Author: Wang Chuan from Silicon Valley, investguru
1/ Much of human anxiety is a stress response triggered by the amygdala in the brain. To alleviate anxiety, simply offering platitudes like “Don’t be anxious—it won’t help” is ineffective, as it does not address the underlying mechanisms generating the stress response. The truly effective method is to encourage anxious individuals to slowly write down the events and reasons causing their anxiety on paper. The act of writing transfers control from the amygdala—responsible for heightened alertness and tension—to the prefrontal cortex, which governs rational thought. The more detailed and deeply analyzed the writing, the more thoroughly control is transferred, and anxiety naturally decreases. Even if you cannot finish in one sitting, simply beginning the process and having a general sense of how long it will take to clarify your thoughts is enough to start reducing anxiety. Without consciously putting thoughts down on paper and relying only on motivational clichés, this transfer of control cannot be completed, and anxiety cannot be fundamentally resolved.
In May 2026, one concerning development was the widespread surge in semiconductor industry stocks. The renowned Philadelphia Semiconductor Index rose 150% compared to a year earlier. Stocks of various semiconductor memory companies surged across the board; as of May 8, the closing price of SNDK reached $1,562—38 times its value a year ago. Although I do not hold investments in semiconductor companies, I aim in this article to dissect the underlying logic behind this rally and share insights with readers.
3/ This story begins with the rapid rise in valuations of various AI startups. Many outside observers, focused solely on stock price fluctuations, lack basic常识:
- First, the vast majority of AI startups are operating at significant losses, yet many people completely confuse the basic concepts of revenue and profit (some even intentionally so), and continue to repeat these misconceptions despite being corrected.
- Second, when a company says its ARR (Annual Recurring Revenue) has reached $100 million, it does not mean its total revenue over the past year was $100 million—it means at most that its revenue last month was approximately $8.33 million. If you misunderstand or overinterpret it, that’s your own issue.
- Three, what a company doesn’t say reveals more than what it does. It reported ARR but didn’t mention profitability, implying severe losses. It stated six months ago that its ARR reached $100 million, but has been silent since—likely indicating declining revenue and even greater losses.
4/ Taking the currently highly prominent large model company Anthropic (hereinafter referred to as Anth) as an example, some media reports at the end of April indicated that the company’s ARR had approached $40 billion. In other words, the company’s monthly revenue in April was close to $3.3 billion ($40 billion divided by 12). In March, Anth’s ARR was still below $30 billion, implying that its monthly revenue in March was less than $2.5 billion. In a public filing in March, Anth’s CFO disclosed that the company’s cumulative revenue from its founding in 2021 through March 2026 would exceed $5 billion. In other words, Anth’s cumulative revenue since its founding over the past five years (once again, this refers to revenue, not profit) is approximately $10.8 billion.
5/ There’s also an issue with how revenue is defined. A large portion of Anth’s revenue comes from Google Cloud and AWS. Cloud providers take at least a 20% cut of what customers pay, and the remainder goes to Anth. Therefore, Anth’s publicly reported ARR should be discounted by another 20%. But we’ll overlook this for now, since Anth is still losing a significant amount of money each month.
6/ How much is the loss? Asked several different AIs, and based on limited public information, assuming monthly revenue reaches $3.3 billion, estimated monthly losses range from $1.1 billion to $1.7 billion. This is precisely why Anth continues to seek external funding. As of the end of April, Anth has raised a total of $72.3 billion over five years, but its cumulative revenue is only $10.8 billion—and it still needs to raise more. People who only read headlines mistakenly believe Anth has already become immensely profitable, without understanding the immense pressure on management to continuously secure new external funding just to stay afloat.
Anth is not without competition; its CEO recently acknowledged in public remarks that other U.S. companies’ AI models are only one to three months behind Anth, while Chinese companies’ AI models are six to twelve months behind. Product features across companies are rapidly evolving and shifting, making it difficult to determine the ultimate winner based on performance over one or two years. For example, Cursor was the most popular programming assistant among AI developers from 2023 to early 2025, and the company was once highly sought after. However, Anth’s Claude Code, launched in the first half of 2025, can automatically generate large volumes of code, enabling many non-programmers to participate in coding—this has been the core driver behind Anth’s rapid rise in revenue and valuation over the past year. Recently, some developers have complained that Claude Code’s user experience and value proposition are inferior to OpenAI’s Codex in certain areas, leading them to switch back. For many users, the switching cost is low. In addition to OpenAI, SpaceX’s Xai began collaborating closely with Cursor last month to develop a competitor to Claude Code.
Investors bullish on Anth often cite scenarios where Claude Code achieves profit margins as high as 70% during inference tasks. However, this claim ignores Anth’s enormous ongoing costs for continuously training new models, and assumes no competition exists—thus no downward pricing pressure. This clearly does not reflect reality. When Anth raised funds in February, its valuation was $380 billion. Just three months later, it began seeking new funding with a valuation target above $900 billion. Recently, in a private equity secondary market transaction, its valuation was reportedly as high as $1.2 trillion. (Some podcasts have even suggested a long-term valuation of $5 trillion.) This is a company with only five years of history, cumulative revenue (not profit!) of $10.5 billion, yet still losing hundreds of millions of dollars each month. Simply because it experienced rapid growth over the past year, some believe it can sustain this indefinitely. Raising $7.2 billion while operating at a loss for five years to generate $10.5 billion in revenue—and then being valued at $1 trillion on capital markets—is an outcome many entrepreneurs naturally aspire to! 😊
Berkshire Hathaway (hereinafter referred to as BRK), under Warren Buffett, generated $370 billion in revenue in 2025, $45.9 billion in operating cash flow, $66.9 billion in after-tax profits, and holds nearly $400 billion in cash on its balance sheet, yet its market capitalization is only around $1 trillion. Put another way, BRK’s ten-day revenue equals Anth’s entire cumulative revenue to date—yet some investors are willing to believe that Anth, which still heavily relies on external funding to survive, is valued higher than BRK. The immense courage required to hold such a belief is truly astonishing.
The valuation growth of 10/Anth and OpenAI is crucial for the surge in capital expenditure (capex) across the entire AI industry, and subsequently, the boom in the storage sector. Looking back at the developments over the past few years, this logical chain is clear. To find out what happens next, stay tuned for the next episode.