Author: Ed Elson

Translated by: Deep Tide TechFlow

Deep Tide Introduction: Last week, the software industry evaporated $1 trillion in market value, plummeting 14% in a single week, with a year-to-date decline of approximately 20%. Major brands like Shopify, Atlassian, Salesforce, and Adobe all plummeted.

The reason? Anthropic released Claude Cowork and its plugins, and OpenAI also released similar tools. Investors panic-sold, believing that "AI killed software."

But Ed Elson believes this is irrational panic; we have seen this movie before: when ChatGPT appeared in 2022, Google dropped 40%, when TikTok emerged, Meta dropped 70%, and when DeepSeek appeared, Nvidia dropped 30%.

The result? These companies surged 630%, 270%, and 55% respectively from their lows. On Thursday, he bought Adobe, Salesforce, ServiceNow, and Microsoft, reasoning that: high switching costs, strong AI integration capabilities, and extremely low valuations.

The full text is as follows:

Last week we witnessed what is called a massacre. $1 trillion in market value was wiped out. Sell-off activities reached historical highs. Companies lost 10%, 20%, 30% of their value. If you were looking at retirement accounts, you might not have noticed: the S&P 500 index only slightly dipped. That’s because the massacre occurred in a very specific place—a sector that has dominated portfolios for decades and many consider invincible: software.

All the biggest names went into free fall: 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, this figure is currently about 20%.

Why did this happen? Because of AI. A few weeks ago, Anthropic released a new AI tool called Claude Cowork. Then (last week), they launched new plugins for specific domains: legal work, sales, finance, marketing, and more. OpenAI quickly followed with similar tools.

Investors soon asked themselves an important question: isn’t this exactly what every traditional software company does? After that: has AI just killed software? Their conclusion: sell everything.

Familiar Fright

We have seen this movie before. In 2022, an AI tool called ChatGPT swept the internet. Investors asked themselves an important question: isn’t this what Google does? Within months, Wall Street decided that search was dead. Google lost up to 40% of its value that year.

Before that, a social media application called TikTok arrived. Investors asked themselves: isn’t this what Meta does? Once Meta reported a decline in users, $230 billion in market value disappeared, marking the largest 24-hour sell-off in stock market history. Meta went on to lose up to 70% of its value.

Recently, a Chinese AI model called DeepSeek became popular. Investors asked themselves: *isn’t this what OpenAI does? OpenAI is not publicly traded, so the sell-off was not visible. However, the fear echoed into the public markets. Nvidia lost 30% of its value in the following months.

Since these market-crash events, Nvidia, Meta, and Google have rebounded from their lows by 55%, 270%, and 630% respectively. DeepSeek is not the domestic AI killer investors thought it might be. After TikTok, Meta learned its lesson and launched its own version, Reels, which now boasts a user base of 2 billion active users. After ChatGPT, Google doubled down on AI, eventually launching Gemini, the fastest-growing competitor to ChatGPT. Google is now considered the undisputed heavyweight champion of AI.

The pattern here is simple. Transformative technologies arrive. Investors indiscriminately decide it's "over." Their judgement on the technology is not incorrect, but they overestimate its impact. They panic-sell, assuming the game is zero-sum. Valuations plummet. Suddenly, America's greatest companies are half-priced. Meanwhile, they continue to deploy armies of talent and capital to sharpen focus and neutralize competition. Earnings growth is larger, valuations soar again. Years later, we look back at the charts and wonder: what were we thinking? That is to say, all the panic sellers among us.

There’s Panic, and Then There’s This

I believe what happened in the software industry last week was nothing new. This is not a correction, but a full-blown crisis. Let me paint you a picture: the Relative Strength Index (RSI) captures buy/sell pressure through a formula. An RSI score of 30 indicates a stock is oversold. Last week, the average RSI of software stocks hit 18. I usually don’t favor technical analysis, but in this case, it accurately describes what we saw: apocalyptic.

On one hand, the concerns are justified. Will AI disrupt software? Yes. Will it put pressure on profit margins? Absolutely. Do SaaS companies need to rethink their distribution? Undoubtedly.

On the other hand, that is not what the market told us last week. The market told us software was over—no matter who you are or what you sell. This position is more dubious. Although I was initially willing to listen to the dissent, I concluded it did not come from a place of reason but from a place of fear. In other words: it is irrational.

