Software stocks plunged 20%—but I have significant positions in Adobe and Salesforce.

Author: Ed Elson; Translator: Deep Tide TechFlow

Last week, the software industry lost $1 trillion in market value, falling 14% in a single week and approximately 20% year-to-date. Shopify, Atlassian, Salesforce, Adobe, and other major players all experienced sharp declines.

The reason? Anthropic released Claude Cowork and its plugins, and OpenAI released a similar tool. Investors panicked and sold off, believing that "AI killed software."

But Ed Elson believes this is irrational panic. We've seen this before: when ChatGPT appeared in 2022, Google fell 40%; when TikTok appeared, Meta fell 70%; when DeepSeek appeared, Nvidia fell 30%.

The result?

These companies have risen 630%, 270%, and 55% respectively from their lows. On Thursday, he bought Adobe, Salesforce, ServiceNow, and Microsoft, citing high switching costs, strong AI integration capabilities, and extremely low valuations. The full text is as follows: Last week we witnessed what was called a massacre. $1 trillion in market capitalization was wiped out. The sell-off reached an all-time high. Companies lost 10%, 20%, and 30% of their value. If you're looking at retirement accounts, you might not have noticed: the S&P 500 only fell slightly. That's because the massacre happened in a very specific place—an industry that has dominated portfolios for decades and many considered invincible: software. All the biggest names are in freefall: Shopify, Atlassian, Salesforce, Adobe, and the list goes on. The software industry lost about 14% of its value in just one week. Year-to-date, that figure is currently around 20%. Why is this happening? Because of AI. A few weeks ago, Anthropic released a new AI tool, Claude Cowork. Then (last week) they released new plugins for specific sectors: legal work, sales, finance, marketing, etc. OpenAI quickly released similar tools. Investors are quickly asking themselves an important question: Isn't this what every traditional software company does? Then: Has AI just killed software? Finally, their conclusion: Sell everything. This seems familiar. We've seen this movie before. In 2022, an AI tool called ChatGPT swept the internet. Investors asked themselves a crucial question: Isn't this what Google did? Within months, Wall Street decided search was dead. Google lost up to 40% of its value that year. Before that, a social media app called TikTok arrived. Investors asked themselves: Isn't this what Meta did? Once Meta reported a decline in users, $230 billion in market capitalization vanished—the largest 24-hour sell-off in stock market history. Meta continued to lose up to 70% of its value. Recently, a Chinese AI model called DeepSeek has become popular. Investors asked themselves: Isn't this what OpenAI did? OpenAI isn't publicly traded, so the sell-off wasn't visible. However, fear reverberated through the public markets. Nvidia lost 30% of its value in the following months. Since these market crashes, Nvidia, Meta, and Google have risen 55%, 270%, and 630% respectively from their lows. DeepSeek wasn't the domestic AI killer investors thought it was. After TikTok, Meta learned its lesson and launched its own version, Reels, now boasting a 2 billion active user base. After ChatGPT, Google doubled down on AI, eventually launching Gemini, ChatGPT's fastest-growing competitor. Google is now considered the undisputed AI heavyweight champion. The pattern here is simple. A transformative technology arrives. Investors indiscriminately decide "it's over." Their judgment of the technology isn't wrong, but they overestimate its impact. They panic sell, assuming the game is zero-sum. Valuations plummet. Suddenly, America's greatest company is half-price. Meanwhile, they continued to deploy armies of talent and capital to sharpen their focus and neutralize the competition. Earnings grew even more, and valuations soared again. Years later, when we look back at the charts, we think: What were we thinking? That is, all of us who sold off.

