AI Impact on SaaS Software Stocks: Deconstructing the Bottom-Fishing Logic of Salesforce, ServiceNow, and Snowflake

Compiled & Edited by: Deep Tide TechFlow

Guest: Nico

Original Title: SaaS Software Stocks in the AI Nightmare: CRM vs NOW vs SNOW, Who is the Real Misjudged Doubling Opportunity? A 10,000-Word Interpretation of the Next Wave of Software Stock Opportunities

Podcast Source: Nico's Frontier Alpha

Broadcast Date: May 21, 2026

Editor's Guide

Over the past half year, Wall Street has described a brutal sell-off as the 'SaaS Apocalypse'. Salesforce, ServiceNow, and Snowflake have been halved from their peaks. Meanwhile, JPMorgan's crowding model shows institutional holdings in the semiconductor sector have soared to 99.3%, while the software sector's crowding is only 22.8%, indicating a historic level of emotional polarization. At this juncture, investor Nico offers a judgment contrary to the mainstream narrative: AI is not here to kill the software industry; it is eliminating companies that sell only functional interfaces while rewarding those selling infrastructure and governance platforms. Although the software sector currently has weaker industry momentum than hardware, it offers higher odds and better value for money.

The most valuable part of this episode is placing the three companies under the same evaluation framework for detailed deconstruction. Salesforce (13-14x forward PE, $14.4 billion free cash flow, $50 billion buyback authorization) represents the 'Margin of Safety' faction; ServiceNow (AI Control Tower narrative, Jensen Huang's endorsement for three consecutive years) represents the 'Clearest AI Narrative' faction; and Snowflake (usage-based billing, RPO up 42% YoY, but still GAAP loss) represents the 'High-Beta, High-Risk' faction. On May 27th, Salesforce and Snowflake will release earnings on the same day, followed closely by Snowflake's annual summit and Microsoft's Build conference. These catalysts will form the most direct near-term observation window.

Highlighted Quotes

'SaaS Apocalypse' and Extreme Market Sentiment

• "The software sector has been hammered to pieces. This isn't a problem with one company, but the entire software sector being sentenced to death by the market."
• "JPMorgan's crowding model shows institutional holding crowding in the semiconductor sector has soared to 99.3%, while the software sector's crowding is only 22.8%, indicating a historic level of emotional polarization."
• "The good news for the hardware sector is that everyone is already invested, already priced in by the market; the bad news for software is that people have mostly sold off, leaving room for an upward rebound. Looking at industry momentum over the next 3 months, hardware will undoubtedly be stronger; but considering upside potential, odds, and value for money, software might actually be better."

AI's Impact on SaaS Business Models

• "Many functional interfaces that SaaS companies relied on for charging can now be prototyped into a usable demo in an extremely short time using AI, requiring no programming experience whatsoever. What the market is truly worried about is the collapse of scarcity and moats at the SaaS functional layer."
• "If one AI Agent can do the work of 10 people, then a company that originally needed to purchase 1,000 accounts now only needs 100. This is what Wall Street recently calls 'Seat compression'."
• "Agents don't need UI, dashboards, or beautiful interfaces; they only need data and APIs. This means SaaS software is being dimensionally reduced by AI, demoted from the main entry point for enterprise workflows to a backend for data storage."

Salesforce's Transformation and Valuation

• "Buying Salesforce is essentially not betting on a high-growth story with a multiple of tens times PE, betting it will ultimately succeed in AI transformation. It's a trade-off between intrinsic value and actual price; it is currently at a relatively undervalued position."
• "Agentforce changes the charging logic from 'per head' to 'per task'. Past revenue was tied to employee count; future revenue is tied to overall workload volume. As long as the task-based billing logic proves out, Salesforce can smoothly transition from a seat economy to a task economy."
• "Microsoft's Dynamics 365 plus Copilot is Salesforce's biggest medium to long-term threat. If salespeople in the future don't even open Salesforce, but let Copilot automatically update customer records in Outlook or Teams, Salesforce could regress from a workflow entry point to a backend database."

ServiceNow's AI Control Tower Strategy

• "ServiceNow's goal is not to rebuild another ChatGPT, but to become the governance layer, orchestration layer, and execution layer for enterprise AI Agents. Regardless of which AI an enterprise uses, as long as that AI enters enterprise processes, calls enterprise systems, and executes enterprise tasks, it must be governed and orchestrated through ServiceNow."
• "This positioning is similar to Apple's iOS. Apple doesn't create every app itself, but all apps run on iOS. ServiceNow aims to follow a similar path in the future."
• "Jensen Huang's exact words were: 'ServiceNow is essentially the enterprise operating system for the AI era.'"

Snowflake's Consumption Model Paradox

• "What Snowflake fears most is not customers not using it, but customers using it too efficiently. When enterprises find their Snowflake bills too high, they push engineering teams to optimize queries, compress storage, and even replace low-value tasks with open-source tools. This is the double-edged sword of the consumption model."
• "Snowflake's net revenue retention rate has dropped from 131% to 126%, then to the latest 125%. This is still healthy, but the downward trend indicates that expansion speed among existing customers is not as fast as before."
• "Snowflake is the fastest-growing among the three, with the most direct AI data infrastructure logic, and is naturally unaffected by traditional SaaS business models; but it is also the highest valued, faces the most intense competition, and has the weakest profitability. High odds, high risk."

