Jaluria Software Download 37: Token Optimization; Innovation Drives LT AI Benefits; Citizen Devs

研报英文原文证据摘录

Jaluria Software Download 37: Token Optimization; Innovation Drives LT AI Benefits; Citizen Devs

RBC Capital Markets, LLC

Rishi Jaluria (Analyst)

(415) 633-8798,

rishi.jaluria@rbccm.com

Max Persico, CFA (Senior

Associate)

(646) 618-6895,

max.persico@rbccm.com

June 22, 2026 Matthew Hedberg (Head of

Global TIMT Research)

(612) 313-1293, Jaluria Software Download 37: Token Optimization; matthew.hedberg@rbccm.comRESEARCH Innovation Drives LT AI Benefits; Citizen Devs

Our view: In this series, we offer a collection of thoughts or insights in software that are top of mind,

primarily AI-focused, and (we believe) relevant for investors but may not (yet) deserve a full note. Please

see Part 1... Part 34, Part 35, Part 36 (detailed index on page 2). We publish these notes on an irregular

cadence. We welcome any investor feedback, especially in terms of making this note more useful.

Token optimization: the straightforward bull case for software (but only for so long). TokenmaxxingEQUITY as a status game is collapsing; token consumption as a workload is exploding. As we wrote about in

our previous Jaluria Software Download (see here), whereas high token burn was once a productivity

proxy, enterprises now focus on optimization and no longer treat raw token usage as a measure of AI

leadership (witness Meta taking down its token leaderboard and companies like Uber, ServiceNow, and

Microsoft emphasizing reduced token burn). One of the straightforward ways to reduce token burn is by

using the most efficient model for the right job, both across vendors, as well as across model generations

(e.g. leveraging Claude Sonnet and GPT 5.3, older models for repeated tasks, open source Qwen and

DeepSeek for routine workloads, and only using Claude Opus and GPT 5.5 for the most complex tasks).