The AI Earnings Record High Is a Concentration Mirage - and the Plumbing Is Tightening

Things started getting a bit ridiculous on Tuesday. The S&P 500 and the Dow Jones both hit record highs, and the headline says AI earnings are powering the market higher. The mainstream narrative has the beat-by-beat: Palantir raised its full-year revenue guidance to between $8.15 billion and $8.16 billion, well above the $7.65 billion to $7.66 billion it had previously forecast, and Caterpillar raised its annual revenue growth outlook citing AI data center buildout driving demand for its power-generation and construction equipment. But here is the part most commentary skips: seven of the eleven S&P 500 sectors were trading lower. Energy was down 2.5%. The S&P 500 posted only 10 new 52-week highs and two new lows. On the NYSE, advancing issues outnumbered decliners by a 1.36-to-1 ratio - that is a modest breadth reading for a market claiming it just hit an all-time high.

This is concentration, not conviction.

The SPY, which tracks the cap-weighted S&P 500, was up 4.1% over the past five days. The RSP, which tracks the equal-weight S&P 500, was up 1.2% over the same five-day window. That gap - the top 100 companies doing all the work while the rest of the index lags - is the dispersion mirage. The index looks strong because a handful of mega-caps are pulling the entire rope. When the leader stumbles, there's nothing underneath to catch it.

And here's what makes this setup treacherous: the plumbing isn't supporting it.

The Fed's overnight reverse repo facility - the parking spot for excess cash that acts as a cushion for funding markets - has been drained to near-zero levels. We were told back in mid-2025 that the reverse repo fund was at its lowest since April 2021, and since then, cash has moved from that facility into Treasury bills, leaving bank reserves as the next source to drain. The Fed's balance sheet had been cut from roughly $9 trillion down toward the $6.7 trillion range by August 2025. As of mid-2025, bank reserves stood at $3.3 trillion, and Barclays at the time estimated they would fall below $3 trillion by September 2025. When reserves fall below the "ample" level, funding markets get noisy. SOFR (the secured overnight financing rate) spikes. Money funds lose patience. And equity markets, which don't think about the plumbing until it backs up, start to move lower regardless of whether AI earnings are good or bad.

Same deficit. Same bills. Different impact - because the funding source changed. That pool of excess cash is gone.

Now add gamma to the mix. The put-to-call open interest ratio on SPY options is 2.30, meaning there are roughly two puts for every one call sitting in the options chain. That's a massive put wall underneath the market. When dealers are short that much put protection, they have to hedge by selling into weakness and buying into strength - which amplifies rallies in the short term but creates a spring-loaded downside risk once price breaks below key strike levels. The implied volatility on SPY options sits at 12.5%, and the VIX closed at 15.76. That's low. That's complacent. The market is simultaneously hedging heavily for a drop (the 2.30 put/call OI ratio) while pretending it doesn't expect one (the VIX at 16). Understanding what I understand about options positioning and spreads tells me that disconnect doesn't last.

On the name-level, NVDA's put-to-call volume ratio was 0.45 today - aggressive call buying. But that's the symptom, not the diagnosis. When one stock has that much single-name directional gamma stacked in one direction, dealers on the other side of those trades have to dynamically hedge. The mechanism: if NVDA rises, dealers selling calls have to buy the stock to stay delta-neutral, which pushes the price higher, which forces more hedging buying. It's a feedback loop that looks like conviction but is actually mechanics. And feedback loops break when volume dries up or when the gamma sign flips at expiration.

So let me be direct about where we are. The market just set a record. The S&P 500 closed at roughly 7,700. The narrative says AI earnings are validating the bull case and the question is when the bubble bursts. I think the question is narrower than that. The question is whether this concentration can hold when the plumbing keeps tightening.

Here's the conditional chain. If Palantir's earnings and Caterpillar's guidance are genuine signs that AI spending is flowing through the economy - not just circling inside hyperscaler balance sheets - then we need to see breadth return. We need the equal-weight index to start catching up to the cap-weight index. We need more sectors participating, not just tech and industrials while seven others fade. If we see that over the next few weeks, the bull case strengthens.

But if the next record is also driven by a handful of names while the rest of the index stays flat or lags, and if reserves keep falling toward that $3 trillion threshold, then the plumbing breaks before the story does. That's not a bubble bursting in the dramatic sense. It's a slow suffocation - rising funding costs, tighter liquidity, forced liquidation as leveraged positions get squeezed. The VIX at 16 doesn't price that in. Nobody is talking about it. And by the time the plumbing starts showing stress, the narrative will still be about AI earnings, which means the move will catch people flat-footed.

I'm not saying the bull case is wrong. Yes, we could still go higher. The bubble case, if you want to put a number on it, is somewhere around 25 times 12-month forward earnings for the S&P 500. But do you really want to play for another few hundred points when the liquidity cushion is disappearing, the concentration is extreme, and the options market has already positioned for a bigger move down than up?

What to watch: reserve balances next, as the drain from the RRP into T-bills continues to pressure banking system liquidity. Then the SPY versus RSP divergence - if it widens further, the mirage is getting thinner. Then the gamma regime heading into month-end and quarter-end expirations, where dealer positioning forces can snap in either direction. When any of those three signals starts flashing red, the AI earnings narrative will look like a lot of noise against a mechanical backdrop.

Views expressed here are personal and are not investment advice. Always conduct your own research and consult a qualified financial advisor before making investment decisions.