Historical Analogy and Final Judgment

• "The narrative that 'AI kills software' is oversimplified. What's really happening is that AI is eliminating software that sells only functional interfaces, while at the same time rewarding platforms that sell infrastructure and governance. Not all software will be disrupted."
• "When the internet bubble burst in 2000, the market trend was 'the internet will kill all traditional companies.' But ultimately, those who survived weren't just internet companies; they also included traditional companies that were the first to embrace the internet and integrate these tools into their businesses. Twenty years later, the logic of this AI wave is the same."

SaaS Apocalypse and Counter-Signals

At the beginning of 2026, the narrative 'AI kills the software industry' ignited the entire US stock market. Since then, the software sector has been shrouded in a nightmare of being disrupted by AI. Even Microsoft, the software sector's leader, couldn't escape, falling over 25% intra-year at one point. Calculating from its historical peak, the maximum drawdown approached 40%, close to the decline during the 2022 US stock bear market. Past favorites like Salesforce, ServiceNow, and Snowflake have seen their market capitalizations halved. This isn't a problem with one company; it's the entire software sector being sentenced to death by the market. Wall Street named this event the 'SaaS Apocalypse'.

Over the past nearly half year, both retail and institutional investors have been doing the same thing: going long on hardware and shorting software. The software sector has been battered. However, recently, several unusual signals have quietly emerged. JPMorgan's crowding model shows institutional holding crowding in the semiconductor sector has soared to 99.3%, while the software sector's crowding is only 22.8%, indicating historic emotional polarization. And just at this time, US President Trump quietly spent several million dollars bottom-fishing software stocks; Wall Street's best bottom-fisher, hedge fund manager Bill Ackman, simultaneously took a large position in Microsoft, the largest software company; NVIDIA's CEO Jensen Huang, leading the world's most valuable company, personally flew to Las Vegas for the third consecutive year to endorse a software company.

So, is AI meant to kill the entire software industry, or is it giving us a once-in-a-decade bottom-fishing opportunity? In today's episode, I will deconstruct three of the most representative software companies from start to finish: Salesforce, ServiceNow, and Snowflake.

Claude Cowork and the SaaS Sector Collapse

The story about AI killing the SaaS industry and the software stock crash starts this January. On January 30th, Anthropic (the company behind the Claude large model) quietly released 11 plugins on GitHub called Claude Cowork, just a simple code repository plus a blog post. However, within 48 hours of release, global software stocks bled. According to market estimates, the software sector lost a total of $285 billion in market capitalization.

Why is everyone so panicked? A CNBC reporter conducted an experiment that kept all SaaS company executives awake at night. Using Claude Code, he spent an hour replicating a website called Monday.com for only $5-15. Monday.com is a publicly traded project management software company valued at tens of billions of dollars. A reporter, in one hour with a few dollars, created a project management demo that looks similar to Monday.com.

Of course, this doesn't mean he actually replicated a public company. The real Monday.com has enterprise permissions, data security, integration ecosystems, and sales channels – things AI can't handle in an hour, requiring time to accumulate. The most frightening aspect of this experiment is that many functional interfaces SaaS companies relied on for charging can now be prototyped into a usable demo in an extremely short time using AI, requiring no programming experience. Behind this story, what the market truly fears is the collapse of scarcity and moats at the SaaS functional layer. The traditional SaaS model charging per seat may not hold up under AI's impact. This also reflects the ambition of underlying AI model vendors – not just optimizing model performance but directly entering the application layer to carve up this massive cake.

SaaS Business Models and Two Layers of Panic

SaaS stands for Software as a Service. Its essence is simple: moving traditional on-premise software installed on enterprise servers to the cloud, where customers pay monthly or annually to obtain software usage rights. Over the past 20 years, this model has been the biggest wealth generator in the software industry.

The core billing logic for all SaaS companies is almost universally per seat. If a company has 1,000 employees needing the software, it must buy 1,000 accounts and pay ongoing subscription fees, ranging from tens to hundreds of dollars per account per year. The higher the frequency and longer the usage, the stronger the customer stickiness, as the entire company's workflows and data are deposited into this SaaS software, making migration and switching costs high in the short term. This is fundamentally the logic allowing the asset-light SaaS industry to make easy money and the core reason Wall Street was willing to give SaaS companies high P/E valuations of tens or even hundreds of times over the past 20 years.

However, with the AI wave, especially entering the Agent era, this logic's foundation has begun to shake. Market concerns about the SaaS industry mainly have two layers.

First Layer: Seat Compression

The most direct layer of panic is that Agents replacing employees lead to a sharp decline in SaaS subscription numbers and sharp reductions in revenue/profits. SaaS companies charge per seat based on how many employees use it. But with the Agent era's arrival, this logic is completely overturned. If one AI Agent can do the work of 10 people, a company that originally needed to purchase 1,000 accounts now only needs 100. This is what Wall Street recently calls 'Seat compression'.

The SaaS company revenue formula is 'Customer Count × Seats per Customer × Price per Seat'. Over the past 20 years, all three variables increased. However, under AI's impact, the 'Seats per Customer' metric faces structural downward risk for the first time. The market fears SaaS business models may be disrupted by AI.

Second Layer: Agent Workflows Bypass SaaS Interfaces

A deeper layer of panic is that under Agent-based workflows, SaaS software is directly bypassed and becomes a supporting player. This is the core reason the market truly freaked out. Traditional SaaS business models have an implicit premise: software is for people to use. Salesforce designs UI, beautiful dashboards, and workflows, essentially to cultivate user habits and increase stickiness. But Agents don't need UI, dashboards, or beautiful interfaces; they only need data and APIs.

