A 10x increase in compute may only yield a one-tier improvement in model performance. This appears inefficient but can be the difference between a useless "6-year-old" intelligence and a highly valuable "16-year-old" intelligence, unlocking entirely new economic applications.

AI is expected to create a new generation of "model busters": companies that grow so rapidly and for so long that they consistently shatter conventional financial forecasts. Like Apple post-iPhone, whose performance was underestimated by 3x, these AI firms will deliver value far exceeding any spreadsheet's predictions.

Historical tech cycles like the cloud and mobile demonstrate a consistent pattern: the application layer ultimately generates 5 to 10 times the value of the underlying infrastructure capital expenditure. With trillions being invested in AI infrastructure, future value creation at the application layer will be astronomically larger.

In the current market, AI companies see explosive growth through two primary vectors: attaching to the massive AI compute spend or directly replacing human labor. Companies merely using AI to improve an existing product without hitting one of these drivers risk being discounted as they lack a clear, exponential growth narrative.

Unlike COVID's growth, which had a hard population limit, AI's potential is tied to energy and computation, which have vast room to expand. However, its real-world application will manifest as a series of S-curves, as different technologies and industries hit temporary plateaus before the next breakthrough occurs.

The AI industry's exponential growth in consuming compute, electricity, and talent is unsustainable. By 2032, it will have absorbed most available slack from other industries. Further progress will require potentially un-fundable trillion-dollar training runs, creating a critical period for AGI development.

Most of the world's energy capacity build-out over the next decade was planned using old models, completely omitting the exponential power demands of AI. This creates a looming, unpriced-in bottleneck for AI infrastructure development that will require significant new investment and planning.

AI's computational needs are not just from initial training. They compound exponentially due to post-training (reinforcement learning) and inference (multi-step reasoning), creating a much larger demand profile than previously understood and driving a billion-X increase in compute.

Elad Gil argues that the total addressable market for AI companies is not limited to traditional seat-based software pricing. Instead, it encompasses the multi-trillion dollar human labor market that AI can augment or automate.