Tuesday 15 September 2026
Evening EditionRSS

ZOTPAPER

News without the noise

ZOTPAPER
Business

Salesforce blames Anthropic Claude spending for dented profit margin guidance

Deputy CFO says token costs with LLM partner prevented margin lift; company now optimising model usage

edit
By LineZotpaper
Published
Read Time2 min
Salesforce has blamed its heavy spending on Anthropic's Claude AI model for a drop in its full-year operating margin guidance, with a senior finance executive telling investors last week that token costs were sufficient to manage expectations. The CRM giant's operating margin is forecast at 20.1 percent for the full year, down from 20.5 percent in the second quarter, as the company pours resources into AI development.

Speaking at the Deutsche Bank Technology Conference last week, Salesforce deputy CFO and head of finance Mike Spencer said the company's extensive use of Claude tokens in its R&D cycle had prevented it from raising margin guidance for the year. "Roughly about six months ago we unleashed Claude in our R&D cycle. It's part of the reason we didn't raise margin guidance on the year because we're covering some of the token spend that we've got going," Spencer said. "And the goal of that really was to, let's see what we could break. Let's see what kind of advancement our R&D teams can make on accelerating the product road map. Worst-case scenario, we would pull back."

Salesforce is now moving to what Spencer called "refinement mode" — selecting the most cost-effective AI model for each task rather than always using the most advanced one. "We're going into the zone of 'prescription model choice for task at hand,' and what that really means is you don't need to use the latest and greatest model for every single task you might want to do," he said. Spencer noted that the company is experimenting with models from multiple vendors, including OpenAI, Cursor, Claude, and X's Grok, each with different cost structures.

The spending with Anthropic is substantial. In May, Salesforce CEO Marc Benioff said the company expected to spend $300 million with the Claude maker in 2026. Salesforce is not alone in seeking to optimise AI spending: Cockroach Labs CEO Spencer Kimball told The Register in June that his company was turning to open-source models to control costs, while Gartner has warned that the cost of coding agents could soon exceed developer salaries in some regions.

§

Analysis

Why This Matters

  • Salesforce's decision to absorb significant AI token costs instead of raising margins signals that even well-resourced enterprises are feeling the financial weight of frontier AI models.
  • The company's shift to "prescription model choice" reflects a broader industry realisation that using the most powerful model for every task is economically unsustainable.
  • If AI spending continues to dent margins, it could push enterprises toward smaller, cheaper models — potentially reshaping the competitive dynamics of the LLM market.

Background

Salesforce has been a major enterprise customer of Anthropic, the maker of the Claude large language model. In May 2026, CEO Marc Benioff disclosed that the company expected to spend $300 million with Anthropic that year. The partnership is part of Salesforce's broader AI strategy to embed generative AI capabilities into its CRM platform. The company's operating margin guidance of 20.1 percent for fiscal 2026 comes after a second-quarter margin of 20.5 percent, a decline that the company directly attributes to Claude token costs.

Key Perspectives

Salesforce: The company views its AI spending as an investment in R&D acceleration, but is now prioritising cost efficiency through model selection and multi-vendor experimentation. Anthropic: As a primary beneficiary of Salesforce's spending, the company stands to gain from continued enterprise adoption, but may face pressure to offer more competitive pricing if customers begin optimising away from its most expensive models. Industry observers and analysts: Gartner's warning about coding agent costs exceeding developer salaries, and the example of Cockroach Labs shifting to open-source models, suggest that the current AI spending pattern may not be sustainable for many enterprises.

What to Watch

  • Whether Salesforce's full-year margin guidance is revised upward or downward in future quarters as the company refines its model usage.
  • How Anthropic responds to potential pressure from enterprise customers seeking volume discounts or tiered pricing for its models.
  • The extent to which other large enterprises publicly disclose AI spending impacts on their financials, potentially prompting a broader industry reckoning with token costs.

Sources

newspaper

Zotpaper

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.

Editorial ProcessSubscribe