Salesforce.com (CRM.US) held its earnings conference call for the fourth quarter of fiscal year 2026. The company forecasts first-quarter revenue to be between $11.03 billion and $11.08 billion, representing an estimated nominal growth rate of approximately 12% to 13%. Concurrently, it anticipates current remaining performance obligation (cRPO) nominal year-over-year growth for Q1 to be around 14%. For the full fiscal year 2027, revenue is projected to be in the range of $45.8 billion to $46.2 billion, implying a growth rate of about 10% to 11%. Bolstered by strong Q4 performance and confidence in the integration of Informatica, the company has raised its revenue target for fiscal year 2030 to $63 billion. The company announced an increase in its stock repurchase authorization to $50 billion, with management expressing the view that the current share price is undervervalued and represents an excellent buying opportunity. Management emphasized that AI is no longer just an assistant but is redefining workflows. AgentForce completed 29,000 orders within its first 15 months post-launch, growing 50% quarter-over-quarter. The combined annual recurring revenue (ARR) for the AgentForce and Data 360 businesses, including Informatica, has surpassed $2.9 billion, marking a substantial 200% year-over-year increase. In Q4, over 75% of the top 100 deals included AgentForce and Data360. The company introduced the "Agent Work Unit" (AWU) to measure the actual value generated by AI. The platform has now delivered 2.4 billion AWUs (with 771 million contributed in Q4), signaling a transition for AI from "conversation" to "delivering outcomes."

Q&A Session

Q: Regarding the divergence between AgentForce's rapid growth and the cRPO growth (9%), which only met expectations. How does management view AgentForce's role in driving the overall product portfolio? Is there confidence in achieving broad-based business acceleration in the second half of the year? A: Salesforce is a comprehensive business entity. We are not only signing new deals and developing new technologies but also carrying forward our "heritage business," driving growth through continuous renewals. This structure also provides us with predictability for future fiscal years. In fact, FY26 performance was significantly better than initial expectations at the start of the year. Particularly in the third and fourth quarters, the results exceeded my expectations. The performance of AgentForce and Data360 also surpassed projections. We are fully capable of maintaining innovation while completing more renewals and driving the overall business. We are satisfied with our current achievements. We are commercializing AI in multiple ways. Currently, we are seeing robust growth in high-end SKUs, and the business is accelerating. A crucial point is that our seat count continues to grow both year-over-year and quarter-over-quarter. With the implementation of AgentForce and the "agent system" logic, we are seeing the incremental value of software materialize. The future growth model will be hybrid: seat count will remain a core component, while usage-based consumption models will serve as an important supplement. We expect to see more incremental value released from agent technology and capabilities.

Q: Against the backdrop of declining valuation multiples for tech stocks, why opt for a massive $50 billion buyback instead of more aggressively using these funds for strategic mergers and acquisitions (M&A)? A: Regarding capital allocation, I have a very clear framework. First is the dividend; we just increased it by 5%, which is a very important component. Second is traditional share repurchases, where we have been very active in recent years. Regarding M&A, we haven't stopped, but we now strictly adhere to a "new formula." Looking back at Salesforce's history, I almost wish we had adopted this logic earlier, as it allows us to more clearly judge which acquisitions add value to the business rather than merely diluting shareholder equity. Furthermore, debt is a key dimension. Frankly, our balance sheet is currently under-leveraged and not being utilized effectively. Considering we will generate over $16 billion in cash flow this year, and given that some past acquisitions (like Slack and Tableau) did dilute investor interests, now is an excellent opportunity—to buy back shares at such an attractive price level. We must use capital correctly, and debt is an effective tool to achieve this. I hope CFO Robin repurchases as much stock as possible. A large-scale buyback does not mean we are abandoning growth. With strong free cash flow and cash reserves, we are fully capable of pursuing both paths simultaneously. In fact, we just completed acquisitions of 10 companies while returning over 99% of free cash flow to shareholders through buybacks and dividends. When discussing such a large cash flow scale—$15 billion last year, expected to reach at least $16.5 billion this year—the core question is how to use it correctly. Dividends, buybacks, M&A, and debt management: these four pillars are all essential. We are completely open to balancing these four and are committed to optimizing the balance sheet to achieve both organic growth, inorganic growth, and shareholder returns.

