Data Center Buildout Going Too Far: Bokeh Capital’s Forrest

Key Concepts

• Data Centers: Physical facilities that house computing infrastructure, including servers, storage, and networking equipment.
• Large Language Models (LLMs): A type of artificial intelligence model trained on vast amounts of text data to understand and generate human-like language.
• Brute Force Method: A computationally intensive approach to training LLMs, requiring significant processing power and energy.
• AI Innovation: The development of new and more efficient methods for artificial intelligence, potentially reducing reliance on current resource-intensive approaches.
• Semiconductor Companies: Companies that design and manufacture microchips, such as NVIDIA, AMD, Micron, and Intel, which are crucial for AI development.
• Valuation: The process of determining the current worth of an asset or company, often used by investors to assess investment opportunities.

Data Center Investment and the "Too Much, Too Soon" Perspective

The discussion centers on the significant investment flowing into data centers, exemplified by Anthropic's $50 billion commitment and Matter's focus in Wisconsin. However, a Wall Street perspective suggests this might be "too much, too soon." This skepticism stems from the inherent challenges of physical infrastructure development, which contrasts with the virtual world often inhabited by Silicon Valley.

Key Points:

• Physical World Challenges: Data centers require substantial power, specialized construction, and uniformity, presenting logistical hurdles.
• Capacity Issues: Reports indicate some established data centers are already struggling to sell their capacity, suggesting a potential oversupply or miscalculation of demand.
• Skepticism on Necessity: A core argument is that the current scale of data center build-out is not necessarily required. The belief is that human ingenuity will lead to more efficient AI training methods, moving away from the "brute force" approach.

The Innovation Argument: Moving Beyond Brute Force

The central thesis presented is that human intelligence will find innovative solutions to current AI challenges, rather than relying solely on massive physical infrastructure. This perspective questions the current paradigm of training LLMs.

Key Arguments:

• Innovation Over Infrastructure: The belief is that humans will "work around" the current brute-force method of training LLMs.
• AI as Presented vs. Future AI: While acknowledging the reality and future potential of AI, there's a distinction made between AI as it is currently presented (resource-intensive) and what it could become through innovation.
• "Thinking Outside the Box": The example of China's DeepSeek product is cited as an instance of innovative thinking in AI development, contrasting with the U.S. approach of potentially dedicating vast physical resources (hyperbolically, "Texas be one big data center").

Investment Strategy in the Face of Uncertainty

For investors and fiduciaries, the question arises of how to invest to support this thesis of innovation over massive infrastructure.

Investment Approach:

• Diversified Portfolio: The strategy involves holding a diversified portfolio of semiconductor companies, including NVIDIA, AMD, Micron, and Intel.
• Belief in AI, Not Current Methods: The investment is predicated on the belief that AI is real and will be significant, but the deployment of capital will be different from current expectations.
• Instrumental Chips: Even if the brute-force method is abandoned, the chips produced by these companies will remain instrumental in future, more efficient AI development.

The AMD Valuation Debate

A specific point of contention is the valuation of AMD. The current market perception is that AMD's valuation is priced for an enormous data center boom.

Key Questions and Perspectives:

• AMD's Current Valuation: Is AMD's current price reflective of an enormous data center build-out, or does it account for a less significant build-out?
• Supporting the Thesis: The argument is made that AMD's chips will be crucial even if the brute-force method is superseded, as they will be instrumental in the new, more efficient AI paradigms. This suggests that AMD's value is not solely tied to the current data center build-out trend.

Conclusion: A Shift in AI Development and Investment

The core takeaway is a call for a re-evaluation of the current trajectory of AI development, particularly concerning the massive investment in data centers. The argument is that innovation will lead to more efficient AI solutions, reducing the need for the current scale of physical infrastructure. Investors should therefore consider a diversified approach, recognizing that the future of AI will likely involve different methodologies, but still require the foundational technology provided by semiconductor companies. The focus should be on supporting the evolution of AI rather than solely on the current, resource-intensive build-out.