The physical buildout behind AI

Key Concepts

• Hyperscalers: Large-scale cloud providers (e.g., Google, Amazon, Meta) driving massive capital expenditure (CapEx) in AI.
• AI Infrastructure "Three-Legged Stool": The framework of Data Centers, Compute (hardware/software), and Energy.
• Behind-the-Meter Power: On-site energy generation to bypass grid limitations.
• Small Nuke: Small Modular Reactors (SMRs) identified as a long-term energy solution for AI.
• CapEx Cycle: The $3–$5 trillion investment commitment by hyperscalers to build AI capacity.

---

1. The AI Infrastructure Framework

Shawn O’Hara, President of Pacer ETFs, argues that the "die is cast" regarding AI investment, with hyperscalers committed to spending $3–$5 trillion. He categorizes the AI ecosystem into three essential pillars:

• Infrastructure (Data Centers): The physical buildings housing AI operations.
• Compute (Hardware/Software): The internal components, networking, and intelligence layers.
• Energy: The power source required to run high-density computing environments.

2. Investment Strategy: The "Three-Pack" ETFs

O’Hara proposes a targeted investment approach using three specific Pacer ETFs to capture the end-to-end AI value chain:

• SRVR (Data Centers): Focuses on companies that own and manage data centers (e.g., Equinix, Digital Realty). It has evolved to include "behind-the-meter" power generation—companies like GE Vernova, Caterpillar, and Rolls-Royce—to address the power bottleneck.
• TRFK (Compute/Networking): Designed to capture everything inside the data center. This includes hardware, networking chips, software, cybersecurity (e.g., Snowflake), and specialized cooling systems (e.g., Trane, Johnson Controls).
• USAI (Energy): Targets the short-term energy solution: natural gas. This ETF holds pipelines and gas producers, which O’Hara identifies as the primary energy source for the current AI build-out phase.

3. Key Arguments and Market Perspectives

• The "Bottleneck" Thesis: O’Hara emphasizes that the primary constraints for AI are not just software, but physical space (data centers), cooling capacity, and, most critically, power.
• Profitable Intermediaries: A core investment argument is that while it remains uncertain if hyperscalers (Meta, Google, Amazon) will successfully monetize their AI operations, the companies supplying the infrastructure are already profitable and guaranteed to receive the committed CapEx.
• Beyond Semiconductors: O’Hara distinguishes his strategy from chip-only ETFs. He argues that the AI boom is a holistic hardware/software/cooling/energy story. While GPUs (Nvidia) and memory chips (Micron, Intel) are vital, the "traffic" of data requires a broader ecosystem of networking, cybersecurity, and thermal management.

4. Technical Terms and Definitions

• Behind-the-Meter: On-site power generation that does not rely on the public utility grid, essential for the high-energy demands of AI data centers.
• Hyperscalers: Companies that provide cloud computing services at a massive scale; they are the primary drivers of the current AI infrastructure spending cycle.
• Small Nuke: Refers to Small Modular Reactors (SMRs), which are viewed as the long-term, sustainable energy solution for the massive power requirements of future AI data centers.
• Cooling Systems: Critical infrastructure components required to maintain the high-performance environment of data centers; companies like Trane and Johnson Controls are highlighted as key players.

5. Synthesis and Conclusion

The AI boom is transitioning from a speculative phase to a massive capital expenditure cycle. O’Hara’s framework suggests that investors should look past the "Magnificent Seven" and focus on the "picks and shovels" of the AI revolution. By targeting the physical infrastructure (SRVR), the internal compute and cooling ecosystem (TRFK), and the immediate energy requirements (USAI), investors can gain exposure to the foundational layers of AI. The strategy prioritizes companies that are currently profitable and essential to the build-out, regardless of the long-term success of the AI models themselves.