Benioff’s September 15 keynote framed Salesforce as the operating system for the ‘Agentic Enterprise,’ with Anthropic’s Dario Amodei and NVIDIA’s Jensen Huang appearing alongside him. The headline launches: AIforce (spanning Claudeforce, now in open beta, Slackforce, and Agentforce Coworker); Koa, Salesforce’s first CRM-native reasoning model, built on NVIDIA Nemotron; and expanded partnerships wiring Agentforce to Google’s Gemini Enterprise over MCP and to AWS (Amazon Quick, Bedrock model choice, Agentforce Voice). Salesforce said 30,000 customers are now live on Agentforce.

BMO, Canaccord, Citi, Guggenheim, JPMorgan, and Stifel all turned more constructive around Dreamforce, with several (Canaccord, Guggenheim, Stifel) landing at $300 targets and Citizens’ Patrick Walravens reiterating Market Outperform at $315 on September 18. The common thread: Agentforce ARR up 240%+ toward a ~$1.5B FY27 target, premium-license ARR past $1B, and a six-path monetization framework across premium editions — enough to move the debate from ‘can Salesforce monetize agents?’ to ‘how fast?’ Stifel reaffirmed the FY30 $63B revenue target.

Despite the announcement blitz and the upgrade wave, Salesforce closed around $237.92 on September 18 (down ~2% on the day and roughly 4% on the week from $247.72), as the September 16 Fed hike to 3.75–4% pressured high-multiple software broadly. Consensus across 56 analysts nudged up to $275.87 (from $270.37), leaving ~16% implied upside — the Street is treating the pullback as a rate event, not a thesis change, with every bull-camp target still above the current price.

Salesforce put hard figures behind the platform: Agentforce ARR up more than 240% year over year toward a ~$1.5B FY27 target, having crossed $1B in Q1 FY27; premium-license ARR past $1B; and top-100 customers generating over 2x their ARR within 18 months. Of the seven job-ready agents (Casey, Paige, Carter, Marshall, Piper, Fin, Hunter), six are now generally available, with Hunter’s outbound-sales agent slated for GA in November 2026. Koa, built on NVIDIA Nemotron, becomes the reasoning engine underneath them.

The most strategically telling part of Dreamforce was who stood with Salesforce rather than against it. Claudeforce (Anthropic) entered open beta; Koa runs on NVIDIA Nemotron; an expanded Google Cloud deal connects Agentforce and Gemini Enterprise over the Model Context Protocol; and a widened AWS collaboration spans Amazon Quick, Bedrock model choice, and Agentforce Voice. The message to enterprises weighing GPT-6 Astra or a DIY agent stack: bring any model you like — Salesforce wants to be the multi-model orchestration and trust layer, not the single-model bet.

ServiceNow launched Action Fabric, a major overhaul of its AI Control Tower that opens the full ServiceNow platform to any AI agent — its own, Anthropic’s Claude, Microsoft Copilot, or a customer’s homegrown bot — and deepened its Microsoft integration so ServiceNow AI specialists appear in the Microsoft Agent 365 Marketplace as governed digital employees. It is the same bet Salesforce is making from the CRM side: whoever owns the governed execution layer for a multi-vendor agent world owns the enterprise relationship. Two of the largest workflow platforms converging on identical positioning in a single week is the signal.

Microsoft completed the migration of consumer and enterprise Copilot into a single unified product, with Windows and Mac desktop apps shipping by mid-September, and continued pushing Copilot toward autonomous, multi-step ‘AutoPilot’ agents. For Salesforce and ServiceNow, Microsoft is the one rival with both the distribution and the model access to contest the control-layer position directly — which is precisely why both spent this week making their platforms agent-neutral rather than model-neutral.

A week after posting 62% cloud growth and a $664B RPO, Oracle slid roughly 3% on September 18, down about 24% year to date, as rising interest expense and heavy AI capital spending — not demand — became the dominant concern. The timing is the point: the same Fed hike that pressured Salesforce hits debt-funded infrastructure builders like Oracle even harder. It is the cleanest illustration of this week’s theme — AI demand is not in question; the cost of financing the buildout suddenly is.

The FOMC voted 12–0 on September 16 to raise rates 25 basis points to 3.75–4%, the first increase since 2023, as persistent inflation overrode growth concerns. The dot plot was the hawkish part: 16 of 18 officials see at least one more hike this year, with year-end projections of 4.1–4.4%. For enterprise SaaS this is the mechanism behind the week’s price action — a higher discount rate compresses the multiples on long-duration growth names regardless of how strong the product news is, which is exactly what CRM, Oracle, and the broader complex experienced during Dreamforce.

Against the rate-driven multiple compression sits an unchanged demand backdrop: Gartner projects 2026 enterprise software spend rising about 14.7% past $1.4 trillion, with generative AI the primary accelerant, and roughly 40% of enterprise applications carrying task-specific AI agents by year-end (up from under 5%). That tension — hostile discount rate, expanding end market — is the frame for the whole sector into Q3 prints: the addressable market for agentic software is still compounding fast; the argument is now entirely about who captures it and at what financing cost.