Two software earnings reports landed in the same week and collectively produced the most substantive challenge to the SaaSpocalypse narrative that Wall Street has seen since the fear took hold. Snowflake (NYSE: SNOW) delivered its best day ever as a public company — a 36% single-session surge — after reporting product revenue of $1.334 billion, growing 34% year-over-year in a quarter that CEO Sridhar Ramaswamy called “the strongest sequential dollar growth in company history.” Net revenue retention improved to 126%, the first acceleration in several quarters. Non-GAAP operating margin expanded more than 300 basis points to 12%. Snowflake raised its full-year growth outlook from 27% to 31%.
Salesforce (NYSE: CRM), meanwhile, reported a 24% EPS beat and record non-GAAP operating margins — yet saw its stock fall roughly 3% after hours. Revenue of $11.13 billion grew 13% year-over-year. Agentforce ARR crossed $1 billion for the first time. The company processed 28.6 trillion tokens, up 152% sequentially, and delivered 3.8 billion Agentic Work Units, up 111%. And then Marc Benioff invited two customers onto the earnings call specifically to rebut the SaaSpocalypse thesis in real time.
The narrative — that AI agents will replace human software users, collapsing the per-seat subscription models that built the entire enterprise software industry — is not dead. But it just received its first credible, data-grounded challenge. The question for the next 6 to 12 months is whether this week’s evidence represents an inflection or an outlier — and the answer is different for each company.
What The Market Priced — & What It Got Wrong In Both Directions
The first-order reaction to Snowflake was straightforward. Product revenue accelerating four hundred basis points to 34% growth, NRR improving, record customer additions — those are clean, unambiguous signals that the AI data cloud thesis is not only intact but compounding. The market’s immediate 36% move reflected genuine re-rating from investors who had been pricing Snowflake as a structurally threatened business.
The first-order reaction to Salesforce was more revealing about how deeply the SaaSpocalypse narrative has taken hold. A 24% EPS beat, record operating margins, Agentforce ARR crossing $1 billion, and seven of the company’s top ten deals adding net new seats — and the stock fell anyway because Q2 revenue guidance came in $90 million below consensus. That $90 million miss, against $11.13 billion in quarterly revenue, was enough to send a stock down that had already fallen 33% year-to-date.
The divergence between the two reactions tells you something specific about market psychology. Snowflake was being priced for decline. Any evidence against that thesis triggered a violent repositioning. Salesforce was being priced for an AI monetization story that needed perfect execution to justify even its beaten-down valuation — and a guidance miss, however narrow, was treated as structural confirmation of the bear case.
This publication’s May 25 pre-mortem on Salesforce identified cRPO as the single most important metric, flagged the gap between 29,000 deals and $800 million in Agentforce ARR as the unanswered question, and predicted that the guidance direction would matter more than the headline beat. The cRPO came in at 14%. Agentforce ARR crossed $1 billion. The guidance missed by $90 million. All three pre-mortem calls landed — and the market’s reaction to the guidance miss was exactly the response that the pre-mortem’s bear scenario had described: a stock already priced for structural damage treating any shortfall as confirmation. The more important question is whether the underlying platform data — 28.6 trillion tokens, 111% sequential AWU growth, seven of ten top deals adding seats — changes the structural verdict, or simply buys time.
Snowflake: The Cleanest Counter-Narrative In Software
Snowflake’s quarter is the more analytically clean of the two. Product revenue growth accelerated for the third consecutive quarter. Net revenue retention, the metric that most directly measures whether customers are expanding usage rather than contracting, improved from 125% to 126% — the first improvement after a prolonged period of normalization from post-pandemic peaks. The addition of 616 net new customers — 38% more than the same quarter a year ago, and the most in company history — directly contradicts the narrative that AI was cannibalizing the need for enterprise data platforms.
