Written by: Discussing with the poor Dao on K-line,潮向 Research

Today, Microsoft's annual developer conference, Build, opened in Fort Mason, San Francisco. Nadella's keynote speech conveyed a single message: AI is no longer an assistant that answers questions; it is now an employee working for you.

This conference falls at an interesting point in time. Over the past five months, the U.S. stock market's software sector has experienced a massacre.

The market dubbed this slaughter “SaaSpocalypse,” the SaaS apocalypse. From the beginning of the year to mid-May, Salesforce fell by 33%, Intuit by nearly 30%, and even Workday and Adobe were not spared. The logic of the panic is simple: AI agents can do the work of ten people, so companies no longer need to buy software seats for ten people. The per-seat pricing model has supported the entire SaaS industry’s business model for twenty years, and now the foundation has been pulled away.

But just last week, a group of people suddenly stood up in the slaughterhouse.

On May 28, Snowflake's stock surged 36.5% in a single day, marking its largest single-day increase since going public. Datadog's stock price has doubled this year and hit an all-time high on May 29. On the same day, MongoDB rose 10%, and Palantir rose 8%, with all three major indices hitting new highs.

Another group remained down. Intuit plunged 19% briefly after its earnings report. Salesforce's EPS beat expectations by 24%, yet its stock price still fell after the earnings report, down 28% for the year.

In the same massacre, some companies doubled their value while others were cut in half. What is the difference?

The Spark Lit by Snowflake

What allows Snowflake to light this spark? Because of its pricing model.

Over the past five months, the core of the market's panic centered around one very specific issue: per-seat pricing. The logic is simple: AI agents can do the work of ten people; therefore, companies do not need software seats for ten people. Atlassian reported a decline in enterprise seat numbers for the first time in history this year, and this fear is supported by real data.

Snowflake stands exactly opposite to this fear. It does not charge per person; instead, it charges based on the amount of computing power and data processing used. AI has not decreased its usage but has instead massively increased it: AI accounts on the platform grew from 9,100 to 13,600 in one quarter, the product revenue grew by 34% year-on-year, and the company raised its full-year guidance while announcing a $6 billion AWS computing power purchase.

Datadog tells the same story from a different angle. Snowflake demonstrates that "AI is generating volume for data platforms," while Datadog shows that "AI is generating volume for monitoring platforms." Q1 revenue broke through $1 billion for the first time, growing 32% year-on-year, with growth accelerating for three consecutive quarters (from 25% to 29% to 32%). Full-year guidance was raised to between $4.3 billion and $4.34 billion. The logic is simple: the more AI workloads companies deploy, the more they need to monitor and debug, and Datadog's usage-based pricing system spins faster. Its RPO (Remaining Performance Obligations) grew 51% year-on-year to $3.48 billion, indicating that customers are not only using it but are also signing longer-term contracts. Its stock price has doubled in price this year, reaching an all-time high on May 29.

A single phrase can summarize the logic of this rebound: AI is creating more workload for certain platforms, rather than replacing them. Snowflake and Datadog are the two cleanest examples.

In the Same Week, Another Face of the Market

If we only look at Snowflake and declare "software stocks are saved," we will fall into another pit.

In the same week, Salesforce released its Q1 earnings report, telling a story far more complex than "guidance is weak."

First, let's look at the positive aspects: Q1 revenue was $11.13 billion, growing 13% year-on-year, surpassing expectations; adjusted EPS of $3.88 exceeded Wall Street's expectation of $3.12 by 24%; the most critical indicator, Agentforce (its AI agent platform), reached an annual recurring revenue of $1.2 billion, growing over 200% year-on-year. The company processed 380 million agent work units, generating 286 trillion AI tokens. This is real AI monetization, not just a PPT.

Salesforce is even actively moving toward a "consumption-based" billing approach. It launched "Flex Credits," which no longer charges solely based on seats but rather based on the workload completed by agents. Six of the top ten deals in Q1 were bound by Flex Credits from the start. This company is desperately trying to cross the boundary between "per-seat" and "usage-based" billing.

Now look at the market's reaction: after the earnings report, the stock price fell in after-hours trading. As of last Friday, Salesforce was still down approximately 28% for the year. The reason is that the Q2 guidance is slightly lower than the most optimistic expectations, while Tableau and Commerce Cloud businesses are weak.

What does this indicate? It means that the boundary truly exists, but crossing it takes time. The market is willing to grant a 36% increase in a single day to a company (Snowflake) that has already crossed into the consumption-based side, but is unwilling to give credit to a company (Salesforce) that is trying to cross over. The willingness to transform does not equate to the completion of transformation.

Intuit is another counterexample. Its stock dropped nearly 19% after the earnings report. Its TurboTax tool, which charges per task and is aimed at individuals, is the most direct target of the “AI replaces human” fear.

Build Conference: Three Signals to Note

The Build conference is in progress, and there are more noteworthy things than expected.

Signal One: Microsoft is cutting down on its reliance on OpenAI.

Project Polaris announced at Build is Microsoft's self-developed AI programming model, which will replace GPT-4 as the default engine for GitHub Copilot in August this year. This model runs on Microsoft’s own Maia AI accelerator, meaning that from the model to the chip to developer tools, Microsoft has taken back the entire chain. The relationship between OpenAI and Microsoft has always had a commercial awkwardness, with both companies sharing user bases with overlapping interests. Polaris is Microsoft's formal answer to this issue.

Signal Two: Agents are no longer demos; they are becoming part of the operating system.

Agent Mode has become the default mode for Office 365 Copilot. When you open Word, Excel, or PowerPoint, AI runs as an "agent," capable of planning and executing multi-step tasks. The Windows Agent Framework has been open-sourced (MIT license), and the Windows Agent Store offers 85% revenue sharing to developers; Adobe and Zoom are already among the first partners. Nadella’s exact words were that AI has transformed from "synchronous assistants" to "asynchronous colleagues capable of independently executing long-term cross-domain tasks."

Signal Three: A $9.7 billion contract with the Pentagon.

The day before Build, the Pentagon announced a $9.7 billion five-year software integration contract, unifying Microsoft 365 subscriptions scattered among various military departments, intelligence agencies, and the Coast Guard into a single agreement. This money is not new spending; it consolidates previously dispersed procurements and renegotiates prices. But the signal it conveys is clear: at the world's largest single software buyer, Microsoft's seat model has not been weakened by AI; rather, it has been further secured.

How Should That Boundary Be Drawn?

Returning to the core question: Who has been rewarded in this rebound, and who has been neglected?

Software companies can be divided into four categories:

First Category: Consumption-based Platforms. Represented by Snowflake, Datadog, MongoDB, and Oracle Cloud. AI is generating more data processing, monitoring, and computing demands, and their billing models are spinning faster. Datadog, in particular, is wo

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。