When Claude can directly connect to your Salesforce, Notion, Google Drive, and Slack plugins, workflows fundamentally change. In the past, salespeople directly opened Salesforce to check customer data, follow up on contracts, view after-sales situations – daily work was inseparable from Salesforce's software interface. Now, salespeople can directly open Claude to complete these repetitive tasks, with Claude calling Salesforce via API to read and write data, without the salesperson ever touching Salesforce's software interface.

This means SaaS software is being dimensionally reduced by AI, demoted from the main entry point for enterprise workflows to a backend for data storage. The terrifying aspect is that it directly changes the value distribution chain. In the past, users interacted most with SaaS software; now, users spend more time interacting with Agents. Where users spend the most time holds the greatest pricing power. In this scenario, SaaS software becomes a supporting player to AI Agents. The strongest moat SaaS had – long-term user habits and workflow sedimentation – was fundamentally based on the premise that 'people heavily use UI interfaces,' but Agents are changing this. This is enough to trigger widespread market panic.

Market Crowding and Counter-Signals

Meanwhile, with a tight macro interest rate environment and big tech's capital expenditures almost entirely flowing into AI infrastructure, enterprise software procurement budgets are being squeezed, and long-duration software growth stocks' valuations are compressed the most. Year-to-date, the entire software sector has significantly underperformed the S&P and Nasdaq during the same period. The market has also become polarized, with everyone mindlessly going long on hardware and shorting software.

JPMorgan's crowding data analysis shows semiconductor industry crowding has reached a historical high of 99.3%, meaning almost all investor holdings are in the same direction. More notably, short positions in the software industry are steadily increasing, with the squeeze risk indicator reaching an extreme level of 100%. When panic reaches its peak, market inflection points and counter-signals often begin to appear.

This data doesn't mean funds will immediately withdraw from the hardware sector and shift to software. It's more of a risk signal – hardware has become the most crowded sector for both retail and institutional trading, making mindless long positions in hardware increasingly less cost-effective. Funds naturally have demand for sector rotation; switching from high hardware to low software is akin to moving from an extremely crowded, short-term fully priced sector to one still suppressed by 'ghost stories' but where fundamentals may improve.

The good news for the hardware sector is that everyone is already invested, already priced in by the market; the bad news for software is that people have mostly sold off, leaving room for an upward rebound. My judgment on this issue is clear: over the next 3 months, looking solely at industry momentum, hardware will undoubtedly be stronger; but considering upside potential, odds, and value for money, software might actually be better. In other words, hardware remains the main AI theme, but it's become too crowded in the short term; software is the catch-up direction with higher elasticity and odds over the next 3 months.

This is mainly because the software sector has been hammered too severely in recent months. With AI panic, software stocks saw widespread, indiscriminate selling – the market sold first and asked questions later, leading to many quality software companies with business moats, data sedimentation, and actively embracing AI being ruthlessly misjudged.

Moreover, in the coming dozens of days, the software sector has many catalysts. For example, on May 27th, Salesforce and Snowflake will release their latest earnings reports on the same day. These reports will answer a core question: is AI devouring SaaS or repricing SaaS? Immediately after, from June 1st to 4th, Snowflake holds its annual summit in San Francisco, themed on data infrastructure and enterprise AI implementation; on June 2nd-3rd, Microsoft holds its Build conference, focusing on AI Agents, Copilot, developer workflows, and enterprise AI applications. These catalysts combined may strengthen the software stock rebound trend. If the market begins to believe AI Agents aren't meant to kill software but to be implemented through software platforms, then software stocks like ServiceNow, Salesforce, and Snowflake may all benefit.

Company Deconstruction 1: Salesforce (CRM)

Company Background

Salesforce's ticker is CRM, matching its business name. It is the world's largest customer relationship management (CRM) software company and one of the most symbolic companies of the SaaS era. Simply put, it helps enterprises manage customers. But 'managing customers' here isn't just about salespeople opening a webpage and entering a few customer details; its real value is being the core system of record for enterprise customer data.

Who the customer is, which employees followed up, what products were bought, contract stages, after-sales complaints, marketing touches – the most critical data throughout the customer lifecycle is deposited in Salesforce. These are enterprises' most core customer assets. AI can help generate emails, summarize meetings, and automatically write sales scripts, but without a trusted customer database, AI wouldn't know how to do these things. This is Salesforce's most core position. AI may impact Salesforce's front-end functions but may not necessarily kill its core.

Salesforce is, on one hand, the most typical traditional SaaS company, directly facing Agent seat compression impact; on the other hand, it is the data foundation for many enterprise customers, not a small tool easily replaced. This is also the core entry point for our Salesforce analysis: is it an old-era software company about to be disrupted by AI, or a cash flow machine overly pessimistically priced by the market?

Salesforce currently has over 150,000 enterprise customers, from startups to Fortune 500. The company was founded by Marc Benioff in 1999. Benioff hails from Oracle, was its youngest vice president, and was an early protégé highly regarded by Oracle founder Larry Ellison. Later, he started his own company with a radical idea at the time: enterprise software shouldn't be sold on CDs for installation on customer servers but should run in the cloud with monthly or annual subscriptions.

This concept was very radical in 1999. Back then, traditional giants like Microsoft, Oracle, and SAP predominantly sold software to enterprises for on-premise deployment. Benioff alone shouted the slogan 'No Software.' Later, the SaaS business model indeed won, and Salesforce became synonymous with the SaaS industry.

Benioff's characteristic is keen嗅觉 and willingness to bet on directions. When he first mentioned Agentforce last year, the entire market thought it was a marketing gimmick. But over the past few quarters, Agentforce has indeed shown some promising data. The latest disclosure shows Agentforce's ARR has reached $800 million, up 169% year-over-year. So, whether you believe Salesforce can complete its AI transformation largely depends on whether you believe in Benioff himself.

