It is currently impossible to predict whether model providers or application-layer companies will capture the most value. The outcome hinges on the level of competition between frontier models, which will determine token prices and, consequently, the profitability of the entire ecosystem built on top.
Despite AI's limited adoption (<5%) in the broader economy, leading model companies are already adding more monthly revenue than established giants like Meta, Google, or Microsoft. This signals that the ultimate market size for AI will be extraordinarily large, potentially consuming 10% of Fortune 500 profits.
The benchmark for a top-tier venture capital exit has dramatically accelerated, jumping from $10 billion in 2022 to $32 billion in mid-2024. This rapid inflation redefines the scale of success and the magnitude of potential outcomes in the current tech cycle, driven by AI.
Contrary to expectations, even cutting-edge companies are not yet using AI to automate internal operations. Their best talent and resources are focused on the larger prize of building new AI-driven products, leaving internal efficiency as a latent, uncaptured opportunity for now.
Unlike typical tech bubbles characterized by excess supply, the current AI boom is severely constrained by shortages in compute, power, and data centers. This fundamental supply-side bottleneck makes a speculative bubble less likely in the short term, as overinvestment cannot easily flood the market.
In venture capital, a portfolio with no failed investments is a sign of a flawed, risk-averse strategy. The goal isn't to avoid losses but to back the market leader in a potentially huge category. Losing money on the leader if the entire category fails is an acceptable and expected outcome.
The current generation of AI founders operates with a fundamentally different ethos. They build extremely lean, aggressive teams that work constantly and leverage advanced AI tools like agent swarms from the start, a stark contrast to the less efficient, headcount-driven growth of the last decade.