1Data-centre spending is flattering the profits that make stocks look cheap
Ben Inker of GMO, speaking on Excess Returns, made the case that a bubble does not need an expensive multiple — it can sit in the earnings rather than the price. The mechanism is mechanical rather than behavioural. The AI build-out, of which he cites roughly $200bn of data-centre spending at Microsoft , is one company's capital outlay and another company's revenue, and that revenue lands in reported profit today while depreciation on the assets it bought has barely started. Profits are therefore, in his framing, unsustainably good, and a market priced off those profits screens cheaper than it is.
Semiconductors are where he says the setup is live rather than theoretical. Demand is extraordinary and earnings are enormous; if the industry is still fundamentally cyclical, then today's earnings are a peak being read as a run rate, and any valuation built on them understates the risk. The episode reaches for David Einhorn's "successively higher lows" description of the memory cycle, cited via Micron , as the standing counter-case — the argument that each memory trough is shallower than the last, which is what would make this something other than an ordinary cycle. Both cannot be true, and nothing in the day's material settles which one is.
Inker's second distinction is what gives the first one teeth: an easy bubble versus a hard one. In 2000 you could step around the froth and still hold an ordinary portfolio, because small caps, REITs and emerging-market equity and debt all did the job. In 2008 the things that worked were gold, tail hedges and cash, all of which look ridiculous to hold in normal times. An earnings bubble sits closer to the second kind, because the usual defence — rotating into whatever screens cheap — runs through the same distortion that created the problem.
Chris Mayer, on the same episode, supplies the reason a screen will not rescue you: the label is doing the thinking. "AI" means something entirely different at SpaceX, at Google and at IBM, and his complaint is that firms bolt on AI features that change nothing about profits purely to earn the tag. , , and came up in the same register, as illustrations of how a label or a valuation gets applied rather than as calls on any of them. The consequence is narrow and specific: a low multiple is not evidence of safety when the earnings underneath it are the thing being inflated. Worth holding the register in mind, though — this was a conceptual episode, so the argument is a test to apply, not a claim that the test has already been failed.
2The doubt about data-centre profits stayed inside the data-centre names
The same scepticism reached the wires independently of the podcast. CNBC ran "Everything tied to the data center is suddenly suspect — can Big Tech fix it?", and Reuters headlines in the day's feed had AI doubts clouding the outlook. In the tape it was narrow. Nvidia was down 1.64% with data-centre doubts attached to the move, and the steepest falls among the names on Yahoo's movers list were ON Semi at −23.66% and Bloom Energy at −18.49%, neither carrying a stated catalyst in the feed.
The names around them went the other way. Micron joined the $1trn club, Qualcomm rose on a ByteDance deal, Oracle was upgraded to buy, and Microsoft — the company whose $200bn of spending is the exhibit in the argument above — was up 5.71% on the same list. Across the pre-market movers feeds the recurring names were memory and chips: again, alongside , , , and . So the doubt is being expressed as rotation inside the complex, not as a de-risking of it.
Risk appetite ran further out than that. The SpaceX and Quantinuum listings had space and quantum names in a frenzy, and the aggregator feed carried micro-cap spikes that it flagged as low signal itself — up 223%, up 25%, up 17%, on a UK approval. Against that, the same feed noted Friday's tech selloff gathering pace, with S&P and Nasdaq futures slipping. Momentum into the new week is two-sided, not directional.
The episode's own answer to a tape like this was about pace rather than names: on a genuine hundred-bagger you get many bites at the apple, and buying after the first double, with some of the risk already taken out, still compounds. The post-listing run-up in SpaceX was the cautionary example attached to it.
3The escalation in Hormuz is being priced through rates, not through risk
Overnight the wires carried renewed US–Iran strikes and a tanker struck in the Strait of Hormuz, described as the worst escalation since the peace deal, followed by a reported agreement between the two sides to halt attacks and reopen talks. Equities took the second half of that sequence. Pre-market futures were up — S&P 0.73%, Nasdaq 1.09%, Dow 0.36% — and the VIX eased 0.60% to 18.30. Asia was mixed, with tech weakness in Japan and Korea.
The instructive move is gold's. Oil climbed on the strikes, and gold slipped, which Reuters ties to firming bets on a Fed rate hike as the oil move feeds a hawkish repricing. That is the mechanism in one line: a supply shock in energy reads as an inflation event before it reads as a risk event, so the rates leg moves first, and gold trades as a real-yield asset rather than as a haven.
Which matters mostly for what happens when the headline turns. Equities here are priced off the halt, and a halt is a headline that can reverse inside a session. The rate-hike leg is the part that persists either way, because it is the oil price doing the work rather than the conflict itself — and it puts duration and gold on the same side of a war headline, which is not the side either normally sits on.
The week's earnings calendar cannot settle either argument
The scheduled reports are small- and mid-cap industrials: , and before the open on 1 July, , , and after it, and on 2 July. None of them speak to the memory cycle or to energy. The week is also shortened around US Independence Day on Saturday 4 July, so an early close and thin liquidity are likely into the back half.
That leaves the fluid inputs as the deciding ones: Hormuz headline risk, which is live, Fed bank-oversight discussions, and and flagged for tech-sector volatility watch. Both of the day's arguments — whether AI capex is flattering profits, and whether an energy shock is really an inflation event — get marked by headlines this week rather than by anything on the calendar.
