AI's Three-Body Problem: no single force can dictate the outcome | Fortune
Ai's Shifting Landscape
The artificial‑intelligence sector has undergone a series of recent events that have heightened instability across the ecosystem. A primary driver of this turbulence is the surge in momentum experienced by frontier laboratories, with Anthropic at the forefront. These labs have recorded demand levels never seen before, translating into rapid revenue expansion. While this surge has helped spread AI capabilities more widely throughout the broader economy, it has also sharpened scrutiny on measurable return on investment and intensified the hunt for lower‑cost alternatives.
Spending and ROI Concerns
Current estimates place AI‑related expenditures at roughly 0.5 % to 1 % of all white‑collar wages in the United States. At that magnitude, analysts argue the spending deserves close examination. In July, Palantir chief executive Alex Karp told CNBC that “something has gone completely wrong” with the way frontier labs market their offerings, warning that enterprises are “tokenmaxxing”—pouring large sums into token usage without a corresponding boost in productivity. At the same time, rivalry among the leading labs has sharpened, as Meta (with Muse Spark 1.1), xAI (with Grok 4.5), Anthropic, OpenAI and Google each roll out increasingly sophisticated models.
Open Models Gain Traction
Progress among open‑weight models, especially those emerging from Chinese firms, has been striking. Zhipu’s GLM 5.2 and Moonshot’s Kimi K3 now rank at or near the frontier on several key benchmark tests, yet they are priced at a fraction of comparable closed‑source solutions. Their competitive pricing and near‑par performance have generated strong momentum for the open‑weight ecosystem. In the United States, domestic open models are also gaining credibility. Projects led by Thinking Machines’ Inkling and Nvidia’s Nemotron 3—both highly capable, though not yet fully frontier—provide a home‑grown alternative to the Chinese releases.
Application Companies Respond
A number of leading AI‑application firms have accelerated efforts to build on top of open‑weight models, seeking to lower costs while gaining greater control over their technology stacks. This shift reflects a broader industry pattern of aligning product development with more affordable, customizable model options.
Outlook for the Second Half of 2026
Looking ahead to the latter half of 2026, three broad trends appear likely. First, the unease surrounding frontier‑lab pricing should subside as competitive pressures push prices lower and as the tangible benefits of AI spending begin to materialize. Much of today’s anxiety stems from a timing mismatch: adoption has outpaced utility, unlike earlier technologies such as automobiles or mobile phones where mass uptake followed price declines. Second, the move toward a multi‑model environment will continue, driven by competition and genuine differentiation in each model’s strengths. Third, U.S. open‑weight offerings are expected to evolve into credible substitutes for Chinese models, achieving meaningful market adoption and presenting clearer business cases that enable customers to make longer‑term commitments.
Convergence and Value Capture
Both frontier labs and application developers are likely to deepen their involvement in each other’s stacks, expanding moats and preserving high margins—software firms typically enjoy gross margins above 70 % when customers perceive sufficient value. This convergence is rational: tighter integration can enhance product differentiation and profitability. The central issue is not merely whether AI will deliver returns, but which participants will capture the resulting value—whether the frontier laboratories, the open‑model challengers, or the application companies that maintain direct relationships with end users.
Source: fortune.com · 2026-08-22