Salesforce’s Agentforce Booming at 169% ARR Growth—Could AI Agents Be the Next $800M Revenue Engine?

Salesforce is making a clear play to become the foundational layer for the next enterprise computing paradigm. Its bet centers on Agentforce, its platform for building and deploying AI agents, which has reached an annual revenue run rate of around $800 million by the end of fiscal 2026. The platform's adoption is accelerating rapidly, with more than 75% of Salesforce's top deals in the fourth quarter including Agentforce. This isn't just a feature addition; it's a strategic move to embed AI agents directly into the core workflows of business.

The company's partnership with NVIDIA is the linchpin of this strategy. By integrating NVIDIA's open-source Agent Toolkit into its platform, Salesforce gains access to a shared foundation for building autonomous agents. This toolkit provides the essential software stack-models, runtime, security frameworks-that agents need to operate. For Salesforce, this partnership is a first-principles move: it leverages NVIDIA's infrastructure to accelerate its own platform's development and ensures its agents are built on a standard optimized for the hardware driving the AI boom.

The purpose of this collaboration is to move AI from the lab to the factory floor. As NVIDIA's CEO Jensen Huang stated, the goal is to reposition the IT industry for its next great expansion. Salesforce and NVIDIA are working to make AI agents usable in regulated, real-world enterprise workflows, from autonomous customer service agents to workflow automation. By embedding agents into CRM and customer data platforms like Data 360, Salesforce aims to become the essential infrastructure layer where the next wave of enterprise productivity is built.

The Reference Architecture: Blueprint for Exponential Adoption

The Salesforce-NVIDIA partnership is more than a software deal; it's a meticulously crafted reference architecture designed to lower the barrier to AI adoption across the enterprise. This blueprint provides a clear, plug-and-play model for building agentic systems, turning abstract potential into a tangible path for customers.

At its core is a conversational interface. Employees interact with agents through Slack as the collaboration and engagement layer, using natural language to request actions. Slackbot serves as the coordination layer for enterprise agents and workflows, receiving the user's prompt and triggering the underlying process. This seamless integration means agents don't disrupt existing workflows-they become part of them.

The intelligence behind the scenes is powered by NVIDIA's Nemotron 3 Nano model, integrated directly into Agentforce. Its 1 million-token context window is a critical enabler, allowing agents to reason across vast amounts of data without losing track. This capability is essential for complex tasks, like analyzing a customer's entire history across multiple touchpoints or orchestrating a multi-step service resolution, which would otherwise require breaking the workflow into smaller, context-lossy pieces.

This architecture isn't theoretical. Salesforce has already demonstrated its power in a physical setting. At its own headquarters, the company connected Cobalt's security robots to Agentforce. These robots, equipped with NVIDIA's visual AI as their "eyes," can now perceive issues in real time-like a door left open or a blocked exit-and automatically trigger fixes through the Salesforce system. The result is a dramatic shift from reactive to proactive operations, with human interventions reduced by 99%.

This physical demonstration is the ultimate validation of the blueprint. It shows how the reference architecture-Slack for input, Agentforce for reasoning and action, Nemotron for deep context, and NVIDIA infrastructure for processing-can be applied to the physical world. By providing this clear, end-to-end pattern, Salesforce and NVIDIA are effectively giving enterprises a ready-made template. It reduces the risk and complexity of AI adoption, accelerating the shift from pilot projects to enterprise-wide deployment. In doing so, they are building the fundamental rails for the Agentic Enterprise, where the next wave of productivity is not just digital, but embodied.

Financial Engine and Adoption Metrics

Salesforce's financial engine is firing on all cylinders, providing the capital and visibility to fund its aggressive pivot. The company's total remaining performance obligation (RPO) now exceeds $72 billion, a 14% year-over-year increase that signals strong future revenue visibility. This massive backlog, built on its core SaaS contracts, is the fuel for its transformation. It provides a stable foundation while the company invests heavily in its AI infrastructure, with operating cash flow scaling to $15 billion last fiscal year.

The adoption metrics for its AI layer are where the exponential curve begins to steepen. Agentforce, the platform for building enterprise agents, is moving from concept to operational scale. To date, it has delivered more than 2.4 billion agentic work units and processed over 19 trillion tokens. These numbers represent a tangible shift from AI experimentation to AI-driven productivity. The platform's growth is staggering, with its annual recurring revenue jumping 169% to $800 million in the past year, and it has closed over 29,000 deals since launch.

