#美股AI概念股普跌 The AI industry is still growing, but the market is starting to discount expectations of “unlimited investment.”
On November 5, the Nasdaq fell 5.52%, and the Philadelphia Semiconductor Index fell 12.25%. Chip stocks fell the hardest, while cloud providers like Microsoft and Amazon saw much smaller declines. This shows that investors are reassessing companies, not that AI demand has suddenly disappeared.
Why did the market fall this time?
Policy expectations shifted: After the midterm elections, data center approvals and AI regulation may tighten.
Returns are being questioned: Free cash flow at Google, Meta, and others has turned negative, putting the problems of heavy investment and slow returns in the spotlight.
Interest rates are weighing on valuations: Rising Treasury yields are hitting high-valuation tech stocks first.
Crowded trades are unwinding: Leveraged ETFs and quantitative funds cut positions in unison, amplifying the declines.
What signs suggest the trade is really cooling?
Earnings expectations have not been significantly revised down; stock price declines are mainly driven by P/E multiple contraction, indicating that valuations are being repriced.
Companies with higher capital expenditures have fallen harder, as the market starts scrutinizing returns on investment.
Pure-play concept stocks and highly leveraged computing-power rental companies face more pronounced shakeout pressure.
Bernstein believes the long-term thesis remains intact, but a full recovery will take time.
Looking ahead, markets will remain volatile in the short term. For the market to truly enter a cooling phase, watch whether data center construction actually slows, cloud providers cut spending, and AI revenue continues to materialize. For now, this looks more like a shift from “telling stories” to “checking results,” not a bubble bursting like in 2000. #WCTCS9

🚀 Gate Square Content Mining Upgrade | Post to Earn Up to 60% Commission Rebates, Win Rewards for Quality Content!
🎁 Triple Benefits, Bigger Creator Rewards!
1️⃣ Base Rebates: Up to 30%! The more actively you post, the more engagement you get, and the higher your ranking, the higher your rebate rate!
2️⃣ Newcomer Perk: Newly onboarded and returning creators get double rebates for 30 days, up to 60%!
3️⃣ Upgraded Content Rewards: Keep publishing quality content and driving genuine engagement to earn extra content rewards!
🔥 Drive Trading with Posts, Earn More with Quality Content!
Post Now 👉 https://www.gate.com/post
Event Rules 👉 https://www.gate.com/announcements/article/102544
‍#内容挖矿

#OpenAI第三季度营收升至100亿美元 Bullish on AI, but retail investors can’t buy “model companies”
I. First, read OpenAI’s report card in full
Q3 revenue was $10 billion, up nearly 50% quarter over quarter, with an annualized run rate now at about $50 billion; the previous quarter was $6.7 billion.
But there’s another side to the same report card: estimated losses of about $14 billion in 2026, and cumulative investment projected to reach about $115 billion by 2029. Altman himself has said that continuing to grow at this rate will require “trillions” in infrastructure investment.
The valuation tells an even clearer story: in March, it raised $122 billion at an $852 billion post-money valuation, setting a record for a single funding round by a global startup; the target valuation for the current round is $1.25 trillion, but only $11 billion has come in during the first tranche, with both the size and valuation expectations shrinking from the previous round.
So the company’s profile is this: revenue is growing exponentially, losses are expanding in proportion, and the faster it grows, the more it depends on capital markets to keep funding it. It is the world’s most expensive “work in progress.”
II. The key question: Can you buy it?
OpenAI is not publicly listed. The March round was the first to open to individual investors through bank channels, raising more than $3 billion from individuals, while also being included in several ARK ETFs.
In other words, ordinary investors have only two ways to get exposure to it: qualified-investor shares through bank channels, or buying ETFs that hold it or shares of its corporate shareholders.
So “model companies or the supply chain” is really a false choice for most people—you can only pick from the supply chain, or use the shell of a public company to buy exposure to a model company.
III. What you should really bet on is where profits migrate along the chain
The reality today is that hardware takes most of the profits. The AI infrastructure sector accounts for half of S&P 500 earnings growth, while Nvidia is still growing revenue at 100%; profits decline at each step along the “chips—computing infrastructure—large models—applications” chain.
But institutions’ medium-term outlook points in the opposite direction: hardware’s share of profits will shrink, while revenue for models and applications could grow 26.9-fold from 2026 to 2030 (versus 12.5-fold for cloud services), driving nearly two-thirds of incremental earnings.
These two statements don’t contradict each other; they describe different points in time.
Selling shovels is “the best business now”; models are “the best business of the future.” They aren’t opposing camps, but upstream and downstream parts of the same chain—the real bet is on which end profits will migrate to.
IV. My allocation strategy: 70% supply chain, 30% model exposure
Within the supply chain, I only buy companies with contract-backed demand and reasonable valuations:
Nvidia at $168.89 and 21x PE has the longest order visibility, but its November 17 earnings report and November 18 congressional hearing are two simultaneous events, so it’s not suitable for a full-position bet.
TSMC at $396.48 and 25.7x PE has the most diversified customer base, so it doesn’t have to bear the risk of relying on a single customer.
Oracle is a “long-dated payoff” play—its $300 billion in contracts will take effect in stages under 5-to-6-year capacity agreements, with most of it not coming due until 2028. Buying it means buying patience.
AMD has 6 gigawatts of computing capacity agreements, but its 110x PE already prices in the full expectations, so it’s only for swing trades.
I don’t buy model exposure directly; I use the shells of public companies:
Microsoft at $533.16 and 27.8x PE is one of OpenAI’s largest shareholders, while also being a platform in its own right.
Amazon at $270.87 and 21.7x PE has invested $50 billion in OpenAI and also holds an equity stake in Anthropic.
Google at $287.38 and 15x PE has a stake in Anthropic as well as its own models, and is also the cheapest of the Magnificent Seven.
What these three have in common is that their cash flows can cover AI investment, so they don’t need to rely on fundraising to stay alive. Buying them is like buying a “public company shell + model company option”—the only viable path for retail investors.
V. When will this answer change?
Anthropic’s IPO. Its Q3 revenue was 70% higher than OpenAI’s, with an IPO valuation cap of about $1.41 trillion; Amazon’s $18 billion investment could swell to $270 billion, while Google’s $33.3 billion investment could reach $139 billion.
Once it goes public, model companies will have a public-market valuation benchmark for the first time. At that point, “models or the supply chain” will finally become a choice you can actually make, rather than one you can only watch.
The easiest mistake to make when bullish on AI is to equate “being optimistic about the industry” with “buying the purest play.” The purest plays are often unavailable, and the ones you can buy are often not the purest.
As profits migrate along the chain, upside potential matters more than purity.
Would you buy private shares for “purity,” or the supply chain for “upside potential”? Let’s discuss in the comments#Gate股票畅聊季