Tokenomics: The New Divide Between AI Winners And Renters

AI token costs are falling, but enterprise AI budgets are being rapidly consumed by agentic workloads, reshaping software economics. Salesforce and Adobe are offsetting the 'token tax' through consumption-based or hybrid pricing models, which should help preserve or expand gross margins. The piece is constructive for software vendors adapting monetization to AI usage.

Analysis

This is a margin architecture story, not just a pricing story. The key second-order effect is that AI spend is becoming a variable cost embedded in product usage, which favors vendors with high customer retention and strong workflow lock-in because they can pass through inference costs without obvious sticker shock. That creates a wedge versus smaller SaaS names that may have to absorb token costs to defend seat-based pricing, quietly compressing gross margins over the next 2-4 quarters.

CRM and ADBE are positioned to benefit because both can reprice around value creation rather than raw seat counts, especially as agentic use cases drive more API calls and longer reasoning chains. The hidden winner is any platform layer that becomes the system of record for AI actions: once customers pay per outcome or per workflow, the vendor gains pricing power and better unit economics than pure model wrappers. The likely loser set includes point solutions and workflow software that lack proprietary data or distribution, since they will see AI usage expand faster than monetization.

The contrarian risk is that the market may be too early in rewarding “AI monetization” stories while underestimating customer resistance. If enterprise buyers start capping agentic usage, adding internal governance, or demanding budget-based contracts, consumption growth can decelerate even as per-token costs keep falling. That would turn the thesis from margin expansion into slower top-line conversion, with the danger showing up over months rather than days.

Near term, the catalyst path is earnings guidance: management teams that can quantify AI attach rates, usage-based expansion, and gross margin stabilization should see multiple support. The cleanest expression is to own the vendors that can bundle AI into existing workflows and avoid the names where AI is an incremental COGS line with weak pricing power. In that sense, this is a relative-value trade within software more than a sector-wide bullish call.

Key Decisions for Investors

• Long CRM vs. a basket of seat-based SaaS names that lack usage monetization power; hold through the next 1-2 earnings cycles. Best risk/reward is if CRM can show AI-led expansion without GMs rolling over, which should support both revenue and margin multiples.
• Long ADBE on 3-6 month horizon into guidance updates; use dips to add. The setup is attractive if management continues to demonstrate that AI features are accretive rather than dilutive to gross margin.
• Pair trade: long CRM / short a legacy workflow SaaS peer basket. Thesis: CRM can pass through token costs into value-based pricing, while weaker vendors face margin leakage or discounting pressure over the next 2-4 quarters.
• Avoid or underweight small-cap AI application names that cannot reprice usage. Their downside risk is a slow squeeze on gross margin as agentic workloads scale faster than monetization.