17% annually from 2022 to 2030, and that AI server farms could represent a meaningful share of U.S. electricity demand by 2030. [14]
For companies like Applied Digital, this matters because power availability is destiny. Even the best designed AI data center is just a concrete shell if the transmission upgrades arrive late — and the market is increasingly pricing that reality.
Europe adds another giant build: Amsterdam hyperscale campus
Also on December 15, Reuters reported that Pure Data Centres, backed by Oaktree Capital Management, unveiled a €1 billion ($1.17 billion) investment to build a hyperscale data center campus in Amsterdam — one of Europe’s more power-constrained but strategically important markets. [15]
Key details in the report underscore how “industrial” the AI data center business has become:
- The campus is expected to be leased to a single hyperscaler.
- Construction is scheduled to begin in January 2026, with phased completion from 2028.
- The facility is described as 78 MW, supplied by a private substation. [16]
The bigger signal: even with regulatory and grid constraints in parts of Europe, capital is still flowing to AI-grade facilities — especially when a tenant is lined up early.
Sustainability is lagging — even as electricity demand surges
A separate Dec. 15 update comes from a Conference Board survey distributed via PR Newswire. It found that while AI is driving a major surge in electricity demand, only 13% of surveyed sustainability leaders said environmental impact is a major consideration in their companies’ responsible AI strategies. [17]
The report also highlights what sustainability leaders say they’re most worried about:
- data center energy demand,
- energy consumption,
- and emissions tied to electricity use — with water use also on the list. [18]
This is relevant for AI infrastructure investors because the sector’s next phase may be shaped not only by demand, but also by:
- local opposition,
- permitting timelines,
- water and power constraints,
- and potential policy responses.
In other words, the AI data center boom is increasingly fighting a “social license to operate” battle — one that could influence timelines and costs.
The risk section investors can’t ignore: tenants, timing, and the cost of capital
Applied Digital’s story sits at the intersection of real estate, infrastructure, and high finance — and that intersection can cut both ways.
The two risks most often flagged in recent commentary are:
1) Construction and delivery risk
If projects slip, cash flows slip — but debt and preferred equity costs don’t pause. That dynamic is why timeline compression (and execution discipline) is so central to the bull case. [19]
2) Financing and tenant credit risk
A Reuters Breakingviews column earlier this month warned that the AI data center boom could be vulnerable if lenders reassess the creditworthiness of some data center tenants — particularly “neo-cloud” intermediaries that rent GPU capacity. The analysis used Applied Digital’s relationship with CoreWeave as an example of how financing assumptions can swing project economics. [20]
This doesn’t mean leases are bad — it means that as the market matures, who the tenant is and how the project is financed may matter just as much as headline megawatts.
A smaller but notable Dec. 15 datapoint: institutional positioning
Not all daily news is about new campuses and megadeals. Some of it is quieter.
A MarketBeat filing-focused report dated December 15, 2025 said B. Riley Wealth Advisors Inc. reported a new position in Applied Digital, citing a holding of 204,798 shares valued at roughly $2.06 million based on the firm’s filing disclosure. [21]
On its own, a single institutional filing doesn’t change the fundamentals — but it reflects how widely APLD has entered the “AI infrastructure” conversation across both retail and professional investor circles.
What to watch next for Applied Digital and the AI data center sector
Looking beyond today’s headlines, the next phase of the story is likely to revolve around a handful of measurable milestones:
- New lease announcements: additional hyperscaler contracts remain the clearest potential upside catalyst mentioned in recent market commentary. [22]
- Delivery and energization timelines: execution against Polaris Forge buildouts and other pipeline projects will drive confidence (or skepticism). [23]
- The power market: utility dealmaking, grid capex, and power procurement strategies are becoming inseparable from AI growth — and today’s Reuters Breakingviews piece suggests private capital is already adapting to that reality. [24]
- Sustainability and regulation: as the Conference Board survey indicates, sustainability planning may lag demand — which could invite more scrutiny and policy friction. [25]
- European capacity buildouts with pre-secured tenants: the Amsterdam project is a reminder that the hyperscaler footprint is global, and long-cycle projects are being set now for late-decade delivery. [26]
Bottom line: Applied Digital has become a symbol of a larger market transition. AI infrastructure is moving from an abstract software story to a physical-world race for power, land, cooling, financing, and speed — and the winners may be the companies that can deliver “energized” capacity reliably, not just announce ambitious pipelines.
References
1. www.nasdaq.com, 2. www.reuters.com, 3. www.nasdaq.com, 4. www.nasdaq.com, 5. www.nasdaq.com, 6. www.nasdaq.com, 7. www.nasdaq.com, 8. www.nasdaq.com, 9. www.nasdaq.com, 10. www.nasdaq.com, 11. www.nasdaq.com, 12. www.nasdaq.com, 13. www.reuters.com, 14. www.reuters.com, 15. www.reuters.com, 16. www.reuters.com, 17. www.prnewswire.com, 18. www.prnewswire.com, 19. www.nasdaq.com, 20. www.reuters.com, 21. www.marketbeat.com, 22. www.nasdaq.com, 23. www.nasdaq.com, 24. www.reuters.com, 25. www.prnewswire.com, 26. www.reuters.com