Yet there's an apparent disconnect in the near-term outlook. The company's fiscal 2027 revenue guidance of $45.8 billion to $46.2 billion implies only 10% to 11% growth. This muted projection likely reflects a near-term plateau in its traditional seat-based SaaS model, which has matured. The guidance underscores that the real growth story is not in the legacy business, but in the adoption of its new agentic infrastructure. The $72 billion RPO provides the runway, but the market is now watching the conversion of that backlog into the next wave of AI-driven work.

Valuation and Market Sentiment

The market's view of Salesforce's AI bet is one of deep skepticism, reflected in a stock price that has fallen nearly 26% so far in 2026. This steep decline is driven by a fundamental fear: that AI could disrupt the very seat-based SaaS model that built the company. The sell-off has priced in a slowdown, with the company's own muted fiscal 2027 revenue guidance of $45.8 billion to $46.2 billion reinforcing concerns about a plateau in its traditional business. In this light, the stock trades at a significant discount to its historical valuation, ignoring the explosive growth of its new agentic layer.

This skepticism is mirrored in insider actions, revealing a clear internal debate. While CEO Marc Benioff sold $110 million worth of shares last year, a rare board buy by former Nvidia scientist David Blair Kirk stands out. Kirk's purchase of just over $500,000 worth of shares is an outlier amid dozens of sizable insider sales. His background as a former Nvidia chief scientist adds weight to the move, signaling a belief that the partnership with NVIDIA is a strategic validation of Salesforce's platform play. Yet the sheer volume of sales from top executives suggests many see near-term risks or simply want to cash out.

The partnership with NVIDIA is the clearest validation of Salesforce's strategy. By integrating NVIDIA's Agent Toolkit into its platform, Salesforce is building its agentic infrastructure on a shared foundation that is optimized for the hardware driving the AI boom. This collaboration, announced at GTC 2026, is a first-principles move to become a key layer in the next enterprise computing paradigm. However, the market appears to be pricing in execution risk. It sees the partnership as a necessary step, not a guaranteed path to growth, and remains focused on the near-term headwinds to its legacy model. The bottom line is that while the blueprint for the Agentic Enterprise is being laid down, the market is still waiting to see if Salesforce can successfully navigate the transition without a costly detour.

Catalysts, Risks, and What to Watch

The success of Salesforce's infrastructure bet hinges on a few critical catalysts and risks. The company is building the rails for the Agentic Enterprise, but the market will judge it on execution and adoption speed.

The most immediate technical catalyst is the integration of NVIDIA's Nemotron 3 Nano model into Agentforce. This isn't just a software update; it's a fundamental efficiency gain. The model's Mixture of Experts architecture promises to increase computational efficiency, reduce reasoning tokens, and lower compute demand and cost in multi-step workflows. For enterprise customers, this translates directly to more powerful, cost-effective agents. It accelerates the shift from expensive, experimental AI to a scalable, operational layer. Success here will be measured by how quickly this efficiency trickles into customer deployments and reduces the total cost of running agentic systems.

The dominant competitive risk, however, is a paradigm shift within the market itself. AI agents are designed to automate tasks traditionally handled by human employees. This creates a potential threat to the very CRM market Salesforce built. If AI agents can handle customer service, sales development, and support autonomously, the need for human agents-and the seat licenses that underpin Salesforce's legacy model-could diminish. The company's strategy of embedding agents into its core products is a direct attempt to capture that value. The risk is that AI adoption simply cannibalizes the existing revenue base before the new agentic layer can fully scale.

The key success metrics will be the growth of Agentforce's revenue run rate and its penetration into new physical workflows. The platform's annual recurring revenue (ARR) grew 169% year over year to $800 million by the end of fiscal 2026, a powerful signal of adoption. Investors must watch for this growth to accelerate beyond that explosive rate, indicating the transition from early adopters to mainstream enterprise use. Equally important is the expansion into the physical world. The deployment of Cobalt's security robots at Salesforce Tower is a blueprint for a new category. Success will be measured by how many more physical workflows-inventory audits, facility management, asset tracking-can be automated through Agentforce. Each new physical application validates the platform's ability to act as the central nervous system for an agentic enterprise, moving beyond digital labor to embodied intelligence.

The bottom line is that Salesforce is racing against two timelines. One is the exponential adoption curve of its agentic infrastructure, fueled by partnerships and efficiency gains. The other is the potential disruption of its own core business by the AI it is championing. The company's massive $72 billion backlog provides a runway, but the market will demand proof that the new layer is not just a side project, but the essential foundation for the next computing paradigm.