Capital punishment

11h

S&P 500 futures and Nasdaq 100 futures are modestly lower this morning in the final trading day of the week.

Thursday’s intense but localized selloff was one of those days for the history books — a session typically only seen during a time of acute hedge fund pain and/or the inflation and popping of the dot-com bubble.

Days when the Nasdaq 100 finished down at least 2.5% and the S&P 500 Equal Weight Index gained 0.5% or more:

Yesterday
March 8, 2021 (when GameStop and other heavily shorted names ramped)
April 26, 2001
March 8, 2001
March 2, 2001
Oct 30, 2000
April 24, 2000
April 3, 2000
March 30, 2000
Jan 6, 2000
April 14, 1999
March 1, 1996

Retail traders on Robinhood were undaunted by the tech-centric drubbing. Net purchases of single stocks were higher on Thursday than any other session so far this week. Most heavily-traded companies that saw above-average volumes yesterday were net bought, including Meta and GE Vernova. On the other side of the spectrum, the selling parade in Adobe accelerated.

Some key technical hot spots to keep an eye on…

A Goldman Sachs index of hedge funds’ favorite stocks is now below its late July low:

The VanEck Semiconductor ETFis now a hair above its 200-day moving average and that same Leopold low, as well as on the verge of oversold territory for the first time since April 2025.

I’ve largely been of the view that market volatility has caused a lot of the narrative volatility we’ve seen in 2026 rather than the other way around. That is, stocks swing rapidly much more than fundamental realities as factors or industries move in and out of favor, and we backfill stories to explain the price action.

What’s going on in the AI trade recently looks and feels a little different this time: a trifecta of supply, capital, and political constraints have all seemingly combined to weigh on the cohort’s performance.

That being said, narrative-wise, it does also look like we’re simultaneously pricing in a lot of stories that would be hard-pressed to coexist:

A downward revision to medium-term demand for compute picks-and-shovels beneficiaries;
Less than stellar ROI on AI capex;
Hundreds of billions in hyperscaler debt/equity issuance; and

Cybersecurity spending to the moon to protect us from all the AI vulnerabilities

As a subsidiary of Robinhood, Sherwood Media is restricted from writing about any company in which Robinhood is or was a selling group member of the IPO during the regulatory "quiet period" for that company.

Capital crosscurrents

In a stunning reversal, Meta went from up more than 7% after announcing better than expected quarterly results to down 7% as investors soured on the prospective dilution that will be required to fund its AI ambitions.

Recently, concerns about the cost and availability of capital have centered squarely on marginal players in the ecosystem. We’ve seen instances of neocloud IPOs being pulled (Firmus) or flopping out of the gate (Elice, whose rebound to above its offering price has received much less attention!) while credit spreads widen to distressed levels (CoreWeave). Further testifying to this dynamic is OpenAI reportedly set to raise capital at a valuation of “only” $1.25 trillion amid lackluster demand.

These speak to two distinct though related issues: a high cost of capital and an unwillingness to provide it. If you can’t get capital, you can’t grow. If the cost of capital is too high, it can’t justify the growth you’re seeking.

Meta’s filing unpacking a new equity class raised one potentially more concerning issue: capital is getting scarce for even the stronger hands.

Per the proxy statement:

“The Special Committee also recognized that timing is important to secure capital before competitors saturate various funding options. Recent significant capital raises at other notable companies, and others soon anticipated to take place, to fund the same AI growth—for example, Alphabet, SpaceX, Anthropic, and OpenAI—have resulted in an increasingly challenging environment for raising capital. The Special Committee considered that, while the Company presently has sufficient capital resources for its current needs, in order to finance its funding needs going forward it would be prudent for the Company to do so before even more companies have absorbed available capital in the market.”

An important caveat here: this document is designed to rationalize a path the powers that be at Meta have decided is appropriate. That is, it’s going to explain why it’s a smart move for Meta to raise equity soon in a potentially unfriendly environment rather than raise equity later, or lots more debt now.

“The Special Committee also considered whether debt financing could be an alternative for the financing needs of the Company but recognized that debt financing alternatives of the magnitude that Company management was considering could have an adverse effect on the Company’s credit ratings and would be a less attractive alternative for the Company,” also per the filing.

Yes, in order for the AI boom to get the money it seemingly needs, it’s important for its obligations to be deemed investment grade. Credit-wrapping and asset-based financings are designed to make debt from riskier companies less risky. That would be undermined if the creditworthiness of the publicly traded giants that demand a lot of this compute were to come into question.

But as investors were reacting to Meta’s financials and filings on Thursday, we also had some more constructive items to consider — albeit, news that probably got overwhelmed by the sea of red in AI.

Nebius booked a $3.5 billion funding facility from Citi, and Blackstone touted the “generational opportunity” to deploy capital into the AI boom.

Here’s Stephen Schwarzmann, CEO of the alternative asset management giant, on the earnings call (emphasis added):

“Over the past few weeks, however, we've seen an increased focus in the market on the sustainability and risk of AI-related investment. The significant financing needs of companies involved in the AI build-out in the context of higher interest rates and the levels of valuations.”
“Demand for compute is accelerating and we see no signs of that slowing down as AI use cases ramp for both businesses and consumers and the world increasingly engages with the vast potential of this technology. At the same time, there's a critical scarcity of supply that is only intensified. Seldom in my career have I seen such a supply-constrained market of this scale.”
“Overall, we keep coming back to the fundamental supply demand imbalance, which supports our conviction that this is the right place to invest significant capital. As the AI revolution progresses, the rising demand for capital is creating generational opportunities for those with the platform, relationships, capabilities, and discipline to meet this demand.”

Do these sound like the words of a man who’s had all the AI he could fund?

More than FCF worries

One issue for Meta and other Mag 7 hyperscalers may be that more flies are poised to land in their operational ointment in addition to the deteriorating free cash flow outlooks.

We’ve discussed before how capex booms are an immediate boon to aggregate index earnings. One company’s spending gets booked as revenue and profits for another, but that outlay does not count as an expense until the capital enters into service and begins to be consumed (that is, depreciation).

Well, Wall Street appears to be forecasting that chickens will be coming home to roost.

“Hyperscalers margins are forecast to decline next year, even while margins across the rest of the index are expected to expand,” write Dennis DeBusschere and Sophia Wang of 22V Research. “This suggests weaker operating leverage among hyperscalers, potentially reflecting rising depreciation and operating costs as AI infrastructure enters service.”

(However, it must be noted that at many points throughout the past five years, we’ve seen calls for the complete convergence of Mag 7 earnings growth and the rest of the S&P 500, and surprisingly strong profitability among the heavyweights is a key reason that hasn’t happened.)

“Hyperscaler next twelve months earnings estimates have broadly plateaued since late August, contrasting with continued increases among other AI beneficiaries. September’s -0.6% decline – the first monthly deterioration since February 2023 – marks a break in the sustained upward trend,” they add. “Overall, the evidence points to more visible near-term investment pressures among hyperscalers, alongside resilient earnings expectations elsewhere in the AI universe and broader market, rather than a generalized AI earnings downturn.”

Cyberpunk’d

One emergent use case for AI is finding cybersecurity vulnerabilities that scare the hell out of everyone because they reportedly “affected around 70% of the internet.”

Ask Shopify CEO Tobi Lutke:

Or Anthropic’s deputy chief information security officer:

The Global X Funds Global X Cybersecurity ETF is now outperforming semiconductor stocks year to date. Kinda seems like the message from the market on AI is, “In order to deploy you must first defend,” or “the best offense is a good defense.”