AI Research Open Models and the Impact on the AI Trade

研报英文原文证据摘录

AI Research Open Models and the Impact on the AI Trade

Global Research

21 July 2026ab

AI Research Equities

AmericasOpen Models and the Impact on the AI Trade

Semiconductors

Karl Keirstead

Summary Analyst

The twin AI trends of token optimization and a share shift to cheaper open models karl.keirstead@ubs.com

(Chinese or otherwise) are dominating our investor discussions over the last several +1-310-734 2455

weeks, as investors judge whether and how these trends will impact AI spending and Timothy Arcuri

specifically chip/hardware suppliers, the hyperscalers (AWS/Azure) and the software Analyst

sector. Last week, we hosted many of the highest-profile AI-native firms on stage at our timothy.arcuri@ubs.com

5th Annual UBS Private AI and Software event in Menlo Park, and we addressed these +1-415-352 5676

subjects in detail. In this note we offer their unvarnished feedback and update our views Dean Marriott

of the impacts across the AI trade. Associate Analyst

dean.marriott@ubs.com

What We Heard +1-212-713 2115

Several themes stood out: 1. The revenue growth rates of these AI-natives have been

extraordinary. This signals that enterprises ARE indeed leaning into AI and not just in the

coding arena. 2. Token spend optimization is real. When we first weighed-in on this

subject last month, we concluded that AI compute/token spend had become a real

concern for ~60% of organizations. That figure now feels higher. 3. Over the last few

months we have quietly passed into a multi-model world, in which frontier labs will

coexist and compete with a plethora of other AI models, some closed-source, many

open-sourced and/or custom-built for companies and use cases. 4. Model routing –

dropping down to cheaper more token-efficient AI models where appropriate – was