Technology 5 min read By Bethany Hadley
AI’s three-body problem: no single force can dictate the outcome
The AI economy in 2026 resembles a chaotic three-body system, with frontier labs, open-weight models, and application companies pulling against each other. No single force can dictate the final outcome, but broad trajectories are emerging for the second half of the year.
The artificial intelligence economy has entered a chaotic phase that mirrors the three-body problem in physics, according to a new analysis. Just as two celestial bodies orbit in stable paths until a third enters and destabilises the system, the AI market is now shaped by three powerful forces pulling against one another: closed-source frontier labs, open-weight models, and the application companies built on top of both. Each is large enough to bend the others’ trajectories, but none can dictate where the system finally settles.
The most dramatic shift has come from the frontier labs, led by Anthropic and OpenAI, which have seen unprecedented demand and extraordinary revenue growth. That momentum has diffused AI’s benefits more broadly through the economy, but it has also sharpened the focus on demonstrable return on investment and the search for cheaper alternatives. Spending on AI now runs to somewhere between 0.5 and 1 percent of all white-collar salaries in the United States, a scale that invites scrutiny.
That scrutiny arrived in July, when Palantir’s Alex Karp told CNBC that “something has gone completely wrong” with how the labs sell their product. Enterprises, he argued, are “tokenmaxxing” — spending furiously on tokens with no matching gain in productivity. Competition at the frontier has also intensified, with Meta’s Muse Spark 1.1 and xAI’s Grok 4.5 fielding increasingly capable models alongside Anthropic, OpenAI, and Google.
Meanwhile, open-weight models, particularly from Chinese companies, have improved dramatically. Zhipu’s GLM 5.2 and Moonshot’s Kimi K3 now perform at or near the frontier on several important benchmarks, while priced at a fraction of comparable closed models. That combination has created strong momentum for the open-weight ecosystem. US open-weight models are adding credibility of their own, led by Thinking Machines’ Inkling and Nvidia’s Nemotron 3, which offer a domestic alternative to the Chinese releases.
The reaction from AI application companies has been predictable: many leading players have ramped up efforts to build on top of open-weight models, targeting lower costs and greater control. Three-body systems are notoriously difficult to predict, but the analysis discerns several broad trajectories for the second half of 2026.
First, discomfort with frontier pricing will ease, partly because competition will push prices down and partly because returns on AI spending will begin to show. Much of today’s anxiety is a timing mismatch: adoption is running ahead of utility. For most prior technologies, like cars and cell phones, mass adoption followed declines in price. In the case of AI, adoption happened much faster, and the payoff will come as faster growth for some companies, cost savings for others, and eventually higher productivity across the economy.
Second, the shift toward a multi-model world will continue, driven by competition and, ideally, by real differentiation in what each model does best. Third, US open-weight models will become genuine alternatives to the Chinese ones and win real adoption as a result, helped by a clearer business model that makes it easier for customers to take longer-term bets.
Eventually, the distinction between open and closed is expected to diminish, as the frontier labs themselves support model personalisation for specific customer needs. In other ways, the players will converge: frontier labs going deeper into the product stack to widen their moats and sustain high margins, and application companies going deeper into the model stack to build moats of their own. That convergence is rational, as software companies typically enjoy gross margins above 70 percent while customers feel they get their money’s worth.
Given these moves and countermoves, the equilibrium remains unsettled. Much of today’s noise — the debate over open versus closed, China panic, and hand-wringing over returns — looks temporary. The genuinely interesting question is not whether AI pays off, but who captures the value when it does: the labs at the frontier, the open models nipping at their heels, or the applications that own the customer.
