Slapped in both directions! "Insufficient electricity" vs "excess computing power", the outcome is "selling AI"

The U.S. stock market's mega-cap cloud providers have just delivered their sixth consecutive quarter of accelerating growth, while Intel has endured its worst single-day plunge in half a century (-29%). Caught in the middle, hedge funds have spent the entire month of October offloading anything with “AI” in its ticker.

What exactly is the market afraid of? Rich Privorotsky, Goldman Sachs’ EMEA Delta One trading chief, wrote in his Monday morning report that, nobody truly knows what to be worried about. He wrote:

The grand debate over AI is becoming somewhat circular.Half the market thinks we simply can’t build enough electricity and data centers to support 2027 capital expenditure plans, so the growth of “picks and shovels” stocks is bound to peak, with multiples compressed as a result.

The other half of the market believes we are about to massively overproduce compute power, which would crush lease economics and likewise demand lower multiples.

In my view, the bigger issue is that no one truly knows what to worry about. Demand is undeniably strong, but the returns on the next trillion dollars invested are far from certain.

Insufficient Compute or Oversupply?

Privorotsky’s observations highlight the factual contradiction behind two bear market logics.

If electricity is the bottleneck, then compute will remain chronically scarce, lease prices will stay elevated, but the problem will be the inability to connect the next GPU to sufficient power, flattening out the “picks and shovels” stocks’ growth curve. Morgan Stanley analyst Stephen Byrd projects the US will still face a net electricity shortfall of 38GW before 2028.

If the industry is about to overproduce compute at scale, then lease economics will collapse and returns on current chip inventory will crumble. Privorotsky warned of this risk as early as June, noting that the market’s core thesis was built around compute scarcity.

These two scenarios can’t serve as the core constraint simultaneously, but both can justify selling. Privorotsky wrote: “When positions are crowded, each new concern provides yet another reason to reduce risk exposure.” The market doesn’t need to settle on a single bear logic—only cling to whichever explanation is most popular that day.

Demand Is Real, Bills Are Real

Last week’s numbers from mega-cap cloud providers were outstanding: Azure’s quarterly growth was 48%, AWS rose 39%, and Google Cloud rocketed 103%. Privorotsky called it “extraordinary demand.” Amazon, Alphabet, and Microsoft combined cloud revenue growth hit 54% year-over-year in Q3, marking the sixth straight quarter of acceleration.

But then comes the “that said.” Privorotsky acknowledges, “Negative free cash flow, equity financing, and steadily growing leverage demands have sparked reasonable doubts about capital returns.”

Goldman Sachs colleague Lee Coppersmith provided specifics: mega-scale cloud providers’ capital expenditures are projected to reach $1.2 trillion by 2027, while their combined free cash flow in Q3 has already turned negative, burning through about $31 billion in a single quarter.

The simple arithmetic: $1.2 trillion in capital spend combined with negative free cash flow means the “next trillion” Privorotsky worries about will have to be financed largely through the bond market, not operating cash. The bond market itself is not exactly relaxing either. AI-related off-balance sheet debt has ballooned to $4.1 trillion, and funding demand is becoming increasingly crowded.

Intel Plunges 29%

Last Friday was a textbook example.

Earlier in the week, the market was basking in a string of relatively upbeat news, but Intel then issued a “surprisingly severe correction in PC demand,” plunging 29% in a single session.

Privorotsky believes Intel’s warning was “more about consumer PC demand, margins, and manufacturing operations than AI demand itself.”

But the market clearly lacked patience.Investors still traded it as yet another victim of the broader AI narrative.

Meanwhile, Amazon—called a “crowded holding” by Privorotsky—saw its share price recover from -7% pre-market to close down just 1.4%. Even as market concerns about its free cash flow lingered, the stock showed resilience. The market began selectively pricing. However, Kioxia’s 15% plunge on Monday was a reminder to investors that this is not a broad-based indiscriminate rally.

Hedge Funds Undergo Largest Deleveraging Since 2015

Is the AI fundamental deteriorating or are crowded trades being forced to unwind? Privorotsky’s positioning data point to the latter.

Goldman Sachs' tracked systematic long/short strategies posted a 2.17% weekly drawdown, “the worst one-week performance since December 2023.” Fundamental long/short net exposure only recovered to the 7th historical percentile, “mainly via index short covering.” Single stocks have seen net selling for seven consecutive weeks, while macro products were net bought for two straight weeks. The American Association of Individual Investors (AAII) bull-bear sentiment index dropped back into negative territory, and the CNN Fear & Greed Index fell to the low 20s.

Morgan Stanley also found that hedge fund AI-related holdings are underweight for the first time since early 2023. The market just saw one of the largest stretches of sustained deleveraging in over 15 years.

Privorotsky’s conclusion: “Sentiment and positioning have been reset.”

But Goldman colleague Coppersmith is cautious. He notes that October saw the biggest single-month decline in global risk appetite since November 2021, but “we are still not at a risk appetite index level that would trigger a contrarian buy signal.” More importantly, despite net exposures being low, total leverage remains at 314% (97th percentile over the past five years). In other words, even as managers cut positions, overall leverage is still high and there is still latent energy for further market volatility.

“Nearly Unmanageable,” But a Year-End Rally May Be Looming

Privorotsky had an unusually candid passage in his report:

It feels nearly unmanageable. Almost every client I’ve spoken with is struggling. Last week felt like a full-on deleverage…momentum unwind, the most shorted stocks getting squeezed, leverage coming out of the system. Defensive stocks and high quality assets soared, AI multiples compressed. The market is pricing in a sudden growth slowdown, AI capex stalling, nominal growth and rates heading lower, and premium for high-quality balance sheets.

He then points out, “Last week’s mega-cap cloud earnings showed no sign of weakening spending…if anything, the opposite.” Whether this supports the bulls or presents a new risk depends on one’s bear camp allegiance.

After passing through such key events as the midterms, the U.S. Treasury’s Quarterly Refunding Announcement (QRA), and non-farm payrolls, Privorotsky offered his outlook:

My instinct tells me we’re preparing for a painful year-end rally…underexposed, low participation, hard to chase higher. Last Friday definitely felt like a capitulation selloff, but I think we need to get past these events before risk gets put back on.

UBS likewise warned of year-end short squeeze risk with short interest near historical highs. Goldman options strategist Brian Garrett also sees seven signs of a market “healing”—with the caveat “we just need to get through 2026 first.”

The elements for a positioning squeeze are in place: net exposure at a historical 7th percentile, sentiment in the gutter, seasonal tailwinds approaching. But as Privorotsky himself asks: demand is real, while the next trillion’s returns are uncertain. With mega-cap cloud providers’ free cash flow having turned negative, total leverage at the 97th historical percentile, and the 2027 capex bill about to hit the bond markets in a big way, any year-end rally could simply present a selling window.