Q: Concerning relationships with model partners (e.g., Anthropic): As these providers potentially evolve downstream, where are the boundaries between competition and cooperation? In which areas does Salesforce believe it has a definitive advantage, and which belong to the model providers? A: Our vision for the world is clear: large language models (whether OpenAI, Anthropic, Gemini, DeepSeek, Mistral, etc.) are new components in our infrastructure that we haven't had in previous years. In the past, we relied on our proprietary Einstein model to understand business. While we still maintain our own models, we are also open to external models, having processed 19 trillion tokens on these models to date. Will these models themselves become platforms? The answer is yes. Just like Windows, Mac, or iOS, applications may be born on these model platforms. This is indeed one future shape. But as a software company, our duty is to use available tools to help customers succeed and connect with them in new ways. Our advantages lie in: First, a deep customer base and distribution network. We have over 150,000 customers in our core business, 1 million customers on Slack, and 15,000 sales representatives in the field helping customers plan for future success. Second, enterprise-grade implementation. The current reality is "humans and agents working together." Our task is to translate existing technology into usable services, which is currently lacking in model platforms. Third, compliance and reliability. For enterprise clients like large banks, automating call centers, sales, and employee processes requires not just AI, but AI that meets compliance, security, scalability, and reliability standards. For example, on help.salesforce.com, we can already automate connections to customer centers, which was unimaginable a few years ago. The future may evolve in various forms, but our current focus is on what we can sell to customers this year and what practical problems we can solve for them. We have a tremendous amount of work to do and a vast array of products to sell.

Q: How is management translating token consumption and "Agent Work Units" (AWUs) into actual revenue? A: We focus not only on underlying token consumption; although we've processed 19 trillion tokens, this is merely a measure of intellectual input for model suppliers (like OpenAI, Anthropic). In the enterprise world, the value of asking AI a question or having it write a poem is limited. The real value lies in whether it can create documents, update records, or assist in decision-making. Therefore, we introduced the "Agent Work Unit" (AWU) concept. We have identified a ratio between token consumption and actual work output. If a customer consumes many tokens but produces little actual work, it indicates an efficiency problem. AWU is a more valuable metric, signifying our ability to transition customers towards becoming "agent enterprises." Tokens are merely a leading indicator on the cost side, while work units represent the true measure of value creation.

Q: As the "agent value" could potentially be 3-4 times that of traditional software value, how will the pricing model evolve? And what are the specific implications for gross margins? A: Regarding gross margins, we believe the short-term impact is neutral. The mentioned differential pricing between tokens and AWUs is crucial. On the cost side, token prices are expected to commoditize and trend downward over time due to market competition. Simultaneously, our engineering team is refining products through technologies like AgentForce scripts, which can significantly reduce underlying costs for the same work output. Additionally, we adhere to a "customer zero" strategy, mitigating costs through internal resource reallocation and efficiency gains (e.g., all employees using Slackbot to prepare meeting minutes). Based on the FY27 financial framework, we are confident in driving operating margin expansion while maintaining robust gross margins through technological optimization and scale effects.

