UNTHINKABLE Risk. UNTHINKABLE Reward. The 4 AI Stocks That Could Make Millionaires

Key Concepts

• AI Data Center Landlords: Companies building large-scale physical infrastructure (data centers) to house the compute capacity required by hyperscalers (Microsoft, Amazon, Meta) for Large Language Models (LLMs).
• Hyperscalers: Large cloud service providers that drive the demand for AI infrastructure.
• Vertical Integration: A business model where a company controls multiple stages of the supply chain (e.g., power, GPU capacity, cloud services, and inference tools).
• Neocloud: A modern cloud platform architecture optimized for AI workloads.
• Capital Expenditure (CapEx) Risk: The financial risk associated with heavy debt-financed construction before achieving profitability.
• Institutional vs. Retail Ownership: The balance of professional investment firms versus individual investors in a stock.
• Physical AI: The integration of AI into physical systems, such as autonomous vehicles and humanoid robots, which is expected to drive future data center demand.

---

1. Overview of the AI Infrastructure Buildout

The current market is witnessing a massive capital investment cycle as companies race to build "the mall" (data centers) for AI. Hyperscalers are committing significant capital to secure space for their AI models. The primary challenge for these "landlord" companies is that they are currently in a high-spending phase, often relying on debt, and are not yet profitable. The central investment thesis is that demand for compute space will continue to outpace supply for the foreseeable future.

2. Risk-Ranked List of Data Center Builders

Marketbeat analyst Chris Marotch categorized four key players based on their risk profiles (1 = lowest, 10 = highest):

Nebius (NBIS) – Risk Level: 4 (Least Risky)

• Profile: A "full-stack" AI infrastructure company and vertical integrator.
• Strengths: Strong balance sheet (more cash than debt); diversified customer base (not solely reliant on one hyperscaler); vertically integrated model.
• Performance: Up ~165% year-to-date.
• Growth: Expanding from 2 data centers (2024) to a planned 16 by the end of the current year.

Iris Energy (IREN) – Risk Level: 5–6

• Profile: A former Bitcoin miner that successfully pivoted to AI cloud infrastructure.
• Strengths: Capital base from previous Bitcoin operations; significant contracts ($9.7B with Microsoft, $3.4B with Nvidia).
• Institutional Interest: Institutional buying outweighs selling by 3:1; institutional ownership is currently ~41%, suggesting room for retail growth.
• Sustainability: Operates using renewable energy (hydropower in Canada, ERCOT grid in Texas).

CoreWeave – Risk Level: 7

• Profile: Considered the "bellwether" of the sector; a pure-play AI cloud landlord.
• Strengths: Deep strategic partnership with Nvidia (Nvidia owns ~13% stake and acts as a backstop for unused capacity).
• Risks: Fundamental concerns regarding profitability; negative profit margins (-23%) despite high revenue; low interest coverage ratios.
• Scale: Rapid expansion from 3 data centers (2022) to 43 currently.

Applied Digital (APLD) – Risk Level: 10 (Highest Risk)

• Profile: Infrastructure layer provider focused on physical power and building sites.
• Risks: Compounded risk; 69% of its $16B in contracted revenue is tied directly to CoreWeave. If CoreWeave struggles, Applied Digital’s revenue visibility is compromised.
• Performance: Up ~550% over the last year.
• Status: Currently a three-campus operation with projects in North Dakota.

---

3. Methodologies and Market Dynamics

• The "Race" for Top-Line Growth: Investors are advised to prioritize "top-line" (revenue) growth over immediate profitability. The goal for these companies is to scale revenue quickly enough to cover debt service before capital runs dry.
• Profitability Timeline: Analysts estimate that adjusted EPS profitability may arrive by late 2027 or early 2028, with GAAP profitability likely occurring later.
• Institutional On-Ramp: Companies with lower institutional ownership (like Iris Energy) are highlighted as having more potential for price appreciation as professional capital enters the space.

---

4. Key Arguments and Perspectives

• Demand Validation: The "bullish sentiment" is supported by the fact that hyperscalers (Microsoft, Meta, Amazon) are confirming the demand by signing massive, long-term contracts with these landlords.
• The "Space" Wildcard: While some speculate about moving data centers to space, the consensus is that this is a long-term "moonshot" that will not impact the immediate, urgent need for ground-based infrastructure in the next 2–5 years.
• Investment Strategy: Marotch suggests that for stocks with aggressive, rapid moves (like Nebius), investors should consider "trimming" positions to lock in gains rather than exiting entirely, anticipating potential pullbacks before the next leg up.

---

5. Synthesis and Conclusion

The AI data center buildout is in its early stages, characterized by massive capital expenditure and high-stakes competition. While the demand from hyperscalers is real and growing, the primary risk remains the reliance on debt and the long timeline to profitability. Investors are encouraged to monitor top-line revenue growth and the physical completion of data center projects. The sector offers significant growth potential, particularly as "Physical AI" (robotics/autonomous systems) creates a new wave of demand beyond current LLM requirements.