Digital Realty (DLR) is positioned as a direct beneficiary of the current AI training boom, as its large-scale data centers are in high demand for power-intensive computing. In the fiber optic space, Cogent Communications (CCOI) is a favored "picks and shovels" play, with analysts bullish on its strategy to repurpose its network for AI and hyperscale customers. A more speculative but high-potential theme involves Bitcoin miners like Core Scientific (CORZ) and Applied Digital (APLD) pivoting to AI infrastructure. These miners are leveraging their existing large-scale power contracts and land to serve the AI industry. This transition could lead to a significant re-rating of their stocks as they capture higher returns from AI data center services.

Digital Realty (DLR) & Equinix (EQIX)

• These are the two main publicly traded data center companies, structured as REITs (Real Estate Investment Trusts). They are essentially landlords who own the physical data center buildings.
• Equinix (EQIX) has historically been viewed as the superior company due to its interconnection-focused business model.
  • It acts as a central hub or "marketplace" where hundreds of different companies (networks, cloud providers, enterprises) connect their equipment directly to each other.
  • This creates a powerful network effect: everyone wants to be in an Equinix facility because everyone else is already there, making data exchange faster and more efficient.
  • This model has historically generated much higher returns on invested capital (ROIC), in the low to mid-teens, compared to more standard data centers.
• Digital Realty (DLR) has a larger focus on the hyperscale market.
  • This involves leasing massive, entire buildings to a single or a handful of large tenants, namely the hyperscalers (Amazon, Microsoft, Google, Meta).
  • Historically, this was a more competitive, lower-return business, with ROIC around 7-8%, not much higher than the cost of capital.
• The AI boom has dramatically shifted the landscape, benefiting Digital Realty more directly in the short term.
  • Training large AI models requires immense power and clustering thousands of GPUs together, which has created a massive demand for the large-footprint facilities that DLR specializes in.
  • This has created supply shortages, causing leasing prices to double and pushing the returns for new hyperscale projects into the low teens, nearly on par with Equinix's historical returns.
• Both stocks were down for the year at the time of the podcast.
  • Equinix announced a major spending (CapEx) plan that will slow its short-term cash flow growth (measured by AFFO, a key REIT metric) from ~10% down to the mid-single digits, which worried public market investors.
  • Digital Realty's stock was seen as being dragged down in sympathy with Equinix.

Takeaways

• Short-Term vs. Long-Term AI Play: The podcast positions Digital Realty (DLR) as a more direct beneficiary of the current AI training phase, which requires massive, power-hungry facilities. Equinix (EQIX) is positioned to benefit more from the future AI inference phase, when AI applications become widespread and need to be physically closer to end-users for speed (i.e., they become "latency-sensitive").
• Valuation Shift: The historical valuation premium that Equinix held over Digital Realty has disappeared. DLR now trades at a slight premium, reflecting the market's current excitement for the hyperscale AI opportunity.
• Investor Patience Required: The reason these stocks are underperforming the broader AI theme is due to the nature of public REITs. They must pay out dividends and are punished by investors for short-term slowdowns in cash flow, even if the spending is for long-term growth. This is a key difference from their faster-growing private competitors.

CoreWeave

• CoreWeave is a private company that operates a GPU cloud, meaning it leases data center buildings and fills them with high-end NVIDIA GPUs and other equipment specifically for AI workloads.
• The company started in 2017 as a cryptocurrency miner, which is how its management team gained expertise in operating large fleets of GPUs.
• It has experienced almost unbelievable growth:
  • 2022 Revenue: $16 million
  • 2025 Projected Revenue: $5.3 billion
• This growth requires astronomical capital spending (CapEx), projected to be $21 to $23 billion in the current year alone, putting it in the same league as companies like AT&T or Exxon Mobil.
• Microsoft is currently its largest customer, accounting for 70% of its business. It has also signed massive deals with OpenAI.
• The analyst believes valuing the company on standard metrics like EBITDA is the "wrong way to look at it" due to the massive CapEx. He prefers a "unit economics" approach, estimating that for every $1 of CapEx spent on equipment, CoreWeave generates 15 to 20 cents in net present value.

Takeaways

• Pure-Play on AI Infrastructure: CoreWeave represents a focused bet on the high-performance computing needed for AI. It is a key enabler for companies like Microsoft and OpenAI.
• Key Risks:
  • Competition: The big cloud providers (Amazon's AWS, Microsoft's Azure, Google's GCP) have "essentially limitless resources" and are expected to become more competitive in the GPU cloud space over time.
  • Customer Concentration: A heavy reliance on a few very large customers like Microsoft and OpenAI gives those customers significant pricing power.
  • Chip Supply: CoreWeave has benefited from a shortage of GPUs. As NVIDIA and others ramp up supply, it could put pressure on pricing and returns.

Fiber Optic Network Operators

• Fiber is the essential "plumbing" that connects data centers to each other and to the rest of the world. The massive increase in data generated by AI is driving significant demand for more fiber capacity.

Lumen Technologies (LUMN)

• Lumen owns the old Level 3 long-haul fiber network, which was famously built during the dot-com bubble with numerous empty "conduits" (pipes for fiber).
• For decades, this extra capacity was unused as the company's legacy business declined.
• The AI boom has made these empty conduits extremely valuable, as Lumen can quickly and cheaply install new fiber for hyperscalers. The company has signed $8 to $9 billion in new contracts related to this.
• Sentiment: The analyst is skeptical. While the AI-related contracts are a positive development, he believes the company's large, declining legacy business will continue to be a major drag on overall growth.

Cogent Communications (CCOI)

• Cogent is a smaller fiber operator that acquired the old Sprint fiber network.
• It is repurposing this network to offer a high-capacity product called "wavelengths," which is in high demand for AI and hyperscale customers.
• Sentiment: The analyst is bullish on Cogent. He believes the company has "a number of interesting advantages" that should allow it to successfully grow its wavelengths business and become a strong third competitor in the market. He notes that he has "been wrong the last couple of years" on the stock, but still finds it "pretty interesting."

Takeaways

• "Picks and Shovels" AI Play: Fiber optic networks are a less obvious but critical part of the AI infrastructure buildout. Companies that own extensive long-haul fiber networks are seeing a surge in demand.
• Turnaround Stories: Both LUMN and CCOI are presented as potential turnaround stories driven by AI demand. However, the analyst clearly favors Cogent's strategy and prospects over Lumen's, which he feels is weighed down by its legacy business.

Investment Theme: Bitcoin Miners Repurposing for AI

• A fascinating trend discussed is the convergence of crypto mining and AI infrastructure.
• Crypto miners have already secured two of the most important and scarce resources for AI training: massive amounts of power and large plots of land in remote areas.
• Because AI model training, like crypto mining, is power-hungry but not sensitive to network delays ("latency"), these remote locations are ideal.
• Several crypto mining companies are now pivoting their business model. They are upgrading their facilities to data center specifications and leasing their power and space to AI companies like CoreWeave, seeing a better return on their assets.
• Companies mentioned in this trend include Core Scientific (CORZ) (which CoreWeave recently attempted to acquire), Applied Digital (APLD), and Galaxy Digital.

Takeaways

• Hidden Data Center Capacity: This trend unlocks a new source of data center capacity and power that was previously dedicated to crypto mining.
• Potential Investment Angle: Investors could look at publicly traded crypto miners not just for their exposure to Bitcoin, but for their potential to transform into valuable AI infrastructure providers. This represents a potential re-rating opportunity for these stocks if they execute the pivot successfully.