The AI System’s Unstable Orbit

The AI economy has been likened to a three-body problem, where no single force can dictate the outcome. This metaphor reveals the current state of the industry as a complex dance between competing interests, with multiple players vying for position.

Rapid growth in the AI economy is striking. Spending on AI accounts for an estimated 0.5 to 1 percent of all white-collar salaries in the United States, making it a significant portion of the economy. Enterprises are investing heavily in tokens, often referred to as “tokenmaxxing,” with no corresponding gain in productivity, as Palantir’s Alex Karp pointed out in July.

The competition between frontier labs is intense, with Meta and xAI fielding increasingly capable models alongside Anthropic, OpenAI, and Google. Meanwhile, Chinese companies like Zhipu and Moonshot are pushing the boundaries of open-weight models, offering capabilities at or near those of closed models at a fraction of the price. US open-weight models, led by Thinking Machines’ Inkling and Nvidia’s Nemotron 3, are also gaining credibility as domestic alternatives.

The timing mismatch between adoption and utility is a significant concern. Adoption has occurred much faster than prior technologies, while prices have remained high. This has created anxiety about returns on investment. However, this situation also means that the industry is ripe for disruption.

As competition drives down prices and promotes real differentiation between models, the shift toward a multi-model world will continue. US open-weight models are likely to become genuine alternatives to Chinese ones, winning real adoption as a result. They may even develop clearer business models, making it easier for customers to take long-term bets.

Frontier labs will go deeper into the product stack to widen their moats and sustain high margins. Application companies will follow suit, going deeper into the model stack to build moats of their own. This convergence is rational – software companies typically enjoy 70%+ gross margins while customers feel they get their money’s worth from the product.

Beneath the surface lies a fundamental question: who captures the value when AI pays off? Will it be the labs at the frontier, the open models nipping at their heels, or the applications that own the customer? The answer is far from clear, and the industry’s equilibrium remains unsettled. For now, we’re stuck in this three-body problem, watching as the players maneuver for position.

The noise around open vs. closed, China panic, and returns on investment will eventually subside. What will be left is a more nuanced understanding of the AI system – one that recognizes the complex interplay between competing interests and the multiple paths to value creation. As we navigate this uncharted territory, it’s clear that the true challenge lies not in whether AI pays off, but in who captures its benefits when it does.