Q: Regarding the market progress, customer adoption, and underlying monetization logic of AI Launch Agreements (ALAs)? A: At a previous Investor Day, we mentioned that revenue could re-accelerate within 12 to 18 months. Today, we can state with high certainty that organic revenue re-acceleration for the Subscription and Support business will occur in the second half of this year. Our confidence stems from the fact that Net New Annual Contract Value (NENU-AOV) growth already surpassed AOV growth in the second half of last year, and this trend is expected to widen in Q1 and Q2 of this year. This growth driven by new orders will formally translate into revenue momentum in the second half. Based on this, we have raised our long-term growth target for FY30 to $63 billion, reflecting our high degree of certainty in achieving this goal. Regarding AI monetization, we have identified a clear formula, primarily across three dimensions: First, leveraging our massive installed base of 100 million seats to drive upgrades to premium SKUs that include embedded AI and Unlimited permissions. Growth in this segment this quarter was astounding, tripling quarter-over-quarter (compared to doubling last quarter). Second, Agentic applications (like AgentForce Sales/Service) significantly enhance customer ROI. Due to the changed value proposition, we are entering areas previously considered too expensive for Salesforce, thereby gaining entirely new seat growth. Third, for consumer-facing AI scenarios, we sell "fuel" i.e., Flex Credits. In Q4 bookings, half came from credit sales and half from premium SKU upgrades. Based on Q4 closing data, this was our best quarter historically. We have never closed 12 deals over $10 million in a single quarter before, with one even exceeding $50 million. Among the top 10 deals, 6 involved existing SKU upgrades, 7 involved new seat additions, and 5 included credits for Agent scenarios, with 3 deals encompassing all three models. This demonstrates our ability to convert AI into financial returns from multiple angles. Looking ahead, I am confident about Q1. The current pipeline is showing double-digit growth, and more importantly, our sales execution capability has been significantly unleashed. A year ago, our trained sales headcount growth was 0%; at the start of this fiscal year, it has reached 15% to 17%, which is undoubtedly "explosive" for performance growth. Furthermore, ILA has become a core product for us, with over 120 sold in Q4, far exceeding the expected 50-100. Among the top 10 deals, 8 included ILA agreements, indicating that top customers are fully committing and establishing long-term partnerships with us.

Q: Given the strong cross-sell of AgentForce and token upsell (contributing 60% of bookings), how does management view the path for acquiring new customers and enabling their rapid deployment this year? Are there any potential obstacles in the adoption process? A: We have currently secured 29,000 AgentForce deals, covering approximately 23,000 customers. The core task for our management team and account executives (AEs) is to engage with customers and articulate the value we can deliver. Just recently (or due to time zones, just occurred) at the World Tour in Australia, 12,000 customers attended, demonstrating market enthusiasm. The core message we convey to customers is: In the LLM era, SaaS is more critical than ever. While we are excited about the emergence of raw intelligence (models), translating it into accurate, secure, scalable enterprise-grade work requires a software infrastructure comprising context systems, work systems, and agent systems. We hold a 40% market share in sales and service, a scale and complexity unmatched by any other company building similar systems. Our agents are connected to real data and can trigger actual actions. Furthermore, Slack, as our collaboration system, its importance is evident in comparisons with demos from partners like Anthropic. While others might demo with clunky UIs, they often end up "copy-pasting" content back into Slack. With Slackbot support, users don't need to switch between different interfaces. We have a deep partnership with Anthropic, but more critically, we possess this native, integrated environment. In recent years, there has been excessive fascination with models and the intelligence layer, but now the application layer and UI are undergoing profound changes. Traditional UIs with complex buttons were designed for human interaction. When humans and agents coexist in the same space, many UI paradigms will be abandoned. This is why Slack is so powerful—it's an environment where humans and agents work together. Slackbot is incredible because it understands not only the system's record data but also all the conversational context occurring within Slack. This conversational data might be our most important asset. When we integrate this data into new user interfaces, the great transformation of SaaS occurs: applications will transform into ecosystems where humans and agents truly collaborate. To add from a customer success perspective, we are doubling down on Field Deployment Engineers (FDEs). These professionals will work alongside our solution sales teams to turn vision into reality. This is a key link in converting AI Launch Agreements (ALAs) into actual consumption. What we want to see is the "wheel of consumption" spinning at high speed continuously; that is the ultimate foundation for business growth.