Reality Check

First, nothing can stop software companies from integrating AI. The story of ChatGPT and Google 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 primary AI interface in the U.S.), and then built their own AI chatbot. If SaaS companies were just ignoring AI, then the bears might have a point, but they are not. Software companies are fully embracing AI.

Second, investors underestimate how painful it is to cancel enterprise SaaS contracts. In more business terms: switching costs are high. The average software sales process can take over six months to close and must be approved by a dozen different decision-makers. This process is arduous because contracts are lengthy. Not to mention the financial costs involved. For instance, a typical Salesforce contract cannot be canceled for free—companies must pay 100% of the remaining contract value. In other words, switching software providers as a business is a very significant undertaking. If you are going to do so, you better have a compelling reason to explain why it’s worth it. Moreover, all other executives in the company must agree with you. Painful. Truly. Painful.

Finally, in enterprise software, security issues are huge. Signing a software agreement essentially means handing over all your private data to a third party and praying they do not lose, misuse, or abuse it. In other words, it requires trust. This is the top priority for 80% of IT leaders. More importantly, trust cannot be coded overnight. Trust must be built over years or even decades. It requires long-term relationships and a track record of success. These are things traditional companies have that Anthropic does not. Trust and security are significant advantages that cannot be overlooked.

Time to Buy In

By Thursday afternoon, I had seen enough. Two voices were in my head. 1) Warren Buffett, who tells me to be greedy when others are fearful. 2) Mark Mahaney, who tells me to find "DHQ" (dislocated high-quality companies). I decided it was time to buy in and gave myself two options.

Option 1: Buy the entire software basket. I looked at IGV, an ETF containing all the large software names, which has been crushed. There may be a few losers in there, but the average multiples have dropped to a level I feel I can’t really go wrong with. This is the safe choice.

Option 2: Stock picking. That is, personally identifying a few software names that I believe are high-quality companies. This is the riskier option since I run the risk of making mistakes and choosing losers. Nevertheless, I chose option 2 because I felt bold.

DHQ (Dislocated 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, and this is not financial advice—I am just telling you what I did. My reasons are as follows.

1. Adobe

Adobe currently has a price-to-earnings ratio of 16, which is less than half of its five-year average. It is also almost half of the average S&P 500 price-to-earnings ratio. It is very cheap. The consensus is that AI will render it irrelevant, but this overlooks two key facts.

1) Adobe has been vigorously integrating AI. In fact, its AI capabilities have generated over $5 billion in annual recurring revenue, which is 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, its product is so deeply integrated across the entire creative workflow that switching solutions is difficult. It is 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 expanding their short video budgets as this medium continues to explode.

2. Salesforce

Salesforce is another company that is considered dead due to AI.

Meanwhile, its AI agent product's ARR quadrupled last quarter, and the company continues to be rated as the most trusted CRM in the industry. It has dropped over 40% in the past year, and its price-to-earnings ratio is now below the S&P average, with a cash flow price-to-earnings ratio at about half of its five-year average.

Even if Claude has more interesting products, I do not believe this will surpass the enormous switching costs—certainly not in the timeframe Salesforce may need to build its own comparable product.

3. ServiceNow

ServiceNow has been hit hard this year—down about 30% by 2026.

The consensus is that growth is about to 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 adequate.

In fact, the company is expected 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 may have noticed I didn’t mention Microsoft. That’s because I hadn’t bought in at the time of recording.

My initial viewpoint was that I didn’t need Microsoft because my exposure was already large. (MSFT constitutes 5% of the S&P 500.) However, upon reflection, I realized the valuation was too cheap to ignore.

At the time, Microsoft's price-to-earnings ratio was only 25 times, the lowest among the Mag 7. Given the reasons I emphasized above, this is relatively absurd, especially absurd for another key reason: Microsoft owns nearly one-third of OpenAI.

Even if Microsoft’s lunch is eaten (which I doubt), the company has contractual rights to get compensated. Few companies are better positioned in AI than Microsoft. The current price does not 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 do not claim to be smarter than them.

However, I also believe that from time to time, extraordinary events occur—political events, natural disasters, global pandemics, or indeed, the arrival of transformative technologies. In these cases, I believe the market can lose its rationality. When that happens, for a brief period of time, the Efficient Market Hypothesis fails.

I take the risk of making mistakes and losing money. But that’s the essence of being an investor. Besides, if you don’t occasionally take some risks, then… what fun is it?

See you next week,

Ed