Product Portfolio

Many think Salesforce is just a CRM tool, but after over 20 years of expansion and acquisitions, it has grown into a massive enterprise software platform.

The core is Sales Cloud, its founding product helping sales teams manage customers, opportunities, and sales funnels. A large portion of global enterprise sales systems are built on this product. After Sales Cloud, Salesforce expanded into Service Cloud,专门 for customer service and after-sales support. Customer phone complaints, email inquiries, online chat questions, backend ticket assignment, and processing workflows all run on Service Cloud. Further extension includes Marketing Cloud for digital marketing, helping enterprises with精准推送, email marketing, ad campaign tracking; Commerce Cloud for e-commerce, helping enterprises sell online.

Combined, these four areas basically cover all stages of enterprise-customer interaction, from acquisition, closing, after-sales to复购, with corresponding products for the entire chain.

But Salesforce's ambitions don't stop there. Over the past few years, it spent heavily on acquisitions: MuleSoft (for system integration; enterprises may use dozens of software systems simultaneously, MuleSoft is responsible for打通 data between these systems), Tableau (for data visualization and business analytics, turning CRM customer data into charts and insights), Slack (for internal enterprise communication and collaboration,类似 domestic Feishu or DingTalk office software). Last year, it acquired Informatica (for enterprise-level data management, helping enterprises clean, integrate, and govern scattered data).

Put together, Salesforce has actually built a complete ecosystem around customer data. CRM is the core, surrounded by layers of integration, analytics, collaboration, and data governance. Salesforce's newest emerging business, and the most critical piece, is Agentforce, the AI Agent platform launched last year and its most important card against AI impact.

Business Model: From Seat Economy to Task Economy

Salesforce's business model is the most typical SaaS – charging per seat. How many salespeople a company has needing CRM determines how many accounts to buy, each costing around $100+ per month, settled on annual contracts. Individual accounts may not seem expensive, but if a large enterprise has thousands or tens of thousands of sales,客服, and operations personnel, this money together becomes very stable recurring revenue. This is the fundamental source of Salesforce's easy money over the past 20+ years.

But with AI's arrival, this easy-money logic begins to loosen. If an AI Agent can automatically conduct customer research, write emails, manage sales funnels, and follow up with customers, does an enterprise still need that many salespeople? This is what the market fears most – seat compression. Salesforce is one of the most representative companies for the market to hype and discuss.

Benioff himself recognized this problem. Starting last year, Salesforce initiated a relatively aggressive yet crucial business model transformation: retaining seat fees but adding a usage-based billing product fitting the AI era called Agentforce. Simply put, the traditional model is 'pay for how many accounts you buy'; the new model is 'pay for how many tasks your AI Agent executes.' Salesforce calls this usage 'Agentic Work Units' (a计量单位 for AI Agent completing work).

The logic behind this new model is clever. If AI can indeed replace some human labor, traditional seat counts may decrease, but simultaneously, the number of tasks executed by AI Agents may大幅 increase. In the past, one salesperson might follow up with 20 customers a day; in the future, one AI Agent could simultaneously follow up with 200 customers. Human seats decrease, but AI-executed task numbers may double or even increase tenfold. As long as task-based billing logic proves out, Salesforce can smoothly transition from a seat economy to a task economy, potentially significantly increasing revenue per customer. Past revenue was tied to employee count; future revenue is tied to overall workload. This is Agentforce's most important significance – potentially重构 Salesforce's entire company billing logic and business model.

Of course, this story isn't fully realized yet. Although Agentforce's ARR has reached $800 million with very fast growth, it still represents less than 2% of Salesforce's $41.5 billion annual revenue. And the seat compression impact Salesforce faces may be more severe than any SaaS company because Salesforce sells seats for salespeople,客服, and marketers. A 10,000-person company might need to buy 3,000-5,000 Salesforce accounts, and these roles are precisely the ones AI Agents will replace first: writing emails, following up with customers, generating sales copy, answering customer inquiries – all things AI large models excel at. Relying on 2% new business to outpace traditional seat revenue decline is very difficult.

Given this, why do I still say Salesforce is worth attention? Not because I believe the Agentforce new business story will definitely outpace old SaaS model revenue, but because Salesforce currently trades at only 13-14x forward P/E, a valuation that already prices in pessimistic expectations. It also has $14.4 billion in free cash flow and a $50 billion buyback authorization.

So, buying Salesforce isn't essentially betting on a high-growth story with a multiple of tens times PE, betting it will ultimately succeed in AI transformation. It's a trade-off between intrinsic value and actual price; Salesforce is indeed at a relatively undervalued position. Of course, this margin of safety isn't unconditional. If AI truly causes significant traditional seat revenue decline and Agentforce can't补上, Salesforce's valuation may still be further compressed. But as long as core business remains stable and buybacks continue, even partial fulfillment of Agentforce could lead the market to revalue it, causing股价反弹.

Moat

Salesforce's strongest moat is the海量 data customers have deposited over the past 20+ years. A company using CRM for 10 years may have stored millions of customer records, hundreds of thousands of sales processes, tens of thousands of custom fields. Moving all this is equivalent to推翻 the entire enterprise digital foundation and rebuilding it; migration costs far exceed the cost of continuing to pay.

