Here is the number that actually moved Salesforce’s stock 23% on Thursday: 14. As in 14% constant-currency growth in current remaining performance obligation, with cRPO reaching $33.5 billion and representing the fastest cRPO growth in four years. That figure tells you what enterprise customers are committing to pay over the next 12 months. The headline EPS of $5.90 does not tell you that.
Net gains on strategic investments totaled $2.61 billion, compared with $6 million a year earlier, and those gains contributed $2.53 per share to Q2 non-GAAP earnings. Set the gain aside, and the math is plain: $5.90 minus $2.53 leaves about $3.37, against $2.91 in the year-ago quarter, which works out to roughly 16% per-share growth. The apparent 80% EPS beat shrinks to roughly 3% after removing the strategic-investment gains. That is not a scandal. It is just not the story most people are telling today.
The Real Thesis: Data Gravity, Not Rented Reasoning
Claudeforce is described as bringing Claude’s intelligence and reasoning together with Salesforce’s enterprise harness, making its data, workflows, business logic, actions, and governance securely accessible to power agentic experiences wherever work happens. Dario Amodei’s version was more pointed: companies can point Claude at the customer information and business context they have been building in Salesforce for decades, and use it to actually run and grow their businesses.
That framing matters strategically. Claudeforce runs in three directions at once: Salesforce inside Claude, Claude inside Agentforce, and Claude inside Slack. The Salesforce in Claude plugin carries 37 prebuilt sales skills, so a seller can run meeting prep, deal-health reviews, and pipeline updates by reasoning over live CRM data without leaving the Claude window. Through Amazon Bedrock, Claude runs entirely inside the Salesforce trust boundary, with early adopters including CrowdStrike and RBC Wealth Management. That regulated-industries angle is where Salesforce’s data moat converts into something a Claude plugin from any other vendor cannot replicate quickly.
The key distinction, and the one that determines whether this is durable competitive positioning or a well-branded API contract, is articulated cleanly by the underlying architecture. The Claudeforce announcement makes Claude the default model across the platform, a significant commercial commitment, but the infrastructure positions AIforce as a middleware layer that preserves Salesforce’s data moat regardless of which AI model ultimately wins the enterprise reasoning market. In other words: Claude is the front end. The decades of customer data sitting inside Salesforce is the moat. If Claude is someday replaced by a better model, the data stays.
The Operating Numbers That Matter
Customers drove 3.2 billion Agentforce work units in Q2, up 97% quarter over quarter. Accounts with agents in production grew 70% over last quarter. Agentforce and Data 360 generated nearly $3.9 billion in combined ARR, up more than 210% year over year, with Agentforce ARR alone passing $1.5 billion and growing more than 240%. One caveat worth noting: Salesforce broadened the ARR definition in Q2 to include Slackbot and Headless 360, so the headline growth rate is not perfectly comparable with earlier periods.
Management raised the fiscal 2027 revenue outlook by $300 million in constant currency to $46.1 billion to $46.4 billion, with $100 million from organic strength in Agentforce, Data 360 and Slack, and $200 million tied to pending acquisitions. Free cash flow of $1.1 billion, up 81% year over year, confirms that margin discipline is holding.
The Competitive Risk
Microsoft, ServiceNow, and Workday are not standing still. ServiceNow has entered the CRM market with Autonomous CRM for Sales, Autonomous CRM for Service, and Configure, Price, Quote capabilities, while its CRM business is already a $2 billion annual contract value operation. Salesforce enters ITSM with limited process depth; ServiceNow enters CRM with limited data breadth. Neither has yet won the other’s core turf, but the overlap is real and growing.
The bear case on Claudeforce specifically is that partnerships can be renegotiated. Benioff is spending an estimated $300 million on Anthropic tokens in 2026, with software coding as the primary use case. That is a material dependency on a single AI provider whose competitive position relative to Google and OpenAI could shift. If Claude loses its reasoning edge, Salesforce’s default model agreement becomes a liability, not an asset.
What to Watch
The metrics that will confirm or undermine the Claudeforce thesis are not on the income statement. Watch cRPO growth for a third consecutive acceleration. Watch whether Agentforce ARR crosses $2 billion before year-end on a consistent definition. The Salesforce in Claude plugin is currently with select pilot customers, with open beta expected September 2026, so the first real adoption data arrives next quarter. That will be the cleaner read on whether Claude is pulling new workloads into the Salesforce ecosystem or simply rebranding existing ones.
The 23% move is defensible on the fundamentals if you ignore the $2.53 investment gift and focus on what actually drives recurring revenue. A four-year cRPO growth high and 97% sequential growth in agent usage are not small numbers. But investors paying for a re-rated CRM at this level are betting that the data moat is real, the Claudeforce integration deepens switching costs, and usage-based pricing converts agent activity into durable ARR. All three conditions need to hold. Only one of them was confirmed yesterday.
