The stock that was supposed to be an AI casualty had its best day in six years

Salesforce shares closed up 22.6% on Thursday, its sharpest single-day move in six years, adding roughly $38 billion in market value in a single session. That's a genuinely unusual outcome for a company Wall Street had spent much of 2026 treating as a likely casualty of the very technology it now credits for the rally. Shares had been down roughly 22% year-to-date heading into Wednesday's earnings report, dragged lower by a persistent fear running through enterprise software: if AI agents can handle tasks that used to require a human logging into a CRM dashboard, why would companies keep paying per-seat licensing fees for the dashboard itself?

Thursday's move didn't just rescue one stock. It rattled a bearish thesis that had been quietly reshaping how investors price the entire software sector.

What actually beat, and by how much

Salesforce reported second-quarter revenue of $11.35 billion, up 11% year over year, with adjusted earnings per share of $5.90, comfortably ahead of what analysts had modeled. Current remaining performance obligations, a measure of contracted revenue not yet recognized that functions as a forward-looking demand signal, rose 14% to $33.5 billion, beating the $33.22 billion analysts polled by StreetAccount had expected. Free cash flow climbed 81% to $1.1 billion. Management didn't stop at the quarter that just closed; it lifted full-year fiscal 2027 adjusted EPS guidance to a range of $16.67 to $16.71, well above the $14.16 Wall Street consensus, alongside a revenue range of $46.1 billion to $46.4 billion.

Trading volume told its own story about how surprised the market was. Roughly 55.3 million shares changed hands Thursday, more than four times Salesforce's three-month average, the kind of volume spike that shows up when a stock's entire narrative shifts in a single session rather than just its quarterly numbers.

The AI numbers investors actually cared about

Buried inside the earnings release were the figures that mattered more than the headline beat. Agentforce, Salesforce's autonomous AI agent platform, crossed $1.5 billion in annual recurring revenue, up more than 240% year over year, accelerating from the roughly 200% growth rate reported the previous quarter. Combined annual recurring revenue from Agentforce and the company's Data 360 product neared $3.9 billion, up more than 210%. Usage metrics moved in the same direction: Salesforce said its platform delivered 7 billion "Agentic Work Units" cumulatively, with 3.2 billion of those in the second quarter alone, a 97% jump from the first quarter's pace. Premium Agentforce bookings more than doubled quarter over quarter.

Those numbers answer a specific question that had been hanging over the stock: are customers actually paying for Salesforce's AI agents, or just experimenting with them for free? Doubling premium bookings in a single quarter is a fairly direct signal that large enterprise customers are treating Agentforce as a genuine budget line, not a pilot project.

The Anthropic deal that changed the framing

Alongside earnings, Salesforce and Anthropic announced an expanded partnership called Claudeforce, integrating Claude's reasoning directly with Salesforce's underlying data, workflows, and governance systems. CEO Marc Benioff described it as delivering "a dynamic interface that thinks, reasons, and acts," and used the announcement to directly confront the bear case that had weighed on the stock all year, dismissing what analysts have dubbed the "SaaSpocalypse" narrative as overblown in a Thursday appearance on CNBC's Mad Money.

The billing structure underneath the partnership is worth understanding, because it points to where enterprise software economics are actually headed. Customers pay Salesforce a platform fee that varies by edition, then contract separately with Anthropic for the Claude inference itself, essentially a two-invoice structure. That arrangement matters because it signals a shift away from flat per-seat licensing toward usage-based, consumption-driven pricing, the same broader trend already forcing Salesforce itself to grapple with rising AI costs; CEO Benioff has said the company's own Anthropic bill will run close to $300 million this year, and has publicly wished for a "smart router" that could send simpler queries to cheaper models rather than always reaching for the most expensive one.

Why the market treated this as bigger than one company's quarter

The rally didn't stay contained to Salesforce. ServiceNow closed up 10.05% the same day, and Workday gained 1.48%, both riding what analysts described as broad enthusiasm across enterprise software after Salesforce's results landed. That kind of sector-wide lift suggests investors read Thursday's numbers as evidence about a structural question, not just a single company's execution: whether large incumbent software vendors can successfully fold AI agents into their existing platforms as a new revenue layer, rather than watching those agents cannibalize the licensing business that built the company in the first place.

Salesforce's specific pitch is that customers are willing to pay a premium for AI agents that come wired directly into their existing CRM data, workflow rules, and governance controls, rather than adopting a standalone third-party agent that has to be integrated from scratch. If that argument holds up across more quarters, it offers a template other legacy software companies facing similar AI-disruption fears might try to follow. The timing also mattered: Thursday's rally landed the same week Nvidia posted its own blowout quarter and Okta jumped 28% on its own AI-security pitch, turning the week into a broader referendum on which companies are actually capturing durable value from the AI buildout versus which ones are simply riding sentiment about it.

What's still unresolved

None of this settles the underlying tension permanently. Salesforce's own chief operating and financial officer, Robin Washington, acknowledged on the earnings call that the company saw "headwinds and volatility" selling licenses for integration and analytics software even as Agentforce accelerated, a reminder that AI adoption isn't lifting every part of the legacy business equally. And the economics Benioff himself flagged, a $300 million and rising AI bill, plus the industry-wide shift toward consumption-based pricing, mean software vendors are now managing their own AI costs as carefully as they're trying to monetize AI for customers. Thursday's rally answered one question convincingly: enterprise customers are still willing to expand their spending with Salesforce when AI is packaged around trusted platform data rather than sold as a standalone replacement. Whether that translates into sustained growth over the next several quarters, rather than one very good earnings reaction, is the test that actually determines if the SaaSpocalypse fear was wrong, or just early.