So where is Salesforce weak? Microsoft's Dynamics 365 plus Copilot is Salesforce's biggest medium to long-term threat. As the world's largest software company, Microsoft's to-B office products have penetrated绝大多数 large enterprises globally. Dynamics 365 is Microsoft's CRM product, directly对标 Salesforce's core business, with growth consistently above 20% over the past few years. Most crucially, Dynamics 365 deeply integrates with Copilot, Teams, Outlook, and other office套件. The most commonly used software entry points for enterprise employees are with Microsoft. If salespeople in the future don't even open Salesforce but let Copilot automatically update customer records in Outlook or Teams, Salesforce could regress from a workflow entry point to a backend database. This is what Benioff fears most and Salesforce's biggest medium to long-term uncertainty.

Latest Earnings Data

Last fiscal year's Q4 data: Full-year revenue $41.5 billion, up 10% YoY; Total RPO reached $72 billion, up 14% YoY; Free cash flow $14.4 billion, up 16% YoY; Full-year shareholder return $14.3 billion, with $12.7 billion用于 stock buybacks and $1.6 billion for dividends. Salesforce also just approved a $50 billion stock buyback plan. Agentforce新业务 ARR is $800 million, up 169% YoY, with 29,000 deals signed.

However, a caveat: 29,000 deals ≠ 29,000 large customers, nor entirely large contracts. This data only indicates rapid product rollout, but what truly determines valuation is whether it can subsequently increase付费金额 per customer and net revenue retention rate. In the last earnings call, the company also raised its FY2030 revenue target to $63 billion.

Overall, Salesforce's fundamentals are indeed very solid. Moreover, during the last earnings call, CEO Benioff himself said this was the company's most辉煌 year historically and the software industry's best year ever. He反而 said now is a great marketing and buying opportunity, so the company raised its buyback authorization to $50 billion. This tone is very clear – management is satisfied with earnings,甚至 directly countering the market, believing the market is overly pessimistic and Salesforce's stock is misjudged.

At the time of making this video, Salesforce's stock price was only $180, with a forward P/E of 13-14x. Compared to software bull market valuations of 30-40x+ in recent years, this is明显压缩了一大截, the lowest valuation in recent years.

Catalysts and Risks

Reasons to be bullish are simple: cheap valuation, stable cash flow, currently massive buyback力度, and Agentforce新业务 accelerating. Salesforce's May 27th earnings are值得关注, the most direct near-term catalyst.

Reasons to be bearish: its growth is only 10%,不算快 in the software industry; doubts about business model disruption by AI aren't eliminated; Agentforce新业务 uncertainty remains high. The market's biggest question: can Agentforce grow large enough to拉动 entire company revenue/profits and help complete full AI transformation? These仍需时间验证.

For the May 27th earnings, focus on these: First, whether Agentforce ARR maintains >100% YoY growth. If growth slows, it indicates AI转型 risks; mainly see how management responds.

Second, whether SaaS seat-fee related business shows明显萎缩. If类似情况 occurs, be cautious; the market may continue hyping 'AI吞噬SaaS' narrative.

Additionally, whether company forward guidance remains乐观, and whether management continues正面回应 AI's impact on SaaS business models. These are比较值得关注 areas.

Looking solely at last quarter's earnings, I think management was very clear and optimistic – they don't believe AI will kill Salesforce, but反而 believe AI will upgrade Salesforce from a SaaS application company to an enterprise Agent platform. But数据上, this story is still in early验证阶段. Personally, I think it's unnecessary to过早下结论 on whether it's disrupted by AI or completed AI business transformation. I value more its valuation being at the most undervalued level in recent years, combined with the company's solid fundamentals, making current buying性价比 and odds relatively high. But the long-term main narrative remains AI; whether Salesforce withstands AI's test还需要时间去验证.

Company Deconstruction 2: ServiceNow

Company Background

ServiceNow is the company I mentioned at the beginning – the one Jensen Huang personally flew to Las Vegas to endorse for three consecutive years. If Salesforce manages external customer relationships, ServiceNow manages internal employees and processes. Simply put, it's the central nervous system for internal enterprise operations.

Many internal enterprise processes requiring approval,流转, execution, and recording can run on ServiceNow. Computer坏了要找IT提工单; new employee入职要开账号,配电脑,走HR流程; system出故障要做事件响应; security告警来了要派发,升级,修复. So it's not just an IT ticketing system; it's more like a unified platform for various internal workflows.

ServiceNow was founded in 2004, headquartered in Santa Clara, California. The current CEO is Bill McDermott, former global CEO of SAP, with decades in the enterprise software industry. After officially taking over ServiceNow in 2019, McDermott led the company from an IT ticketing software company to further expand into a '全企业工作流平台'. His style is distinct,擅长讲大叙事,做大交易,搞大客户. This style becomes an advantage in the AI era.

Product Portfolio

The core founding business is ITSM (IT Service Management). Enterprise IT departments use it to manage tickets, incident response, change releases, IT assets, and service requests. In the ITSM market, ServiceNow is the undisputed global leader. On this foundation, it expanded into ITOM (IT Operations Management). ITSM is more about 'how to handle after problems occur,' while ITOM monitors systems proactively,发现问题,尽量自动修复.

Further business expansion includes HR Service Delivery – from入职,离职,请假,调岗 to various employee requests can be completed on ServiceNow. Also Customer Service Management业务 (enterprise-level客服, overlapping somewhat with Salesforce's Service Cloud, but ServiceNow leans towards complex B2B scenarios like large equipment, enterprise customers, cross-department售后工单); Security Operations for security incident response; Strategic Portfolio Management helping CIOs manage project portfolios, deciding which IT projects to invest in or cut.

