Apollo says the Mag 7 is cracking. The thread says: one month proves nothing

Apollo's pitch: the seven kings are spending their crowns away

Apollo's daily *Spark* deck, written by chief economist Torsten Slok, makes a clean argument across 32 slides: the Magnificent Seven are starting to underperform, and the cause is structural, not a blip. Valuations sit near records on a cyclically adjusted basis, while the same companies pour cash into data centers instead of returning it. The deck pivots the reader's eye from earnings multiples to free cash flow, where the AI buildout shows up as a widening drain. The macro number behind it, surfaced in the thread from a parallel Yahoo Finance piece: Big Tech capex is projected to jump about 70% this year, past $700 billion.

The planned 2026 figures one reader pulled out are the shape of the bet:

The one slide a domain expert singled out as genuinely compelling wasn't the headline chart at all.

"Page 25 'The number of data centers in the US' gives an interesting insight … 60% more data centers are being planned or are under construction … Page 27 'Cyclically adjusted P/E ratio near all-time highs' is certainly concerning and points to a near term correction." — a Google senior staff security engineer working on GenAI

The thread: a trend line through one quarter

The sharpest pushback wasn't about the conclusion but the evidence behind it. An ETF and market-making specialist dismissed the timing outright:

"Apollo should be smart enough to know that you can't draw any conclusions from 1 month of market data (especially when there was a big, relevant IPO)." — a quant whose stated focus is ETFs and market making

Others went after the framing of "Mag 7" itself. One reader noted the label is an arbitrary list coined at a moment, with no rule for swapping members the way the S&P does, which makes any long-run comparison shaky. A trader recalled the irony from the other direction:

"6 months ago when Mag7 was overperforming everyone was worried about it being too high a fraction of the S&P500." — gandalfgeek

The most cited research turned the deck's own logic against recent winners, and in its favor. One reader quoted an SSRN paper: since 1926, the median ten-year market-adjusted return of a top-quintile five-year winner is −17.8%, and recent winners have lagged 93% of the time since WWII. That's a case for diversifying out of the Mag 7, not a verdict that they're broken.

A second camp argued the deck misreads the business. A freelancer pointed out the buildout is now financed with debt and share issuance, unwinding the buyback story that anchored the bull case:

"all those companies are taking significant debt and even issuing new shares. One major part of the investment thesis in these companies were their constant stock buy backs. Now their gargantuan Capex … is back at diluting investors." — a product engineer and consultant

And a structural read framed the whole boom as a treadmill: if a good-enough model can be swapped in every few months, value flows from the model to cheap inference, which is great for users and brutal for returns on huge capex. tptacek compressed it to four words.

"This is like a straightforward Red Queens Race story, isn't it?" — tptacek

The deck's strongest critics and its defenders converge on the same uncomfortable spot: the buildout might be both real and unrewarding for the people funding it. Meta could halt its data-center spend tomorrow and post the same cash flow it always did, as one skeptic noted, which is exactly why one quarter of relative weakness settles nothing.