BNP’s Bull Case: Salesforce, Workday, Snowflake as Agentic AI’s Essential Infrastructure Stack

We are in the steep middle of an S-curve. The global agentic AI market, valued at USD 7.55 billion in 2025, is projected to explode to approximately USD 199.05 billion by 2034, growing at a CAGR of 43.84%. This isn't just growth; it's the acceleration phase of a paradigm shift. The evidence suggests we're past the hype peak and into the adoption ramp. By mid-2026, roughly 40% of enterprise applications will ship with some kind of agent baked in, a threshold that signals the technology is moving from pilot projects to foundational infrastructure.

This is the strategic lens of BNP Paribas. They see the wave not in the flashy new agents themselves, but in the enterprise software platforms that will become their essential rails. Their research identifies a clear competitive landscape. While cloud giants like Microsoft are the top beneficiaries of the AI application wave, global software leaders like ServiceNow and SAP are viewed as the most "resilient" companies in the industry. The thesis is that these enterprise software incumbents are building the critical, durable layers upon which the agentic AI paradigm will run. The question for investors is not just about who's innovating fastest, but who is becoming the indispensable infrastructure for the next computing paradigm.

Salesforce, Workday, Snowflake: The Infrastructure Layer in Action

The agentic AI wave is not just about smarter chatbots; it's about autonomous systems that own decisions and trigger actions. For these agents to function, they need a foundation. This is where Salesforce, Workday, and Snowflake come in. BNP Paribas notes these companies are accelerating the AI-driven enterprise, positioning them as the essential rails upon which the new paradigm runs.

These platforms provide the fundamental layers agentic AI requires. Snowflake offers the vast, governed data reservoirs that agents need to make informed decisions. Workday supplies the workflow and human capital management engine, giving agents the operational context to automate HR and financial tasks. Salesforce delivers the application layer for sales and service, where agents can now own customer interactions from discovery to payment. Together, they form a complete stack that transforms enterprise software from a toolset into an execution layer.

This deep integration is their competitive moat. Unlike new entrants, these companies are already embedded in the core operations of global businesses. Their resilience, as noted by BNP, stems from this entrenched position. As agentic AI moves from pilot to production, the cost of human oversight becomes a major friction point. Enterprises are shifting toward autonomous systems to cut labor costs. In this environment, the platforms that already manage the data, workflows, and applications become indispensable. They are not just software upgrades; they are the foundational infrastructure that enables agents to operate at scale.

The bottom line is that value capture is shifting. While cloud giants build the compute and model layers, the companies that own the enterprise data and workflow layers-Salesforce, Workday, and Snowflake-are positioned to capture significant value as agents automate tasks across sales, HR, and data analytics. They are the rails for the agentic AI train.

Financial Impact and the Long-Term Paradigm Shift

The market's recent volatility in software stocks highlights a classic tension between short-term noise and long-term infrastructure build-out. Shares of cloud banking software company nCino have fallen in recent weeks, mirroring broader declines across the SaaS industry. This sell-off is driven by investor fears that AI advancements will disrupt traditional software models. Yet, this reaction may be mistaking a temporary wave for a fundamental shift. The decline reflects market anxiety and a flight to perceived safety, not a re-rating of the underlying value of enterprise infrastructure.

The core risk here is the massive up-front investment required for AI infrastructure. As noted in the 2026 Investment Outlook, the desire to be first to market is creating an arms race that may result in overbuilding. This requires enormous capital for training models and running compute, with revenues and earnings expected only at some future date. This creates significant uncertainty about near-term returns on invested capital, a risk that echoes the dotcom era. However, a key mitigating factor is that today's leading cloud providers are large, rational companies with strong balance sheets and positive cash flow. They are self-funding their AI capex primarily through operating cash flows, unlike the debt-funded infrastructure plays of the late 1990s.

This is where fundamental adoption metrics become critical. The market is starting to reward companies demonstrating strong AI execution across their core products. Alphabet's stock provides a clear example. Its analyst fair value estimate has shifted modestly higher, supported by a modelled revenue growth rate of around 13.18%. This valuation support is directly tied to its AI execution, with recent research highlighting 34% Cloud growth as a key driver. The market is beginning to price in the exponential growth potential of the infrastructure layer, not just the hype.

The bottom line is that current SaaS declines may be a temporary overreaction to fears of disruption. The long-term paradigm shift is toward agentic AI that automates tasks across sales, HR, and data. Companies like Salesforce, Workday, and Snowflake are building the essential rails for this new era. While the path involves significant near-term investment and uncertainty, the fundamental adoption curve for enterprise AI is steepening. The market will eventually reprice these infrastructure plays based on their actual integration into business operations, not on short-term fears of disruption.

Catalysts and Risks: Navigating the Adoption Slope

The path from today's promising pilots to tomorrow's production systems is the critical slope to watch. The evidence shows a stark gap: while about three-quarters of organizations are testing agents, only roughly 11% of those pilots are currently moving to production. This is the make-or-break transition. The catalyst for BNP's thesis is the acceleration of this shift. When enterprises begin to scale these agents to automate real business functions, the value of the underlying infrastructure platforms-Salesforce for sales, Workday for HR, Snowflake for data-will become undeniable. The economic driver is clear: as systems move from pilot to production, the rising cost of Human-in-the-Loop review becomes a major friction point, pushing companies toward autonomous systems that cut labor costs.

The primary risk to this S-curve is the very momentum that fuels it. The desire to be first to market with leading AI models is creating an arms race that may result in an overbuilding of infrastructure. This requires massive up-front investment with revenues and earnings expected only at some future date, creating uncertainty about return on invested capital. While leading cloud providers are self-funding much of this through operating cash flows, the competitive pressure to build and deploy the most advanced models could strain financial discipline across the ecosystem. This "arms race" risk is a classic bubble precursor, where not all players will be successful, and the resulting overcapacity could compress margins for years.

The watch item for investors is concrete product development. Announcements from Salesforce, Workday, and Snowflake on integrated agentic AI features will signal the acceleration of the adoption slope. These aren't just incremental upgrades; they are the foundational building blocks that allow agents to own customer interactions, automate workflows, and analyze data at scale. When these platforms begin to bake agent capabilities directly into their core applications, it will mark the transition from enterprise software as a toolset to enterprise software as an execution layer. The market will reprice these infrastructure plays based on their actual integration into business operations, not on short-term fears of disruption.