The Hidden AI Trade: Land, Power, & Compute | With Mike Alfred

Key Concepts

• Infrastructure-First Investing: Prioritizing physical assets (land, power, electrical infrastructure) over software or purely digital assets.
• Compute as a Commodity: The view that high-performance computing (HPC) power is the modern equivalent of scarce industrial resources.
• Vertical Integration: The strategy of controlling the entire stack—from physical data center development to the deployment of GPUs and cloud services.
• Neo-Cloud: Emerging cloud service providers that focus on high-performance AI compute rather than traditional enterprise storage/hosting.
• Asymmetric Risk/Reward: An investment strategy focusing on limited downside (based on liquidation value of physical assets) and high upside (based on future demand for AI compute).

---

1. Market Overview and Macro Perspective

Mike Alfred characterizes the current market as a "compressed cycle" where AI has been the primary driver of growth in the S&P 500, while other sectors (healthcare, staples) have lagged.

• The "Noise" Factor: Investors have faced constant warnings of recessions, banking crises (Silicon Valley Bank), and geopolitical tensions (Iran) over the last three years. Despite this, assets tied to AI and Bitcoin have consistently trended higher.
• Idiosyncratic Investing: Alfred advocates for a concentrated, theme-based approach, buying assets when they are out of favor and holding them through market volatility.

2. The Evolution of Bitcoin Miners to AI Infrastructure

Alfred argues that the term "Bitcoin miner" was a misnomer from the start. Many firms were actually industrial infrastructure developers who recognized that the real scarcity was not the digital asset itself, but the physical capacity to host compute.

• Strategic Location: Early developers (like those in West Texas and Oklahoma) focused on areas with excess renewable power, defying the traditional industry preference for Northern Virginia.
• Convergence: The infrastructure built for Bitcoin mining—high-uptime, low-cost energy, and massive electrical capacity—is now being repurposed for AI inference, synthetic biology, and high-frequency trading.
• The Choke Point: The primary constraint in the AI boom is not the availability of Nvidia chips, but the ability to "spin up" heavy-duty infrastructure to house and power them.

3. Investment Methodology and Mental Models

Alfred’s investment approach is tactile and long-term, often involving board-level participation.

• Margin of Safety: He evaluates investments by calculating the liquidation value of the physical assets (land, power, interconnects). In his analysis of companies like Iron (IREN) and Cipher (CIFR) in 2022, he determined that the underlying physical assets were worth significantly more than the market cap at the time.
• Asymmetry: He seeks opportunities where the downside is protected by tangible assets, while the upside is tied to the massive, long-term demand for AI compute.
• Board Constraints: Serving on a board limits liquidity due to SEC regulations (Form 4 filings) and blackout windows, which forces a "buy and hold" mentality that aligns with his long-term thesis.

4. Business Models: Colocation vs. Vertical Integration

Alfred distinguishes between two primary models in the data center space:

• Colocation (Asset-Light): Acting as a landlord, building the site and leasing space to others. This is less capital-intensive but leaves the provider vulnerable to the economics of the tenant.
• Vertical Integration: Controlling the entire stack—developing the site, owning the power, and deploying the compute (GPUs) to provide cloud services directly to customers.
• Case Study (SpaceX/xAI): Alfred highlights how SpaceX pivoted their "Colossus" data center from a proprietary Grok-focused facility to a "neo-cloud" provider for companies like Anthropic and Google, demonstrating the flexibility of owning the infrastructure.

5. Future Outlook and Synthesis

• Long-term Cycles: Alfred compares the current AI boom to the 1999 internet bubble. He expects periods of expansion followed by macro-driven contractions, but believes the underlying trend is a multi-decade "mega-trend."
• Actionable Insight: He suggests that investors should view significant drawdowns (50%+) in high-quality infrastructure plays as buying opportunities rather than signs of a "popped bubble."
• Conclusion: The winners in the next decade will be those who can "wield the most compute at scale." Companies that successfully vertically integrate their infrastructure and cloud services will be the most difficult to compete against.

"The reality is like I've been hearing that [it's a bubble] for 3 years, and yet everything that I own at least is higher. So I'm just going to continue kind of focused on what I'm focused on." — Mike Alfred