The mechanism driving the acceleration is specific and instructive. Cortex Code — CoCo — launched into general availability on February 5, the opening day of Q1. CFO Brian Robins stated on the call that CoCo had “the largest driver to the increase in our forecast” and that the company had “a very unique opportunity to layer CoCo in the model” once observed consumption behavior was available. Critically, Robins confirmed there was no change to Snowflake’s guidance philosophy — they only forecast observed behavior. The $5.84 billion FY2027 product revenue guidance, up from $5.7 billion, reflects CoCo consumption that is already happening, not projected future behavior.
The second-order effect that the market is not yet fully pricing is that CoCo is simultaneously a revenue generator and a core platform accelerator. Ramaswamy was specific on the call: “Customers adopting CoCo are growing even faster,” and the product creates a flywheel where building on Snowflake via CoCo drives incremental consumption of the core data platform. Use cases deployed in the quarter grew 114% year-over-year. Use cases won per account executive grew 86% year-over-year. These are not vanity metrics — they measure the rate at which Snowflake is converting its installed customer base from passive data storage users into active AI workflow producers.
The $6 billion, five-year AWS infrastructure commitment — more than doubling the prior contract — signals something even larger. AWS has committed to an expanded go-to-market investment alongside the capital commitment. That is not a procurement decision. It is a strategic declaration that AWS views Snowflake as a core component of the enterprise AI stack, not a competitive threat to its own data services.
Salesforce: A Mixed Signal That The Market Over-Simplified
The Salesforce quarter resists a clean headline. The EPS beat was real and large. The margin expansion — non-GAAP operating margin at 34.8%, up 250 basis points, a record — was real. The Agentforce ARR crossing $1 billion was real and meaningful, answering directly the question our pre-mortem raised about whether the 29,000 deals were converting from pilots into production commitments. They are: top 10 AWU customers increased their total Salesforce spend by 1.5 times in the past year.
The most important data point in the Salesforce transcript is one that received almost no coverage: seven of the top ten Q1 deals added net new seats. CRO Miguel Milano said it explicitly: “This is a beautiful statistic.” It is the direct empirical refutation of the seat compression thesis that Bank of America’s $160 Underperform is built on. Human users are not being replaced by Agentforce agents in the top deals. They are being augmented — and the companies deploying agents are simultaneously expanding their human seat counts. PenFed Credit Union, which Benioff brought onto the earnings call, operates 76 agents alongside its human employees and describes its staff as “bionic employees” rather than employees being replaced.
The weakness in the quarter is also real and should not be minimized. Tableau bookings and renewals showed increased softness. Commerce Cloud and Marketing Cloud continued to drag. The Q2 revenue guide missed consensus by $90 million. These are not rounding errors in a thesis — they are evidence that the migration to the Agentforce platform is not lifting every application uniformly, and that the companies in Salesforce’s portfolio that have not yet been restructured around agentic workflows are experiencing real competitive pressure.
Benioff explicitly addressed the SaaSpocalypse on the call. He asked his own customer, Pallavi Mynampati of UCLA Health, the question directly: “We’re in the SaaSpocalypse. How do you look at it?” Her answer was not a denial of AI disruption. It was a reframing: the question is not whether AI replaces software seats, but whether AI makes the existing software more valuable. For UCLA Health, deploying Agentforce reduced the work burden on care teams without reducing the number of Salesforce licenses they need. That is the counter-narrative Benioff is building — and Q1 offered genuine evidence for it, alongside genuine evidence of parts of the business that haven’t yet turned.
The Names That Are Still Priced For The Old Narrative
ServiceNow (NYSE:NOW) has not yet reported a quarter that tests the SaaSpocalypse thesis with comparable evidence. The stock is down roughly 41% year-to-date — a steeper decline than Salesforce — and it carries a comparable per-seat exposure in IT workflow automation. The Snowflake and Salesforce prints provide a framework for what the market will expect when ServiceNow reports: acceleration in AI product consumption, NRR stability or improvement, and evidence that agentic workloads are expanding rather than displacing the human-licensed seat base.
