The AI sector, particularly frontier model labs, is experiencing growth at a scale and speed never before seen in capitalism. Anthropic's explosive revenue growth, achieved with significantly less capital burn than competitors like OpenAI, highlights a new paradigm of capital efficiency and market capture.
The AI boom is fundamentally constrained by physical-world inputs: energy availability, data center construction (gated by zoning and regulation), and semiconductor fabrication. TSMC's inability to meet NVIDIA's full demand is acting as a governor on the industry's growth, preventing a potential bubble.
Economic value in the AI ecosystem is currently flowing to the foundational layers: energy providers, data center operators, chip manufacturers like NVIDIA, and the frontier model labs. In contrast, the application layer has been a net destroyer of value, as companies struggle to build sustainable businesses on top of the models.
AI and its underlying compute infrastructure are becoming central to geopolitical strategy. Events like the closure of the Strait of Hormuz can create a relative manufacturing advantage for the US via energy prices, while battlefield AI superiority is proving to be a decisive factor in modern warfare, as seen in Ukraine.
Leading AI labs are moving away from 'all-you-can-eat' subscription plans towards usage-based pricing. This shift, coupled with the ability to offer more powerful, unconstrained models to enterprise clients, is set to unlock exponential revenue growth and further widen the gap between frontier and open-source models.
Keep pulling the thread on Gavin Baker.