Goldman's Moravi: Two Key Signals of Peak CapEx Among Tech Giants Have Emerged
nashnova research
Goldman Sachs trader Bobby Molavi says two of three signals tracking peak tech-giant capital spending have triggered, putting the AI infrastructure narrative at a critical juncture — while professional investors are struggling to make money even as indices hit all-time highs.
Two signals out of three — what are they saying?
Molavi tracks three signals for the peak of the tech-CapEx cycle (capital expenditure — the money Big Tech pours into AI infrastructure). Two have now fired simultaneously.
He calls the pattern the "gap between peak stock price and peak CapEx." This means → share prices may have already front-run the spending high, even as actual outlays keep climbing.
He deliberately withheld the third signal's identity, but two firing at once is enough to open a serious debate: is this AI spending cycle approaching its inflection point?
Indices at all-time highs — so why can't fund managers make money?
The S&P 500 sits at a record, yet U.S. long-short net exposure has dropped to roughly 46% — close to the low hit during the April 2025 "Liberation Day" sell-off.
In plain terms = managers are still in the game, but their bets are so hedged down that they're effectively half-sitting-out.
The spread between single-stock realized volatility and index realized volatility has widened to roughly 26–28 percentage points, well above the ~20-point peak during the 2009 financial crisis. This means → individual stocks are swinging wildly in offsetting directions, making the index look calm while the lived experience of stock-pickers is brutal.
How extreme is market concentration?
The S&P 500's top five companies carry a combined market cap near $22 trillion — roughly 30% of the index, equal to the other 434 constituents combined.
For perspective: in 2013–2014, it took about 200 stocks to match the top five's value; at the 2000 dot-com peak, the figure was roughly 350. In plain terms = today's concentration exceeds the dot-com bubble's.
The "AI Top 10" account for about 40% of S&P 500 market cap, up roughly 58% year-to-date. The median S&P 500 stock is down about 15%. This reflects a market that is not broadly rising — a handful of AI winners are dragging the entire index higher.
Volume is surging — so why is liquidity shrinking?
Since 2023, top-of-book liquidity — the dollar amount the best bid/ask can absorb — has fallen from roughly $14 million to about $11 million, a 21% drop.
Book depth per billion dollars of volume has plunged from $162,000 in 2017 to roughly $11,000 — a 93% collapse — while average daily volume rose from about 6.5 billion shares to nearly 19 billion.
In plain terms = trading activity has exploded, but the capital standing ready to absorb large orders has shriveled. When June 2026 volume hit 22.6 billion shares — the year's high — book depth fell another 33%. This means → if directional selling pressure hits, the risk of a flash crash is far larger than headline volume figures suggest.
Is "forward P/E says it's not expensive" actually reliable?
Molavi directly challenged the prevailing narrative that the S&P 500 is reasonably valued on a forward P/E basis.
His exhibit: in March 2000, the forward P/E was 25.2× — seemingly acceptable. But trailing actual earnings put it at roughly 39.3×, and on a GAAP basis it was about 60×.
This reflects the structural optimism baked into analyst forecasts. In plain terms = arguing "it's cheap" on forward P/E is like arguing "I'm thin" based on target weight — when the actuals come in, the picture can look entirely different.
When does the third signal fire?
Molavi intentionally kept the third peak-CapEx signal unnamed, but stated clearly: whether it fires will be the key test of whether this AI capital-spending narrative can continue.
This means → the market is in a "two lights on, waiting for the third" observation window. If the third signal also triggers, the case for an AI CapEx peak gains full confirmation.
For investors, this is not a sell-everything signal — but it is the moment to re-examine how concentrated your AI exposure really is.
市场有风险，内容仅供研究参考，不构成投资建议。