Source | WhiteLine
Compiled | Wu Shuo Blockchain
The content of this article does not constitute any investment or financial advice. Readers should strictly comply with the laws and regulations in their respective jurisdictions. Please star the Wu Shuo public account; otherwise you may not receive the notifications.
Find direction, before change arrives
(WhiteLine) produced by the Wu Shuo team, taking its way from Crypto to the broader capital markets, focusing on the trend shifts in the AI era.
In this episode, we focus on the most critical main thread in the current AI market: Anthropic’s rapid revenue growth is changing the market’s logic for pricing AI infrastructure investment (AI Capex).
The host, Minta, believes that the reason AI infrastructure hasn’t yet collapsed despite “bubble” skepticism isn’t simply because chips are strong, AI iterations are fast, or models are improving. The key is that Anthropic has proven that large-model commercialization is rapidly materializing. Using data such as revenue, valuation, and enterprise customer payments, the article explains that Wall Street is shifting from doubting “whether AI can make money” to a new framework: “whether compute limits the ceiling of revenue.” It also further breaks down three layers of trading paths around Anthropic, including Pre-IPO allocations in the primary market, shadow targets in the secondary market, and the entire AI Capex supply chain.
Minta believes SK Telecom is a relatively purer Anthropic shadow asset, but he also flags risks such as the ownership ratio, mapping discount, and volatility in core operations. The overall conclusion is: Anthropic validates demand, AI CapEx drives capacity expansion—together forming the core logic of this AI rally.
Below is a text summary of this episode’s video:
I. The core logic of AI CapEx
AI CapEx refers to the infrastructure costs that large companies invest in ahead of time to develop AI, including GPU, data centers, power, liquid cooling, and networks. These investments are difficult to recoup in the short term; fundamentally, they’re to prepare for future capacity.
The reason the AI bubble hasn’t burst for so long isn’t only that semiconductors are strong and AI iterations are fast. More importantly, demand has started to be validated. White Line believes the key company that makes this logic work is Anthropic. It’s Anthropic’s explosive revenue growth that has shifted AI CapEx from a pure “burning cash” narrative to a capacity story that can be realized.
II. Why Anthropic is the AI mainline
Anthropic is the main AI theme of this cycle because it was the first to prove that AI can be genuinely commercialized—not just a money-burning narrative.
The core comes down to three points:
1. Revenue explosion. Annualized revenue grew from about $1 billion in early 2025 to over $47 billion in May 2026—about a 47x increase in one year—showing that it has penetrated enterprise workflows.
2. Repricing of valuation. Valuation rose from $183 billion to $965 billion, reflecting the market’s systemic revaluation.
3. Enterprises pay to make it real. Claude Code revenue is growing rapidly, and the number of enterprise customers paying over $1 million per year has surged too—indicating that growth is driven mainly by enterprise deep-paid usage, not by C-end trial.
So, the significance of Anthropic isn’t that “growth is fast,” but that it has validated AI’s commercial closed loop: enterprises are willing to keep paying, and AI CapEx therefore has a repayment logic.
III. Gavin Baker’s judgment framework
Is Anthropic expensive or not doesn’t depend on static valuation—it depends on where the revenue bottleneck is: is it insufficient demand, or insufficient compute?
If the bottleneck is demand, that means the growth potential is close to its limit; if the bottleneck is compute, then current revenue isn’t the ceiling. According to Gavin Baker’s view, once compute is loosened, Anthropic’s revenue could rise from $50 billion further to $100 billion, or even $200 billion.
Viewed this way, when the valuation is close to $1 trillion and revenue is $50 billion, the implied P/S multiple is about 20x; but if revenue rises to $100–200 billion, the valuation multiple would drop to 5x–10x.
So the core isn’t “is it expensive right now?” but rather this: For every additional unit of compute, how much revenue can Anthropic unlock.
IV. Since Anthropic isn’t listed, how to trade around the main theme
The trading thesis around Anthropic can be broken into three layers: buy Pre-IPO shares, buy shadow stocks, and buy the AI CapEx supply-chain.
1. Pre-IPO
The most direct approach is to buy Anthropic’s Pre-IPO allocation—maximum optionality and closest to the company itself.
But the core risk lies in whether equity transfers will receive official approval. If the platform and shares aren’t compliant, they may not truly be able to capture the subsequent upswing in valuation.
2. Shadow stocks
The standard for shadow stocks isn’t high correlation—it’s actually holding Anthropic shares, and the value of those holdings is large enough to affect your own stock price.
Amazon, Google, Fidelity, Blackstone, Sequoia, as well as Micron, Samsung, and Hynix—all have some connection to Anthropic. But their stock prices are driven more by their own businesses, so the trading logic isn’t pure enough.
In contrast, SK Telecom (SKM) is closer to a pure shadow stock: first, it’s already listed and can be traded directly; second, it holds Anthropic shares; third, its own market cap is smaller and its core business is relatively stable, making Anthropic’s valuation changes more easily transmitted into its stock price.
Based on publicly available estimates, SKM’s stake is about 0.54% to 0.58%. Using Anthropic’s $96.5 billion valuation, that implies an equity value of roughly $5.2 billion, while SKM’s market cap is about $14 billion—so the stake is more than 30%. Therefore, its logic can be summarized as “telecom cash flows + an Anthropic valuation option.”
The risks are clear as well: the equity stake proportion hasn’t been officially confirmed, and future fundraising and options could dilute it. Fundamentally, SKM is still a telecom stock, and shadow stocks can’t map Anthropic’s valuation changes 1:1.
3. The AI CapEx industrial chain
Expand outward from Anthropic into the AI CapEx industrial chain. As long as Anthropic continues to validate demand, the market will keep betting on the expansion of AI infrastructure. The trade can spill over into cloud compute, chips, advanced process nodes, as well as liquid cooling, optical-electrical components, power, interconnect, and other related areas.
In other words, Anthropic validates demand, while the AI CapEx trade is about expansion of capacity.