From Mining Crypto to Manufacturing AI in North Dakota: A $7 Billion Pivot

Today, I'm looking at Applied Digital (APLD). It's the crypto miner turned AI data center that just announced a new data center in Harwood, ND (a few miles north of my house).

A couple of years ago, they were perpetually losing money in their crypto mining operation. Instead of folding, they realized their North Dakota data centers were perfect for AI.

Two quick stats:

• They secured a $7 billion contract with AI compute provider CoreWeave

• Their ND "geographic arbitrage" play saves 40-60% on operating costs

Here is some specific info on why North Dakota offers geographic arbitrage:

• North Dakota = ~220 days of free cooling is modeled to save $.04/kWh vs. TX/VA

• Power costs: (believed to be as low as) $0.03/kWh vs $0.10/kWh in Virginia

The uncomfortable truth: Applied Digital is basically a leveraged bet that AI compute demand will explode. They're building infrastructure 12-18 months before revenue arrives, funded by $688M in debt at a 108% debt-to-equity ratio.

If they're right, they own the picks and shovels for the AI gold rush.

With that, I'll see you tomorrow!

Nick

TL;DR

• Applied Digital transformed from a crypto mining company to a $4B AI infrastructure powerhouse by building specialized data centers in North Dakota

• Secured a $7 billion, 15-year contract with CoreWeave for their Ellendale, ND facility—validating their contrarian "AI factories in cornfields" strategy

• North Dakota advantages: 200+ days of free cooling, $0.03/kWh power costs, and $2.7B in savings versus traditional locations

• Breaking ground on $3B Polaris Forge 2 campus in ND (September 2025), positioning the state as America's AI infrastructure hub

• Key lesson: Geographic arbitrage still exists in tech—ND's climate and power create moats that Silicon Valley money can't replicate

The 30,000-Foot View

Applied Digital operates as the picks-and-shovels play for the AI gold rush, designing and leasing specialized data centers for artificial intelligence workloads. The company generates revenue through:

• Data center hosting: $102.4M (71% of revenue) - mainly crypto mining transitioning to AI

• HPC hosting: Ramping up with new AI-focused facilities

• Cloud services: $17.8M (12% of revenue) - being divested to focus on infrastructure

Key stats:

• Market cap: $4.0 billion

• TTM revenue: $144.2 million

• Gross margin: 29.6% (up from 10.4% YoY)

• Net loss: ($161.0 million)

• Employees: 150-205

• Industry: Technology Infrastructure/Data Centers

Company History

• 2001-2021: Existed as dormant shell company cycling through various identities

• March 2021: Wes Cummins takes control, pivots to Ethereum mining

• April 2022: IPO at $5/share, raises $36M just as crypto winter begins

• November 2022: Rebrands from Applied Blockchain to Applied Digital

• January 2024: Flagship Ellendale facility goes offline for 5 months due to transformer failures

• September 2024: NVIDIA invests $63.66M (in the open market), grants Elite Partner status

• January 2025: Macquarie commits up to $5B in expansion financing

• June 2025: Signs $7B CoreWeave contract, stock soars from $3 to $15+

• August 2025: Announces 280MW facility in Harwood, ND (just north of Fargo)

Show Me the Money

Stand-out financial features:

• CapEx exploded to $681.6M (5x YoY) betting on AI infrastructure

• Gross margins nearly tripled as company optimized operations

• Debt ballooned to $688M funding expansion before revenue

• Projects $1B annual operating income within 3-5 years (7x current revenue)

Financial Data

The N.O.O.B. Nine — Competitive Powers

The Nerd Out on Business Nine is made up of Hamliton Helmer's famous "7 Powers" of competitive advantage (Scale Economies, Network Economies, Counter-Positioning, Switching Costs, Branding, Cornered Resource, and Process Power) combined with two of my own (Data Flywheel and Distribution Advantage).

Average Score: 2.9/5 - ounter-positioning and specialized processes create a defensible niche, but lacks the scale advantages of established players.

Memorable Marketing

Applied Digital's B2B playbook focuses on technical buyers who care about specs over sizzle:

"$2.7 Billion Savings" White Paper (2025)

• Core idea: Quantified North Dakota cost advantages with hard numbers

• Channels: Direct sales tool, PR amplification

• Why it worked: Led with ROI, not features

• Result: Generated CoreWeave interest, validated counter-positioning

"AI Factory" Positioning

• Reframed data centers as manufacturing facilities for intelligence

• Channels: CEO podcast tour, industry panels

• Why it worked: Simple metaphor for complex infrastructure

Tactical Takeaways:

• Lead with concrete ROI calculations in B2B marketing

• Transform geographic disadvantages into strategic advantages

• Use founder-led content for technical credibility

• Skip traditional B2B tactics (trade shows) for white papers that sell

AI Uses & Opportunities

Current AI applications:

• Operating "AI Factories" supporting up to 50,000 NVIDIA H100 GPUs

• Liquid cooling achieving 15-30x traditional data center density

• Cloud services division offering GPU-as-a-Service (being divested)

Future AI potential:

• Multi-gigawatt pipeline including $3B Polaris Forge 2 campus

• NVIDIA Elite Partner status ensures GPU access during shortages

• Potential REIT conversion would unlock institutional capital

• Co-developing next-gen cooling with ABB for AI-specific workloads

Bumps in the Road

• Transformer failure: Ellendale facility offline for 5 months in 2024

• Securities lawsuits: Multiple class actions alleging profitability overstatements

• Debt burden: $688M debt with 108.6% debt-to-equity ratio

• Customer concentration: CoreWeave represents majority of future revenue

• Competition: Hyperscalers building own infrastructure, Equinix's $15B expansion

Your Swipe File

• Geographic arbitrage can be hard to beat: North Dakota's natural advantages (climate, power) create a hard-to-replicate economic cost moat

• Counter-position against what incumbents can't do: Legacy data centers can't handle AI heat loads as retrofitting costs more than their facilities are worth

• One whale beats 100 minnows in infrastructure: Their $7B CoreWeave deal validates focusing on massive contracts over diversification

• Transform bugs into features: Marketed ND location as strategic advantage, not limitation

• Don't "blindly build": Don't build hoping customers will come. secure anchor tenants before breaking ground on material infrastructure investments.