Inside Salesforce’s $300 Million Anthropic Token Commitment
Salesforce CEO Marc Benioff has signalled an aggressive new phase in enterprise AI by projecting that the company will spend USD 300 million (approx. RM1.38 billion) on Anthropic tokens in 2026, largely for software development. Tokens are the basic units of text that Anthropic’s Claude models process, and enterprise AI tokens are billed based on volume, turning this into a massive consumption-based contract for coding workloads. Benioff calls Anthropic “a rocket ship that will not stop” and says coding agents are making “everything cheaper to make.” While neither firm has formally disclosed the deal, this level of spend would position Salesforce as one of Anthropic’s largest commercial accounts and among the most committed enterprise buyers of AI infrastructure, underscoring how cloud software vendors now view foundation models as a core utility, not an experimental add-on.
From Hiring Freeze to AI-Generated Code Oversight
Behind the Salesforce Anthropic investment is a deliberate restructuring of work. Salesforce froze engineering hiring for 2025 after AI tools boosted productivity by more than 30%, with AI now handling between 30% and 50% of overall workload. The company’s roughly 15,000 engineers are still in place, but their roles are shifting: instead of hand-writing every line, they supervise AI-generated code, validate outputs, and manage coding agents from Anthropic, OpenAI Codex, Cursor, and Salesforce’s own Agentforce. Benioff is explicit that human engineers remain necessary, yet “we’re not adding any more software engineers next year” because of AI efficiency gains. At the same time, Salesforce is hiring 1,000 to 2,000 salespeople to help customers adopt AI products. This rebalancing reflects a wider enterprise pattern: fewer net-new coders, more AI-generated code oversight, and expanding go-to-market teams to sell the new AI stack.
Slack Coding Tools and the New AI-First Developer Workflow
Salesforce is not just consuming Anthropic’s models; it is embedding them deep into daily workflows. Benioff revealed that Salesforce is building Slack coding tools to “make it easier for everybody to code,” hinting at AI pair-programming directly inside the collaboration hub it acquired for USD 27.7 billion (approx. RM127.4 billion). That push aligns with the company’s broader platform strategy: Agentforce has reached about USD 800 million (approx. RM3.68 billion) in annual recurring revenue with 29,000 deals, while the new Headless 360 API-first platform exposes 60-plus tools so agents like Claude Code can access Salesforce’s enterprise stack. For developers, this means coding, reviewing, and deploying from within chat, while AI agents handle routine tasks. For enterprises, it shows how AI assistance is being woven into existing platforms rather than bolted on, turning tools like Slack into central hubs for orchestrating AI-powered software creation.
Cost Optimisation and Competitive Positioning in Enterprise AI
Committing USD 300 million (approx. RM1.38 billion) to Anthropic tokens forces Salesforce to treat AI compute like a strategic P&L line. Benioff stresses that not every request should hit a frontier Claude model. Instead, Salesforce is building an “intermediary layer” that routes simpler jobs to smaller, cheaper models and reserves premium capacity for complex reasoning tasks. This tiered approach mirrors how enterprises historically optimised cloud spend, but now the unit is tokens, not virtual machines. The stakes are high: AI agents have already helped Salesforce reduce its support workforce from 9,000 to 5,000, demonstrating tangible ROI. At the same time, the company’s deep integration of Anthropic, alongside OpenAI Codex, Cursor, and Agentforce, positions it in a fierce competitive race where software giants must prove not only that they can embed AI everywhere, but that they can do so profitably at massive scale.