HubSpot (NYSE: HUBS) sits in a structurally different position. Its customer base skews toward mid-market companies that have less institutional complexity and fewer sunk costs in legacy deployments — which makes them theoretically more vulnerable to AI-native alternatives that don’t require the full Salesforce or ServiceNow stack. The Salesforce evidence that AI adoption is driving seat expansion may not translate directly to HubSpot’s market, where the average deal size and switching cost calculus is fundamentally different.
Microsoft (NASDAQ: MSFT) is the most paradoxical name in this context. Its Copilot suite is simultaneously the most credible AI agent offering in enterprise software and the thesis that other software companies need to disprove. Snowflake’s CoCo explicitly supports “other data platforms like Databricks” and is built on model partnerships with both Anthropic and OpenAI — the same AI labs whose tools power Microsoft Copilot. The enterprise AI market is not winner-take-all, and the evidence from both Snowflake and Salesforce this week suggests that the companies with the deepest proprietary data moats are the ones most likely to compound their advantages in the agentic era, regardless of which frontier model is doing the reasoning.
Long-Term Structural Implications: Two Different Versions Of The Same Story
Snowflake and Salesforce are making the same fundamental argument through different architectures. Both companies are claiming that the value of AI is not independent of where enterprise data lives — that a general-purpose AI model with access to a company’s proprietary data, business context, and workflow history is structurally more valuable than a general-purpose AI model without it. Snowflake calls this the “agentic control plane.” Salesforce calls it the “#1 agentic CRM.” Both are describing the same competitive moat: the proprietary data layer that makes AI outputs enterprise-grade.
The structural risk that both companies face is also the same: what happens when the AI models themselves become so capable that the data preparation, governance, and workflow management layer becomes commoditized? Ramaswamy addressed this directly on the Snowflake call, arguing that “deep infrastructure capabilities just take a lot of time to develop” — citing role-based access control at scale, enterprise replication, and governance frameworks as capabilities that cannot be quickly replicated by AI labs or hyperscalers. Benioff made the same argument implicitly by building Headless 360 — making Salesforce’s data and workflow context accessible via MCP to any AI agent, regardless of which lab’s model is running the reasoning.
The SaaSpocalypse narrative rests on a specific version of the future where AI models become so good that enterprises no longer need the specialized data infrastructure these companies provide. Both transcripts this week suggest a different version: AI models become so good that the value of the proprietary data layer increases, not decreases, because better models can extract more value from better-structured, better-governed enterprise data. Which version of the future is correct will determine whether this week’s results represent an inflection or a reprieve.
Final Thoughts: The Narrative Is Cracking — The Verdict Is Not Yet In
The SaaSpocalypse narrative has not been defeated this week. Tableau is softening. Commerce is softening. The Salesforce guidance miss was real, however narrow. The structural question about per-seat models remains legitimately open.
But the narrative has been credibly challenged in ways it had not been before. Snowflake’s CoCo launched a new monetization vector mid-quarter and accelerated the company’s full-year growth outlook by four percentage points in ninety days. Salesforce’s Agentforce ARR crossed $1 billion while seven of its ten largest deals simultaneously expanded seat counts. Both companies’ management teams described AI as an accelerant to core platform consumption rather than a substitute for it.
The companies most exposed to the lingering narrative risk are those that have not yet reported a comparable counter-data quarter. ServiceNow is the most significant. The monitoring signals over the next 90 days are specific: whether ServiceNow’s AI product consumption mirrors the Snowflake/Salesforce pattern, whether Salesforce’s H2 organic revenue reacceleration materializes as Robin Washington committed it would, and whether CoCo adoption continues compounding as the primary driver of Snowflake’s raise-and-accelerate cycle. The SaaSpocalypse was always a hypothesis about the future. The hypothesis just had its first bad week.
Disclaimer: We do not hold any positions in the above stock(s). Read our full disclaimer here.
