Salesforce (NYSE:CRM) just handed software investors one of their biggest relief rallies of 2026. The Salesforce Snowflake SaaS Apocalypse debate is now moving to its next major test. Shares jumped 22.6% Thursday after better results, stronger bookings, and an expanded Anthropic partnership. CEO Marc Benioff even declared, “This is not the SaaSpocalypse.”
That matters after a brutal software selloff erased roughly $2 trillion in market value by some estimates. Investors had feared AI agents could replace traditional applications and weaken seat-based software economics.
But Salesforce’s headline profit deserves a closer look. A huge investment gain helped produce its eye-catching earnings number. That makes Snowflake (NYSE:SNOW), which reports on September 2, an especially interesting follow-up test. Snowflake does not have a giant frontier-model investment gain supporting its earnings story. Its case rests much more directly on whether AI increases demand for enterprise data, governance, and computing.
Salesforce’s Blowout Had A Huge Non-Software Assist
Salesforce reported non-GAAP earnings of $5.90 per share, up 103% from a year earlier. At first glance, that looks like a dramatic acceleration in software profitability. However, $2.53 of that $5.90 came from gains on strategic investments. Salesforce recorded $2.61 billion of net strategic investment gains during the quarter.
The largest driver was Anthropic. Salesforce’s regulatory filing showed about $2.7 billion of unrealized gains tied to that investment. Anthropic had been valued at $965 billion in its May funding round. Strip out the investment contribution, and adjusted EPS falls to roughly $3.37. That represents growth of about 16%, rather than the reported 103%.
That distinction matters for the broader software debate. Salesforce’s rally was partly a software victory and partly an investment windfall. The Salesforce Snowflake SaaS Apocalypse discussion therefore needs to separate operating progress from investment-driven profit. The Anthropic gain does not mean the operating business was weak. It simply means the headline EPS figure overstated what Salesforce earned from selling software and related services.
Snowflake will not have the same type of cushion on September 2. That makes its report a more direct read on operating demand.
The Operating Story Still Challenges The Salesforce Snowflake SaaS Apocalypse
Removing the Anthropic gain does not erase Salesforce’s underlying progress. In fact, several operating metrics directly challenge the idea that AI is simply destroying incumbent software vendors.
Revenue reached $11.35 billion, up 11% year over year. Current remaining performance obligations reached $33.5 billion, growing 14% in constant currency. Management also said net new annual order value growth was the strongest in four years. Customer attrition remained near historic lows.
AI adoption was also measurable. Agentforce ARR reached $1.5 billion, while Agentforce plus Data 360 ARR approached $3.9 billion. That combined figure grew more than 210%. Agentic work units reached 3.2 billion during the quarter, up 97% sequentially.
Salesforce also reported increased seats across Agentforce Sales, Service, and Slack. Benioff said agentic platform usage through MCP and command-line calls increased sixfold. The Salesforce Snowflake SaaS Apocalypse debate is therefore being tested against actual usage rather than theory alone.
There was one important warning. CFO Robin Washington cited license revenue headwinds and volatility in integration and analytics. That area sits closer to Snowflake’s competitive neighborhood. So Salesforce weakened the broad SaaS-apocalypse argument, but it did not eliminate every concern surrounding data and analytics software.
Snowflake Is The Cleaner Test Of AI Economics
Snowflake enters September 2 with a much simpler question: Is AI actually causing customers to consume more data infrastructure?
So far, management says yes. First-quarter product revenue reached $1.334 billion, up 34% year over year. Growth accelerated from 30% in the previous quarter. Net revenue retention increased to 126%, while remaining performance obligations grew 38%. Snowflake also raised its fiscal 2027 product-revenue growth outlook to 31%.
Management specifically connected that acceleration to AI. Cortex Code, also called CoCo, had already reached more than 7,100 accounts. Accounts using Snowflake Intelligence more than doubled sequentially. Snowflake said these AI products are generating revenue directly while also increasing core platform consumption. The Salesforce Snowflake SaaS Apocalypse thesis therefore comes down to whether that consumption continues to compound.
That distinction is important. Traditional SaaS companies often charge according to seats. Snowflake is largely consumption driven. More AI activity can therefore create more queries, pipelines, applications, agents, and data processing.
Snowflake had 13,912 customers at the end of Q1. It added 616 net new customers during the quarter.
September 2 will show whether that AI-driven consumption pattern continued into another quarter.
Trusted Enterprise Data Could Become The Real AI Bottleneck
Frontier AI models are becoming more capable, but enterprises still face a basic problem. An agent cannot do useful corporate work without reliable corporate context.
That is where Snowflake’s positioning becomes relevant. Companies already store sensitive revenue, customer, operational, and financial information on its platform. Snowflake argues this governed information gives AI systems the context needed to perform useful work. Management describes its role as an “agentic control plane” connecting data, models, applications, permissions, and workflows.
Its partnerships reinforce that strategy. Snowflake expanded a $200 million relationship with OpenAI during the first quarter. The aim is to make leading models available around governed enterprise information inside Snowflake. That partnership adds another layer to the Salesforce Snowflake SaaS Apocalypse debate because the models may depend on the software platforms holding trusted business data.
Reuters added another timely example this week. Its news and multimedia archive is now available through Snowflake Marketplace for enterprise AI applications. The archive stretches back to 1987 and covers five languages. Companies can access it without maintaining separate copies of the underlying datasets.
That points toward the larger issue. AI may reduce the value of some interfaces while increasing the value of trusted data underneath them. Snowflake’s next results should provide another data point on whether customers are spending accordingly.
Final Thoughts
Salesforce gave software investors real evidence that the SaaS apocalypse narrative may have been too broad. Its AI products are growing, bookings improved, and customers are still expanding usage. Yet its spectacular earnings headline also received an unusual boost from Anthropic.
Snowflake now faces a different standard. Its operating numbers must support the idea that AI increases demand for governed enterprise data. There is no comparable investment gain to blur that test. That is why the Salesforce Snowflake SaaS Apocalypse argument now matters beyond one company’s quarterly results.
Valuation also leaves little room for weak execution. Snowflake recently traded at 22.34x LTM enterprise value to revenue, 22.67x LTM sales, and 33.28x LTM enterprise value to gross profit. Its trailing EBITDA, EBIT, and earnings remain negative, making those profit-based multiples less useful.
For perspective, Salesforce traded around 5.39x LTM EV/revenue and 4.70x LTM sales, while its LTM EV/EBITDA stood near 17.70x. Its trailing P/E was about 23.36x.
That gap shows what the market already expects. Snowflake carries a substantial growth premium. September 2 therefore needs to answer more than whether AI is helping. Investors will also be watching whether the pace of AI-driven consumption can support a valuation that already assumes significant future growth.
Disclaimer: We do not hold any positions in the above stock(s). Read our full disclaimer here.
