Artificial intelligence competition is undergoing a fundamental shift, a change whose impact on tech stock valuations may far exceed what the market currently prices in.
Rich Privorotsky, head of Goldman Sachs' One-Delta trading desk, notes in a recent report that the AI sector is caught between two opposing forces. On one side is broader adoption and rising demand for computing power; on the other is intensifying token deflation, uncertain monetization prospects, and persistent stock supply expansion. He believes market focus is increasingly shifting towards the latter.
Recent experiments from OpenRouter show that a combination of several low-cost open-source models can match the performance of top-tier proprietary models at roughly half the cost. If validated, this suggests the competitive center of gravity in AI is accelerating its shift from a few elite labs towards model orchestration and the open-source ecosystem. The potential unraveling of AI pricing power is set to shape the market's core narrative this week.
The Dual-Edged Sword of Affordable AI
In his report, Privorotsky cites the OpenRouter experiment results, indicating that a model ensemble of Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro comprehensively outperformed standalone GPT-5.5 and Opus 4.8 in benchmarks. It also narrowed the performance gap to within 1% of Fable 5 at approximately half the cost.
He characterizes this trend as "a direction the market has consistently underestimated." The logic of the AI intelligence race is pivoting from "who has the single strongest model" to "who can most effectively orchestrate multiple models," thereby increasing the weight of the open-source ecosystem.
This shift carries clear dual implications. The bullish case is that falling costs and lower access barriers should ultimately drive a simultaneous expansion in token consumption and compute demand. The bearish case is that this trend accelerates token deflation and fundamentally challenges the sustainability of existing model economics.
Privorotsky distills the core contradiction into a single question: "Does lower intelligence cost create more demand, or does it destroy more pricing power?" The answer to this question directly relates to the justification for the AI sector's current multi-trillion dollar market valuation.
AI Sector Confronts Dual Supply and Demand Pressures
On the demand side, Privorotsky points out that the U.S. government's decision to restrict access to Claude Fable indicates AI competition may be entering a phase of geopolitical control, providing a degree of policy protection for leading proprietary models.
However, supply-side pressures are more direct. The report notes expectations for significant stock issuance in the second half of the year. Furthermore, several hyperscale cloud companies have recently shown clear underperformance relative to the broader market, drawing increased investor attention to their potential financing needs.
Privorotsky believes the two forces currently facing the AI sector—broader application adoption and rising compute demand versus token deflation, monetization concerns, and stock supply expansion—are seeing the latter gradually gain the upper hand. This marks a notable departure from the previously widespread market narrative of "infinite compute demand expansion."
Technical and Sentiment Backdrop: Momentum Persists, But Fuel Partly Spent
From a technical perspective, the backdrop for risk assets this week is broadly supportive. Historical patterns show a tendency for markets to drift higher ahead of Federal Reserve meetings; this week coincides with an options expiration while markets are at elevated levels, potentially generating mechanical buying demand; and the Juneteenth holiday shortens the trading week, reducing liquidity.
However, several of the firm's sentiment indicators—including Bull/Bear, Fear/Greed, and NAAIM—show sentiment is in a relatively subdued state, not displaying signs of overheating.
Privorotsky also highlights that data from the firm's prime brokerage book shows a significant level of short covering in the U.S. last week, particularly concentrated in macro products, implying some upward momentum has already been consumed. He suggests that if energy prices and the long end of interest rates see further movement, the most immediate market implication would be a momentum rotation—with lagging sectors like consumer discretionary outperforming, while AI and tech sectors face periodic pressure.
Comments