APLD: The Gigawatt Bet On The AI Data Center Buildout

The single most underappreciated bottleneck in the AI buildout is not chips. It is not capital. It is not engineers. It is power. Specifically, megawatts and gigawatts of energized, permitted, interconnected, deliverable electricity at sites with the land and water to support a hyperscale data center.

The companies that locked up that capacity years ago, before AI exploded, are sitting on extraordinarily valuable optionality. Applied Digital, ticker APLD, is one of those companies. They are not the biggest. They are not the most famous. They are one of the very few public small caps with real gigawatt-scale power in the queue, sites in the ground, and a leasing model that just got proven by a marquee customer.

This is the asymmetric bet inside the AI infrastructure thesis.

APLD owns and is developing some of the most strategically located, lowest cost, longest energized power capacity available to AI buildouts in North America, and the market is still partially pricing the company like its bitcoin mining heritage instead of its HPC future.

Applied Digital started as a bitcoin mining hosting operator. They built sites in North Dakota and Texas optimized for cheap, abundant power. The competitive edge was always the same. Lock up power at attractive long-term rates. Build the infrastructure to deliver it. Host high-density compute at competitive cost.

In the past two years they pivoted aggressively into HPC and AI hosting. The thesis was that the same sites optimized for bitcoin mining could be repurposed and expanded for AI workloads at higher margins and longer contract durations. The crown jewel of that pivot is the Ellendale, North Dakota campus, which is being scaled to gigawatt class for AI tenants.

The signature event was the CoreWeave lease deal. CoreWeave signed a multi-year lease for a meaningful chunk of APLD's North Dakota capacity. That deal is more than the revenue. It is the proof of concept. A first-tier AI cloud customer signed a long-term, take-or-pay style commitment for purpose-built infrastructure. That validates the entire pivot.

Pillar one: power is the constraint. Every conversation about AI capacity in 2026 starts and ends with power. The US grid is not expanding fast enough to meet projected AI demand. Hyperscalers are paying premiums for sites with energized capacity ready to deploy. APLD has that capacity in the queue at meaningful scale.

Pillar two: leasing model converts capex into recurring revenue. A take-or-pay lease to a credit-worthy AI tenant transforms APLD from a speculative buildout into a contracted cash flow stream. That is how the multiple expands. Bitcoin mining hosting earns one multiple. Long-term hyperscaler-style data center leases earn another. The market is in the middle of repricing.

Pillar three: site advantages. Ellendale and the Texas sites benefit from low industrial power costs, cool climate, available water, and friendly regulatory environments. Those are not easy to replicate. Try permitting a new gigawatt-scale interconnect in California or the Northeast and you will appreciate the value.

Pillar four: expansion runway. APLD has additional sites and additional capacity moving through the development pipeline. Each successful customer signing adds bookings, validates the model, and unlocks the next round of financing.

Lease bookings in megawatts. Revenue mix by segment, splitting cloud services, HPC hosting, and crypto hosting. Realized lease rates per megawatt-month. Energized capacity versus contracted capacity. Net debt and liquidity. Capex versus committed bookings.

The metric I watch hardest is the spread between cost of power and realized lease pricing. As long as that spread holds, the model works. If hyperscalers ever decide to walk away from the colocation model and bring everything fully in-house, the spread compresses. So far they are doing the opposite. They are taking more leased capacity than expected.

The first risk is customer concentration. CoreWeave is a meaningful tenant. Anything that pressures CoreWeave's own customer base ultimately pressures APLD.

The second risk is execution. Building gigawatt-class campuses on schedule is hard. Delays cost money and risk lease penalty provisions.

The third risk is capital structure. APLD operates with a mix of debt, equity, and project financing. The wrong financing decision under stress is a real risk.

The fourth risk is power cost. If the underlying grid costs surge, the spread compresses. APLD has structured pricing to mitigate this but it is not zero risk.

The fifth risk is competition from larger data center operators. Equinix, Digital Realty, Iron Mountain, and a wave of private operators are all chasing the same AI infrastructure dollars. APLD is competing on cost, site quality, and execution speed. They cannot afford to slip.

This is a higher-volatility name than the data center REITs. The right size reflects that. A 1% to 3% position in a diversified portfolio gives meaningful exposure to the asymmetric upside without putting the broader account at risk if the buildout stumbles.

I would not own this as my only data center exposure. I pair it with a more conservative name to balance the risk.

Equinix and Digital Realty are slow growers with predictable cash flows and big dividends. APLD is the asymmetric growth play in the same end market. The thesis is not that APLD becomes Equinix. The thesis is that APLD's gigawatt buildouts come online, get leased, and the market rerates the company from a hosting business to a hyperscale infrastructure platform.

Different risk profile. Different return profile. Same underlying demand wave.

Read APLD's most recent earnings release and the CoreWeave lease announcement. Look at the project pipeline slide. Look at the energized megawatt schedule. Now overlay it against your view of AI capex over the next three years. If you believe the AI capex wave continues, the math is straightforward. If you do not, this stock is not for you. Either answer is fine. Make sure it is your answer.

Read next: NBIS: The Cleanest AI Cloud Pure Play | CEG: Why Nuclear Is The Trade Of The Decade

*⚠️ Important Disclaimer: MentorSurge is not a financial advisor. This post is for educational and entertainment purposes only. Nothing on this site constitutes financial, investment, or trading advice. APLD is a small cap data center stock with elevated business and execution risk. Trading involves substantial risk of loss. Always do your own research and consult a licensed professional.*