There was panic, and then there was this

I believe what happened in the software industry last week was no different. This wasn't a correction; it was a full-blown crisis. Let me draw you a picture: The Relative Strength Index (RSI) is a formula that captures buying and selling pressure. An RSI score of 30 means a stock is oversold. Last week, the average RSI for software stocks reached 18. I don't usually like technical analysis, but in this case, it describes what we saw very well: the end of the world. On one hand, the concerns are valid. Will AI disrupt software? Yes. Will it put pressure on profit margins? Absolutely. Will SaaS companies have to rethink their distribution? Without a doubt. On the other hand, this isn't what the market told us last week. The market told us that software is over—no matter who you are or what you sell. This position is more questionable. While I was initially willing to listen, I concluded that it doesn't come from rationality, but from fear. In other words: it's irrational. Reality Test First, nothing can stop software companies from integrating AI. The ChatGPT and Google story is a perfect example. Just because OpenAI has a more exciting product doesn't mean Google is dead. Google simply enhanced their existing products with AI capabilities (Google Search is now the dominant AI interface in the US) and then built its own AI chatbot. Short sellers might have a point if SaaS companies were simply ignoring AI, but they aren't. Software companies are fully embracing AI. Secondly, investors underestimate how painful it is to cancel an enterprise SaaS contract. In more business terms: switching costs are high. The average software sales process can take more than six months to finalize and requires approval from ten different decision-makers. This process is arduous because contracts are long. And that doesn't even mention the associated financial costs. For example, a typical Salesforce contract cannot be canceled for free—the company must pay 100% of the remaining contract value. In other words, switching your software provider as an enterprise is a very big deal. If you're going to do it, you'd better have an extremely compelling reason why it's worthwhile. Furthermore, all other executives in the company must agree with you. Pain. Really. Pain. Finally, when it comes to enterprise software, security is a huge issue. Signing a software agreement essentially means handing over all your private data to a third party and praying they won't lose, use, or misuse it. In other words, it requires trust. This is a top priority for 80% of IT leaders. More importantly, trust can't be coded in a day. Trust must be built over years, even decades. It requires long-term relationships and a substantial track record of success. These are things traditional companies have that Anthropic doesn't. Trust and security are a huge advantage that cannot be ignored. It's Time to Buy By Thursday afternoon, I'd had enough. I had two voices in my head. 1) Warren Buffett, who told me to be greedy when others are fearful. 2) Mark Mahaney, who told me to find "DHQ" (high-quality companies out of the ordinary). I decided it was time to buy and gave myself two options. Option 1: Buy the entire software basket. I looked at IGV, an ETF that tracks all the major software names, and it's been crushed. There might be a few losers in it, but the average multiple has fallen to a level where I feel I really can't be wrong. This is a safe option. Option 2: Stock picking. That is, personally identifying a small number of software names that I believe are high-quality companies. This is a riskier option because I risk making mistakes and picking losers. Nevertheless, I chose Option 2 because I felt bold. DHQ (Misaligned High-Quality Companies) On Thursday morning, I bought three stocks: Adobe, Salesforce, and ServiceNow. After that, I bought one more: Microsoft. Note: I am not a financial advisor, this is not financial advice—I'm just telling you what I did. My reasoning is as follows. 1. Adobe Adobe currently has a P/E ratio of 16, less than half its five-year average. It's also almost half the average P/E ratio of the S&P 500. It's very cheap. The consensus is that AI will make it irrelevant, but this ignores two key facts. 1) Adobe is already heavily integrating AI. In fact, its AI capabilities already generate over $5 billion in annual recurring revenue, more than half of Anthropic's ARR. 2) Its moat is enormous. Over 98% of Fortune 500 companies use Adobe, and like other software solutions, it's deeply integrated into the entire creative workflow, making it difficult to switch solutions. It's so ubiquitous that most digital creative roles list Adobe proficiency as a job requirement. An additional tailwind is short video. Adobe Premiere Pro is the industry standard for video editing, and most media companies (including ours) are significantly increasing their short-video budgets as the medium continues to explode. 2. Salesforce Salesforce is another AI-enabled company considered dead. Meanwhile, its AI agent product's ARR quadrupled last quarter, and the company continues to be ranked the industry's most trusted CRM. It has fallen more than 40% in the past year, its P/E ratio is now below the S&P average, and its cash flow P/E ratio is about half its 5-year average. Even if Claude had a more interesting product, I don't believe it would outweigh the enormous switching costs—and certainly not in the time it might take Salesforce to build a comparable product. 3. ServiceNow ServiceNow has been hit hard this year—down about 30% by 2026. The consensus is that growth is coming to an end. Meanwhile, its fundamentals tell the opposite story: subscription revenue grew 21% last quarter, and total revenue grew 20%. As for its AI capabilities, ServiceNow is more than capable. In fact, the company is on track to generate $1 billion in revenue from its AI products this year. It has also signed multi-year partnerships with OpenAI and Anthropic—more evidence that the AI ​​revolution is not a zero-sum game. I believe OpenAI and Anthropic will grow significantly this year, and so will ServiceNow. 4. Microsoft If you listened to yesterday's podcast, you'll notice I didn't mention Microsoft. That's because I hadn't bought any at the time of recording. My initial view was that I didn't need Microsoft because my exposure was already quite large (MSFT represents 5% of the S&P 500). However, upon reflection, I decided the valuation was too cheap to ignore. At the time, Microsoft's P/E ratio was only 25, the lowest among the Mag 7. This was relatively absurd for the reasons I emphasized above, especially for another key reason: Microsoft owns nearly a third of OpenAI. Even if Microsoft's lunch is eaten (which I doubt), the company has contractual rights to compensation. Few companies are better positioned in AI than Microsoft. The current price doesn't reflect this.

Efficient Market Hypothesis

In most cases, I believe in the efficient market hypothesis—the concept that markets reflect all available information and are smarter than any individual. I have great respect for the market's predictive ability (especially after they correctly predicted 93% of Golden Globe winners). I don't claim to be smarter than them.

However, I also believe that every now and then, something extraordinary happens—political events, natural disasters, global pandemics, or indeed, the arrival of transformative technologies. In these situations, I believe markets can lose their minds. When that happens, the efficient market hypothesis fails for a short period of time.

I risk being wrong and losing money here. But that's what being an investor means.

Besides, if you're taking risks every now and then, then... what's the fun in that? See you next week!