Put together, ServiceNow has expanded from a simple IT service management software to an internal enterprise workflow platform. This is also the根本原因 for its 97% renewal rate – once an enterprise moves IT, HR, security,客服 workflows onto ServiceNow, replacing it isn't just changing software but重建一整套企业内部运转系统, a highly costly endeavor.

Recent Key Acquisitions

Besides native products, ServiceNow made several crucial acquisitions in the past year.

First is Moveworks, an AI-driven employee service assistant. Employees don't need to search for entry points; they directly ask AI, which can help查政策,提工单,看进度,甚至自动解决一部分问题. Post-acquisition, Moveworks capabilities integrated into ServiceNow's EmployeeWorks.

Second is Veza, mainly for identity governance and权限管理. In the AI Agent era, 'who can access what data' becomes extremely critical, not just for people but also Agents. Veza solves this.

Third is Armis acquisition, for real-time asset visualization in cybersecurity. How many devices in the enterprise network, which have vulnerabilities, which are communicating – Armis sees all.

These three acquisitions share a common指向 – preparing for large-scale AI Agent entry into enterprises. For Agents to work within enterprises, they need to know what employees are asking, who has permissions to动什么数据, and what assets are in the network. These three acquisitions respectively补齐这三块能力. Of course, consecutive acquisitions in a short time also bring integration risks, especially Armis's $7.75 billion大交易, which we'll detail in risks.

Core AI Strategy: AI Control Tower

ServiceNow's core AI strategy is called AI Control Tower. This concept starts from a practical problem. Future enterprises won't use just one AI; they may use OpenAI's GPT for customer service, Anthropic's Claude for contract review, Microsoft's Copilot for document collaboration, Google's Gemini for data analysis, and develop many internal AI Agents.

Here's the problem: with so many AI Agents running simultaneously within an enterprise, who manages them? Who decides what data they can/cannot access? Who ensures they don't overstep权限? How to assign责任 if accidents occur? This is what AI Control Tower solves.

ServiceNow's goal isn't to rebuild another ChatGPT but to become the governance layer, orchestration layer, and execution layer for enterprise AI Agents, responsible for making these AIs act safely, compliantly, and auditably within enterprises. This distinguishes it from many other SaaS software companies. Many companies think 'can we build an AI Agent ourselves to compete with ChatGPT, Claude, Gemini for application layer入口?' ServiceNow cleverly chooses another path: 'I won't compete with you for underlying models but manage execution processes after these models enter enterprises.'

ServiceNow aims to achieve: regardless of which AI an enterprise uses, as long as that AI enters enterprise processes, calls enterprise systems, and executes enterprise tasks, it must be governed and orchestrated through ServiceNow.

Why ServiceNow?

This returns to ServiceNow's底层能力 accumulated over 20+ years. It possesses something called CMDB (Configuration Management Database). Simply put, it's the complete map of enterprise IT asset and system relationships. Which servers the company has, which applications run, user权限 relationships – all recorded here. It also has a workflow engine running for over a decade – all internal approval, execution, collaboration链路 operate on ServiceNow. It has complete audit logs –每一步谁做了什么, when, changed what – the system can留下记录.

After AI Agents enter enterprises, they most need these three things: know what systems can be called, execute tasks according既定流程, and每一步都要留下审计记录. Additionally, ServiceNow通过 Veza补全了 identity and权限验证, through Armis补全了 real-time asset visualization.

At this year's Knowledge conference, this progressed further with ServiceNow releasing Action Fabric. This allows any third-party AI Agent – whether Claude, GPT, Gemini, or Copilot – to call ServiceNow's governance engine to execute enterprise-level tasks. 'I don't care which AI model you use, but execution and governance must go through my layer.' This logic is similar to Apple's iOS. Apple doesn't create every app itself, but all apps run on iOS. ServiceNow aims to follow a similar path.

Jensen Huang Endorsement

The most convincing背书 for this positioning comes from Jensen Huang. NVIDIA's CEO attended ServiceNow's annual conference for the third consecutive year – not just合作伙伴互相站台; NVIDIA itself is a ServiceNow customer. NVIDIA's internal supercomputer quoting system runs on ServiceNow; previously, generating a complete quote document took 5 days; with AI workflows, it's done in 5 minutes.

Huang's exact words: "ServiceNow is essentially the enterprise operating system for the AI era." This year, the two companies jointly launched Project Arc – NVIDIA provides AI compute security sandboxes, ServiceNow provides enterprise governance, a深度绑定 relationship. This indicates ServiceNow's AI Control Tower isn't an isolated software concept; it's entering the enterprise落地叙事 of AI生态伙伴 like NVIDIA, OpenAI, Google, Anthropic.

Latest Financial Data

Q1 this year: Total revenue $3.77 billion, up 22% YoY; Subscription revenue $3.671 billion, also up 22% YoY, exceeding guidance上限; Total RPO $27.7 billion, up 25% YoY; Customer renewal rate 97%. These numbers indicate ServiceNow's fundamentals are fine – still a software platform with ~20% growth, 97% renewal rate, high margins, high cash flow.

AI performance is even brighter. The company raised its ACV (Annual Contract Value) target for AI-related business from $1 billion年初 to $1.5 billion this year. Note this is contract value口径, not当期 revenue, converting to real revenue over time. But raising the target 50% within a quarter indicates its AI products truly have customer买单,高速增长.

Its股价 has retracted over 50% from历史高点, forward P/E now around 21-24x. For a high-growth轻资产 software company, this is indeed a relatively undervalued区间.

