Long Paper, Short Barrels
Two prospectuses arrived on Monday and the tape graded them in opposite directions. The Philadelphia semiconductor index rose 2.8% and still sits 12.2% lower for the month. Meta closed down 1.9%. Anyone who files that under noise should explain why a market that rewards one company for opening its books punishes another for opening its hand.
The rewarded document was Anthropic's S-1. It showed $16.3 billion of first-half revenue and a planned Nasdaq listing under the ticker ANT. Bloomberg put the compute bill at roughly 55% of sales. Nvidia, Micron, AMD and Broadcom each added between two and three percent. The month-to-date hole in the SOX tells you how much of Monday was relief and how little was conviction.
The punished document was Meta's. It plans up to $95 billion of new equity: $22.5 billion of Class A common through forward sales, $12.5 billion of mandatory convertible preferred at $50 a depositary share, then $60 billion of at-the-market sales in non-voting Class C stock from early next year. Meta had already warned investors it would need hundreds of billions of dollars over several years. Monday was the first installment on that sentence.
Read the two filings together and they answer the question that sank the chip complex over the last fortnight, namely who pays for all this. Shareholders pay. Meta's own AI spending pushed quarterly free cash flow below zero, so the operating cash flow that was supposed to fund the build-out has run out of road. And look at the cornerstone list. Nvidia, AMD and Broadcom expressed nonbinding interest in up to $5 billion of the Class A stock, at the same terms as everyone else. Those are the three companies Meta pays for silicon. Five billion is a rounding error against $95 billion. It is also the oldest move in the supplier's playbook: buy a piece of the customer's paper and let the order book call it demand.
The Nasdaq closed at 5,048.62 on March 10, 2000. Nobody rang a bell that day. What had changed over the preceding months was the sheer quantity of stock on offer, and 1999 and 2000 were then the heaviest years for equity issuance on record. Supply does not kill a bull market by itself. It kills it after demand has been spent, and demand has been spent more than once this autumn. Goldman's prime brokerage clients sold about $20 billion of tech over two weeks, and trend-following hedge funds had their worst week since 2024. A headline circulating Monday claimed Americans hold more of their money in stocks than at any point, the dot-com peak included. Meanwhile the ten-year pays 4.99% and the two-year 4.72%. A saver who can lock in five percent from the Treasury needs a reason to buy a mandatory convertible.
Now the calendar. Meta's $35 billion is this week's problem. Anthropic is reportedly targeting the week of November 16, with an offering that could raise $100 billion or more at a $1.5 to $2 trillion valuation. The record before SpaceX came along was Alibaba's $21.8 billion. That makes this one more than four Alibabas. The lockup schedule is spread across more than twenty windows and lets insiders trade from day one, which is a design choice built for orderly selling, so we should expect selling. If Nasdaq's fast-entry rule applies, index funds must buy after 15 trading days. SpaceX got that bid in July and fell nearly 6% on the day it arrived. Forced buyers have a way of showing up exactly when the price is least interested in them.
Meanwhile the physical economy keeps its own books. Brent sat near $115 and WTI near $101 on Tuesday, a shade lower as Saudi, Pakistani and Turkish officials talked in Tehran. A barrel cannot be issued off a shelf registration. Neither can a megawatt, a transformer, or a safe transit through Hormuz. The binding constraints of this economy are physical, and the financial response is to print more claims on future cash flow. This market is long paper and short everything paper cannot buy.
On Wednesday the Fed holds at 3.75% to 4%, which the futures put above 90%, and Warsh is expected to say that tight financial conditions are no substitute for higher policy rates. Fine. Nobody has asked what well over $100 billion of fresh equity inside a month does to those conditions. It drains the same pool of liquidity the Fed is trying to price, and no committee meets to vote on it.
Asia read Monday as an all-clear. The Kospi jumped 3.4% until a sidecar halted program trading, and the Nikkei added 1.7% on chips. None of this says sell. A lot of it says the next leg of the AI trade will be decided by absorption rather than earnings, which is a different exam with different ways to fail. An S-1 can be flawless and the price can still fall, because price is the product of the document and the number of shares offered against it. The market has a good day when someone shows it revenue. It has a bad one when someone shows it the share count.