The Infrastructure Thesis

AI systems continue to grow in scale. Models are larger, compute clusters are denser, and data centers are expanding rapidly. As this expansion continues, a physical constraint becomes increasingly important. Heat. Every unit of electricity consumed by a processor eventually becomes heat. As compute density rises, thermal output rises with it. This creates a new engineering challenge. High-performance GPUs generate enormous thermal loads. Thousands of them operating together produce heat levels that traditional cooling systems struggle to handle. For years, most data centers relied on air cooling. Cold air is pushed through server racks. Fans move heat away from processors. Hot air is then removed from the facility. But as AI clusters become more dense, air cooling begins to approach its limits. Moving enough air to remove massive heat loads requires enormous energy. Air is also relatively inefficient at transferring heat compared to liquids. This is why many AI operators are turning t...

I nfrastructure Thesis Series [01] AI Doesn’t Run on Code. It Runs on Electricity [02] The Real AI Bottleneck Isn’t Chips. It’s Power [03] Why Data Centers Are Becoming the New Factories [04] The AI Boom Is an Infrastructure Story [05] NVIDIA Was Phase One. Here’s Phase Two. [06] AI Winners Will Look Like Boring Industrial Stocks [07] The AI Supply Chain Is Longer Than You Think [08]AI Spending Is Shifting Beyond GPUs [09] Computing Is a Deployment Problem [10]Data Centers Are Creating a New Real Estate Cycle  [11] The Quiet Growth Engine: Server Racks and Cooling [12] AI at Scale Is an Energy Story  AI at Scale Is an Energy Story AI expansion is often described as a software revolution. Narratives move quickly. Markets respond to stories. Investors react to headlines. Technology cycles often begin with excitement about new capabilities. AI is no exception. In the early stage of the AI boom, the dominant narrative focused on soft...

I nfrastructure Thesis Series [01] AI Doesn’t Run on Code. It Runs on Electricity [02] The Real AI Bottleneck Isn’t Chips. It’s Power [03] Why Data Centers Are Becoming the New Factories [04] The AI Boom Is an Infrastructure Story [05] NVIDIA Was Phase One. Here’s Phase Two. [06] AI Winners Will Look Like Boring Industrial Stocks [07] The AI Supply Chain Is Longer Than You Think [08]AI Spending Is Shifting Beyond GPUs [09] Computing Is a Deployment Problem [10]Data Centers Are Creating a New Real Estate Cycle  [11] The Quiet Growth Engine: Server Racks and Cooling [12] AI at Scale Is an Energy Story  This series explaines why AI is becoming on infrastructure industry driven by power, data centers, and capital investment. In the early phase of the AI cycle, capital concentrated around GPUs. Compute was the primary constraint. Training large models required massive parallel processing power. Whoever controlled the compute layer controlled the trade. That...