Catalysts and Risks

Reasons to be bullish on ServiceNow are clear. First, its AI叙事脉络 is very清晰 – AI Control Tower as the enterprise operating system for the AI era; greater AI demand means more need for governance, audit,权限, and execution platforms. Second, its AI新业务 indeed持续放量 – AI ACV rising from $1B to $1.5B, story truly兑现. Third, its生态伙伴阵营很强 – OpenAI, Google Gemini, Claude, NVIDIA all integrating or深度绑定合作 with ServiceNow, strengthening its strategic position as 'Enterprise AI Control Tower.'

But ServiceNow's risks must be清楚. After the latest quarterly earnings,即使超过市场预期, after-hours still fell double digits – market sentiment极度悲观, indicating the trend hasn't扭转 yet; skepticism about SaaS business models and AI转型 persists. Then, ServiceNow's three acquisitions密集落地, especially Armis's $7.75B大交易, needs time to消化; the market will scrutinize how much of the raised revenue指引 comes from acquisitions versus有机增长. External risks include中东地缘政治 factors – last quarter some large projects delayed, causing ~75bps不利影响 on subscription revenue growth.

Personally, I'm still看好 ServiceNow. Among the three, it has the smoothest, most straightforward, most marketable AI narrative. Its AI Control Tower positioning not only avoids AI impact but反而 benefits from AI普及, likely becoming the most critical software platform in enterprise AI落地. Valuation-wise, its股价 halved from highs over the past year, forward P/E low, similar to Salesforce, reaching a比较便宜 level – current buying性价比 and odds are quite good.

Company Deconstruction 3: Snowflake

Company Background

Simplest summary: the super warehouse for enterprise data. If Salesforce manages customers and ServiceNow manages processes, Snowflake manages data. All enterprise data – sales data, user behavior, financial reports, system logs –全部倒入 Snowflake, then analysis, modeling, running AI workloads can be done on this super data warehouse.

Product Portfolio

Snowflake's core foundation remains data warehouse and data lake. Enterprises dump all structured and semi-structured data here, running SQL queries, data analytics – Snowflake's根基, source of most revenue. On this foundation, Snowflake built Snowpark, allowing developers to directly write code in Python, Java, Scala within Snowflake,构建数据管道 and machine learning models without moving data out, completing the全过程 from data processing to model training inside the platform.

Moving up, Snowflake's重点推的 Cortex AI套件 over the past year+, with two core products. Snowflake Intelligence面向业务用户, allowing natural language conversations with data. Based on structured/unstructured data in Snowflake, it自动查询, analyzes, generates insights, can主动执行多步任务,更像是一个 enterprise-level AI Agent. Cortex Code面向开发者, different from普通编程助手在于 it's Snowflake-native AI Coding Agent, understanding Snowflake's data structures,权限设置, compute environment, directly helping write data pipelines, debug queries, build AI applications –功能非常强大.

So the分工很清楚: Snowflake Intelligence for business users, letting non-SQL people directly ask data, use data, let AI act based on data; Cortex Code for technical teams, letting developers/data engineers更快构建数据应用, data pipelines, AI applications.

Besides AI products, Snowflake has two比较独特 abilities. Snowflake Marketplace is a data sharing/trading market; enterprises can directly buy/sell datasets,调用第三方数据 for analysis. Data Clean Rooms enable privacy-preserving跨组织数据协作; two companies can jointly analyze without exposing各自原始数据. Advertising industry uses this for cross-platform attribution,医药行业 for联合临床研究,金融行业 for反欺诈协作. These two abilities are比较难复制的差异化优势.

Put together, Snowflake is transforming from a data warehouse tool towards an AI data platform direction –底层 data storage/compute,中间 development tools/AI engines,上层 business-user intelligent assistants/data market. Snowflake aims not just to帮企业存数据,查数据 but let enterprises analyze data, share data, develop applications on the same governed data platform, truly integrating AI into business data. Customer规模: Snowflake currently has 13,300+ enterprise customers;平台 daily processes 6.3 billion data queries.

Business Model

This is Snowflake's biggest difference from the previous two. Salesforce and ServiceNow core businesses charge per seat, fixed annual subscriptions; Snowflake完全不同, charging based on actual consumption of compute and storage resources – how many queries run, compute used, data stored, pay according to platform's计算公式.

This model has pros and cons. Pros: AI era enterprise data consumption指数级增长; every AI task背后消耗算力/data queries, Snowflake revenue naturally grows with AI workload暴涨. Cons: once enterprises cut budgets or optimize workloads, Snowflake revenue also跟着下跌.

However, Snowflake这两年也开始大力宣传 multi-year consumption commitment contracts. Latest earnings RPO $9.77 billion, up 42% YoY, indicating large customers开始锁定 future compute budgets to Snowflake for several years, not completely说走就走的关系.

Moat and Competitive Landscape

Strength lies in data黏性. After data倒入 Snowflake,上下游分析模型,查询脚本,数据管道全部建在上面; migration costs非常高. This is Snowflake's most核心 moat. Also, its Data Clean Rooms relatively成熟 in隐私保护,跨组织协作,不容易被复制.

Weakness: competitive landscape太激烈. Biggest competitor is Databricks – its latest annualized revenue run rate reached $5.4 billion, 65% YoY growth, double Snowflake's 29%; latest valuation round ~$100+ billion. Databricks stronger in machine learning/AI workloads. If Databricks未来上市, it likely becomes one of the most关注 IPOs in enterprise software市场, forcing Snowflake to接受正面对比 in public markets.

