Dylan Patel - Inside the Trillion-Dollar AI Buildout - Invest Like the Best with Patrick O'Shaughnessy Recap

Podcast: Invest Like the Best with Patrick O'Shaughnessy

Published: 2025-09-30

Duration: 1 hr 59 min

Summary

Dylan Patel discusses the intricate dynamics of the AI infrastructure buildout, emphasizing how companies like OpenAI, NVIDIA, and Oracle are intertwined in a strategic relationship that drives demand for compute resources. The conversation also covers the challenges and opportunities in the semiconductor supply chain and the implications for asset management.

What Happened

In this episode, Patrick O'Shaughnessy sits down with Dylan Patel, founder and CEO of Semi-Analysis, to delve into the intricacies of the trillion-dollar AI buildout. They discuss how the biggest tech companies, including OpenAI and NVIDIA, are engaged in a strategic game of resource allocation and compute demand. Dylan breaks down the relationships between these companies, emphasizing that OpenAI has an insatiable demand for compute, which is crucial for their growth and functionality in the AI space. He highlights how this demand necessitates a robust infrastructure, including the need for data centers and the associated costs that come with it.

The conversation shifts to the realities of the semiconductor supply chain, where Dylan notes the significant increases in costs, such as electrician wages doubling. He explains that we are still in the early stages of AI development, specifically referencing post-training and reinforcement learning as areas ripe for exploration. By analyzing these infrastructure realities, Dylan offers a framework for understanding where value will accrue in the AI stack, particularly as traditional SaaS economics struggle under AI's high cost of goods sold. This episode provides a comprehensive perspective on the physical realities that underpin the AI revolution, making it a must-listen for anyone interested in the future of technology and investment.

Key Insights

Key Questions Answered

What is the relationship between OpenAI, NVIDIA, and Oracle?

Dylan Patel explains that OpenAI pays Oracle, which in turn pays NVIDIA, creating a complex financial ecosystem where all three companies benefit from each other. The relationship is essential for OpenAI, which has a massive demand for compute resources necessary for training and running AI models. This interplay highlights the strategic dependencies among these tech giants as they navigate the competitive landscape of AI.

How does the demand for compute resources affect AI companies?

Dylan emphasizes that the compute demand precedes business growth. Companies like OpenAI require a robust infrastructure to train models and process inference efficiently. If they do not secure sufficient compute resources quickly, they risk being overshadowed by competitors, even if they have a significant user base. The compute resources are critical for unlocking new AI use cases, making this demand a central theme in the AI buildout.

What are the challenges in the semiconductor supply chain?

Patel points out that the semiconductor supply chain is facing significant challenges, including rising costs for labor, such as electrician wages doubling. These realities complicate the landscape for AI development, as companies must manage higher operational costs while striving to innovate. Understanding these challenges is essential for navigating the AI infrastructure buildout effectively.

What does Dylan Patel mean by 'the first innings of AI'?

Dylan refers to the current stage of AI development as being in the 'first innings,' suggesting that we are just beginning to explore the potential of post-training and reinforcement learning. This metaphor indicates that there is much more to come in terms of advancements and applications in AI, and firms must be prepared to capitalize on these emerging opportunities as they develop.

How are traditional SaaS economics impacted by AI?

Dylan discusses how traditional SaaS economics are breaking down under the high cost of goods sold associated with AI technologies. As companies like OpenAI and others scale their operations, they face unique cost structures that challenge conventional revenue models. This shift necessitates a reevaluation of how businesses position themselves within the AI stack, focusing on where value can be generated despite these economic pressures.