#美股AI概念股普跌ALAB跌超9% The anchor of this AI bull market—can it still hold?

Investors holding U.S. tech stocks and the AI hardware supply chain endured another sleepless night. The market was brutal: Intel plunged about 5%, Applied Optoelectronics tumbled 13%, Lumentum fell 5.6%, Micron and SanDisk came close to 5%, and Coherent crashed more than 9%; the Philadelphia Semiconductor Index plunged 3.39% in a single day; Nvidia closed down nearly 3%, wiping out more than RMB 1 trillion in market value.

The blame was once again thrown at OpenAI.
OpenAI is a fake negative catalyst. OpenAI disclosed that its annualized revenue as of the end of September was around $50 billion, while the market had previously circulated figures of $68 billion to $70 billion—a gap of nearly $20 billion.
But this is not a blowup. The definitions are different: Anthropic counts partner revenue from AWS and Google Cloud as revenue, while OpenAI does not. The $50 billion is its own figure.
Its growth is not weak either: overall revenue grew 77% in the third quarter, and enterprise revenue grew 107%. A company still growing at 77% cannot be called a blowup.
This is merely an excuse for AI stocks to fall.
The key is still Nvidia, because Nvidia is the overarching anchor of this market rally.
In both the primary and secondary markets, everything ultimately comes down to pricing computing power, and Nvidia is the outlet for that computing power.
Its quarterly revenue of $96.2 billion is the sum of several hundred billion dollars in downstream capital expenditures; its $108 billion guidance is a forecast of computing investment for the entire industry.
It is the load-bearing wall: if it holds, everyone can keep playing their own game; if it collapses, no one gets away.

The previous anchor of this kind was Cisco in 2000: it also sold infrastructure, also reached the top in market capitalization, and when the anchor collapsed, its stock fell 88%. But there are two types of collapse.
Valuation collapse: the stock price falls while earnings continue to grow—a mid-cycle shakeout, which is what is happening now.
Fundamental collapse: capital expenditures peak, orders miss expectations, and earnings are revised downward. That is what ends a bull market.

The anchor may also be inflated.
Nvidia invests in OpenAI, Anthropic, and xAI; the money makes a round trip and comes back to buy its chips, and Nvidia has also provided OpenAI with more than $100 billion in lease guarantees. Of the $96.2 billion in revenue, how much comes from genuine end demand and how much is the result of a circular flow must be distinguished.
Jensen Huang specifically came out in August to say that the risk was very low.

To determine whether the bull market is still intact, watch three signals: whether cloud providers' capital expenditure growth has peaked, whether Nvidia's order guidance misses expectations, and whether circular financing and in-house chips are beginning to cannibalize revenue.
Nvidia has no problem for now; the problem lies ahead.
Overnight, Nvidia fell 2.94%, the Nasdaq fell 1.5%, and Micron, Oracle, and CoreWeave followed lower.
Nvidia's latest quarter: revenue of $96.2 billion, up 106% year over year; data center revenue of $89 billion, up 117% year over year; net income of $59.7 billion; gross margin of about 75%; and next-quarter guidance of $108 billion, above expectations. There is nothing wrong with its current results.
What the market is selling is not this quarter, but how many more quarters double-digit growth can be sustained.

Three concerns:
First, customer concentration is too high.
Three customers account for 44% of revenue, while the five largest cloud providers account for half. Cloud providers' capital expenditures are expected to reach $710 billion in 2026, up 61%, but growth has already slowed from above 70%. Once capital expenditure turns downward, Nvidia will be the first to cool.
Second, upfront commitments are too large.
The $270 billion in supply commitments lock in HBM, TSMC capacity, and power in advance. When demand is there, this is a moat; when demand turns downward, it becomes inventory and impairment. Third, gross margin has peaked.
Gross margin will fall from 75% to 74% next quarter and could reach 71% to 72% in the fourth quarter. Major customers are still developing their own chips—Google TPU, Amazon Trainium, and Microsoft MTIA—which will divert demand over the long term.

Right now, valuation is wobbling while fundamentals are still holding. Nvidia's gross margin, cash flow, and orders are real, unlike Cisco's purely conceptual story back then. Most likely, valuation will collapse first rather than fundamentals collapsing outright.

The steepest acceleration phase of the AI narrative has passed. The sector is shifting from “rise on storytelling” to “prove it with results.” What should be done?
In the short term, the shakeout is not over, so do not rush to buy the dip.
In the medium term, treating differences in reporting standards and a shift in growth rates as a fundamental collapse and selling into the resulting trough could present an opportunity.
The simplest framework is to watch Nvidia, the anchor. If the anchor wobbles at the valuation level, it is an opportunity; if it collapses at the fundamental level, it is an exit signal.
For now, the anchor is still holding. $NVDA