Salesforce freezes hiring, buys Anthropic tokens

Salesforce is pairing a hiring pause with a large AI usage bill. Marc Benioff, the company’s chief executive, said on the All-In podcast published May 16 that Salesforce expects to spend about $300 million on Anthropic tokens in 2026, with much of that tied to coding and software development work. Separate reports this week linked that spending plan to Salesforce’s earlier decision to pause hiring new software engineers. The juxtaposition matters because it shows where the budget is moving. Salesforce is not describing a small internal pilot. It is describing AI consumption at a scale large enough to sit next to payroll decisions, at least for incremental engineering hiring. That framing comes from Benioff’s own comments, which emphasized productivity and output rather than direct job replacement. (moneycontrol.com) ### When did Salesforce stop adding software engineers? January 2025 was the first clear marker. Benioff said then that Salesforce would not hire any more software engineers in 2025 because AI tools had lifted engineering productivity by about 30%, according to Moneycontrol. The company said at the time it would still add sales staff. (finance.yahoo.com) That earlier decision is what gives the new Anthropic figure its weight. The company had already argued that AI was letting existing teams do more. The May 2026 disclosure puts a dollar amount on how much Salesforce is willing to spend to keep pushing that model. (moneycontrol.com) ### What exactly is Salesforce buying from Anthropic? Anthropic tokens are usage credits for the startup’s AI models. Benioff said Salesforce expects to consume roughly $300 million of them this year, and multiple reports said coding is the main use case. Some of that work is tied to internal engineering productivity and developer tools. (moneycontrol.com) Benioff also used the podcast to praise Anthropic’s coding products. Reports quoting the interview said he called the coding agents “awesome” and said the company was seeing unusually strong productivity gains from them. ### Is Benioff saying AI is replacing engineers? Benioff’s public line has been narrower than that. (finance.yahoo.com) Recent coverage said he framed the tools as making engineers more productive and more efficient, not eliminating the need for engineers altogether. One report said Salesforce still has about 15,000 engineers using tools that include Anthropic models, OpenAI Codex and Cursor. (analyticsindiamag.com) That distinction is important because Salesforce has made similar claims in other functions. Earlier reporting said the company had reduced support headcount as AI agents handled more customer interactions, while Benioff described AI as a “digital labor” layer across parts of the business. (economictimes.indiatimes.com) ### Why are people focusing on the $300 million number? The $300 million figure is unusually concrete. Companies often talk about AI strategy in terms of pilots, partnerships or broad capital spending. Benioff instead attached a specific annual consumption number to one model provider and one broad class of work: coding. (moneycontrol.com) That makes the Salesforce example easier to read than most. The company is signaling that AI inference and model access are becoming an operating expense inside software development, not just an experimentation line item. That is an inference from the size and stated purpose of the spend, based on Benioff’s comments and the company’s prior hiring stance. (finance.yahoo.com) ### What should readers watch next? Salesforce’s next public checkpoints are likely to be earnings commentary, hiring disclosures and product updates around Slack and coding tools. Benioff said on the podcast that Salesforce is working on technology to make coding easier inside Slack, and future filings or calls could show whether the Anthropic spend changes headcount, margins or product rollout timing. (finance.yahoo.com) (businessinsider.com)