Salesforce AI Revenue Nears $3.9 Billion as Agent Buying Shifts to Credits

Salesforce reported its fiscal second-quarter results on August 26, and its AI business grew far faster than the rest of the company. Combined annual recurring revenue for Agentforce and Data 360 reached nearly $3.9 billion, up more than 210% from a year earlier, while total revenue rose 11% to $11.3 billion, Salesforce said in its earnings release. For CIOs, the numbers matter less as a stock story than as a signal of how AI agents are now being bought, priced and metered inside the systems they already run.

What did Salesforce report for Q2 fiscal 2027?

The quarter ended July 31. Subscription and support revenue came to $10.8 billion, up 12%, and current remaining performance obligations, the contracted revenue due over the next year, rose 14% to $33.5 billion. The results included about $456 million from Informatica, which Salesforce now owns, according to the release.

Agentforce ARR alone passed $1.5 billion, up more than 240%. One caveat for anyone comparing quarters: starting this quarter, Salesforce counts its AI offerings, Slackbot and Headless 360 within that Agentforce figure, the release notes. Some of the jump reflects a wider definition, not only new demand.

Chair and CEO Marc Benioff said in the release that "AI is delivering value across every layer of our platform." The company raised its full-year revenue outlook to $46.1 billion to $46.4 billion. About $200 million of that increase is tied to the pending acquisitions of Contentful and Fin, with $100 million from organic growth, partly offset by a $100 million currency headwind.

Investors reacted strongly. By Thursday afternoon, shares were up 19.6% from the prior Friday's close, The Motley Fool reported, and Jefferies and Raymond James both raised their price targets.

How are enterprises actually using Agentforce?

The more useful figures for enterprise buyers came from the earnings call. Salesforce added about 2,000 paying customers with Agentforce in production during the quarter, a 70% increase from the prior quarter, President and COO Miguel Milano said, according to a MarketBeat summary of the call on Nasdaq.

Usage is climbing faster than customer counts. Customers generated 3.2 billion "agentic work units" in the quarter, up 97% sequentially, and 7 billion to date, the company said. Milano also said roughly half of Agentforce bookings came from customers topping up usage credits. That is a consumption pattern, closer to how companies buy cloud compute than how they buy CRM seats.

Data volumes point the same way. Data 360 ingested 104 trillion records in the quarter, 82 trillion of them through Zero Copy connections that read data where it already lives in warehouses and lakes, according to the release.

Why does consumption pricing matter to CIOs?

Milano said Salesforce now sells agents per user, per agent, by consumption and by outcome. He added that customers starting with agents have reached about twice their average contract value, with room for three to four times that, per the MarketBeat summary.

That second number is the one finance teams should notice. It shows the vendor's expectation that agent spending will expand well beyond today's software budgets. Credit-based buying makes pilots easy to start, but it also makes spend harder to forecast, because the bill tracks how often agents run, not how many people use them.

AI Connect speakers have warned about this gap for a while. In a AI Connect webinar on taking AI from prototype to production, Arun Agarwal, Senior Product Marketing Manager, AI/ML, at Snowflake, said many projects die from poor estimates of volume and tokens. "You don't optimize what you don't measure," he said. At COLLIDE 2024, panelists made a similar point about unmonitored cloud and GPU spending eroding AI returns.

What else did Salesforce announce?

Alongside the results, Salesforce and Anthropic unveiled Claudeforce, a deeper integration of Claude models across Salesforce, Slack and Agentforce, which we cover in a separate story. Benioff also used the call to reject the "SaaSpocalypse" argument that generative AI will erode demand for traditional software, citing what CFO Robin Washington described as the strongest growth in net new annual contract value in four years.

What should enterprise leaders do next?

Three practical steps follow from the quarter:

• Ask for usage data before renewal. If half of bookings are credit top-ups, buyers should insist on work-unit reporting by team and use case, and set alerts before credits run out.
• Compare like with like. The broader Agentforce ARR definition makes year-over-year growth look larger. Ask vendors for consistent metrics when judging adoption claims.
• Fix the data first. Zero Copy ingestion shows agents increasingly read from existing warehouses. As Jack Smith, Principal Solutions Engineer at Syncari, argued at COLLIDE 2025 in a talk on why bad data blocks AI at scale, "Integration is not a strategy, neither is siloed warehouses." Agents pointed at messy data will burn credits on messy answers.

The next test comes when the Contentful and Fin deals close, which Salesforce expects in its fiscal third quarter, and when Claude-based features reach customers beyond the pilot group.

Frequently asked questions

How much revenue did Salesforce report for Q2 fiscal 2027?

Salesforce reported $11.3 billion for the quarter ended July 31, 2026, up 11%, with subscription and support revenue of $10.8 billion.

What is Agentforce ARR?

It is annual recurring revenue from Salesforce's agent products. It passed $1.5 billion in Q2 fiscal 2027 and, from that quarter, includes AI offerings, Slackbot and Headless 360.

How is Salesforce pricing Agentforce?

Salesforce sells agents per user, per agent, by consumption and by outcome. About half of Q2 Agentforce bookings came from customers buying more usage credits.

Filed under: Agentic AI, AI Strategy & ROI