Besides Databricks,三大云厂商 threats不小. AWS Redshift, Google BigQuery, Azure Synapse持续进化,天然绑定 respective cloud ecosystems – they're both Snowflake partners and potential替代者. Further down, DuckDB, ClickHouse这些开源/新兴工具蚕食市场 in specific scenarios like local analysis, real-time analysis, low-cost queries. So Snowflake's竞争环境 is more复杂 than Salesforce/ServiceNow.

Counterintuitive Risk of Consumption Model

Another反直觉 point: Snowflake最怕的不是 customers not using it, but customers using it太溜. Because Snowflake's consumption model – more queries, compute, storage equals higher revenue. Conversely, when enterprises find Snowflake bills too high, they push engineering teams to optimize queries, compress storage,甚至 replace low-value tasks with open-source tools.

This is the consumption model's double-edged sword: when growth fast, revenue naturally rises with customer usage; but once customers start optimizing usage, revenue growth跟着慢下来. This trend already shows in数据: Snowflake's net revenue retention dropped from 131% to 126% to latest 125%. This数字依然健康, indicating existing customers still increase consumption annually, but downward趋势 indicates老客户扩张速度 not as fast as before. Behind this are both自然回落 after large基数变大 and customer cost optimization/consumption节奏放缓影响.

Thus, Snowflake更像是一个高增长,高弹性, but竞争强度极高的 AI data platform. This is Snowflake's biggest魅力, also its biggest风险.

Latest Financial Data

Full-year product revenue $4.47 billion, up 29% YoY – fastest growth among the three. Latest quarter product revenue $1.23 billion, up 30% YoY, slightly above full-year growth. RPO $9.77 billion, up 42% YoY. Latest quarter net new customers 740, up 40% YoY. Company also signed largest单笔合同 historically, value exceeding $400 million. These数据 indicate Snowflake demand hasn't slowed;相反,大客户 still signing larger长期合同.

But problems也很明显. GAAP口径, Snowflake full-year still lost ~$1.33 billion –唯一 GAAP-unprofitable among the three. Quarterly stock-based compensation ~$400+ million, full-year >$1.7 billion – shareholder dilution pressure不小.

But Snowflake remains最贵的 among three – forward revenue-based EV/Sales估值倍数 ~9x,明显高于 Salesforce.

Catalysts and Risks

Bullish points: Snowflake has several看点. First, not traditional SaaS but usage model,天然受益于 AI workload growth. Short-term, more AI runs equals more Snowflake赚得. Second, RPO up 42% YoY indicates大客户 still signing larger长期合同,代表未来收入可见性很强. Third, Snowflake Intelligence and Cortex Code快速扩展, 9,100+ accounts already using AI功能.

Additionally, Snowflake近期有两个比较重要 events: May 27th earnings,紧接着 June 1st-4th Snowflake年度大会 in San Francisco. Two催化紧挨, personally利大于空. During that period,股价波动应该会比较大.

Risks必须提前了解. First, GAAP持续亏损 biggest硬伤. In market preferring profitability/cash flow, compared to Salesforce/ServiceNow, Snowflake承受更大的估值压力. Second, Databricks currently Snowflake's最强烈 competitor; Databricks未来上市 may重塑整个 data platform赛道竞争格局. If post-IPO faster growth, stronger AI narrative, more attractive valuation, funds may流向 from Snowflake to Databricks. Also股东诉讼, insider减持 – these公司治理层面噪音 also affect market sentiment, though not当前的主线.

One sentence总结 Snowflake: among three, fastest growth, most direct AI data infrastructure logic,天然不受 traditional SaaS business models影响; but also highest valuation, most intense competition, weakest profitability. High odds, high risk.

Three-Way Comparison and Personal Conclusion

After deconstructing these three companies, I'll share my personal subjective看法.

If you value margin of safety, prefer value投资逻辑, Salesforce is相对最稳 – low teens forward P/E, $14.4B FCF, $50B buyback authorization, stable profitability –建仓 holding safety边际比较大. But growth only 10%,股价上涨爆发力可能没有那么强.

If you认可 AI Control Tower governance layer logic, ServiceNow may have clearest AI叙事 among three – >20% growth, 97% renewal, 22x forward P/E,加上 Jensen Huang's consecutive endorsements – current buying性价比还是挺高的. But前提: accept密集收购 integration risk,承担短期股价高波动.

If you pursue maximum弹性,也能承受最大波动, Snowflake is高赔率赌注 – biggest risk: company not profitable,持续亏损, net revenue retention下滑, competitor Databricks未来 IPO may重塑整个 data platform赛道估值锚. Risk波动确实比较高.

Besides these three, in software板块, if you want most stable压舱石, Microsoft remains最佳选择 – most严重错杀 large-cap software target this round. However, emphasize this is just my personal判断框架, not构成任何投资建议. Everyone must make相应投资决定 based on own actual仓位情况 after理性分析.

Conclusion: Who Does AI Kill?

Finally, back to the开头 question: Is AI meant to kill the entire software industry, or giving us a once-in-a-decade bottom-fishing opportunity?

My judgment: the 'AI kills software' narrative is过度简化. What's really happening: AI is淘汰那些只卖 functional interfaces software,但同时 rewarding platforms selling infrastructure and governance. Not all software will be颠覆.

This is好比 2000互联网泡沫破裂时 – market trend then was 'internet will kill all traditional companies.' But ultimately, survivors weren't just internet companies; they included traditional companies最先拥抱 internet, integrating these tools into their businesses,顺利完成了 internet转型. Twenty years later, looking at this AI wave, logic是一样的. Truly moated, data-sedimented software companies能够充当 AI infrastructure platforms will ultimately become biggest赢家. And now, they may正好站在新一轮上涨周期的起点.