The tech derisk is no longer a US-only story — Asia is selling the same names, which means dip-buying the US session has to fight global flows.
Digest · Mon Sep 14 · 10:25 PM PT
Mon Sep 14 · 10:25 PM
The AI trade spent today arguing with itself, which is my favorite kind of tape. Overnight Asia deepened the tech selloff — weak China retail data (5.3% urban unemployment) did nothing to help — while Broadcom's CEO came out and told everyone the AI buildout is just fine, long-term revenue goals intact. Two credible stories, one market. When the debate is this loud, prices are doing the honest thing by bouncing around. Meanwhile the old-school macro forces kept squeezing. Bonds are back as the market's boss — 'The Bond Market Strikes Back' is the kind of headline that matters more for long-duration growth names than any earnings print — and Saudi supply uncertainty has oil extending gains, which is a quiet tax on everything else. Bitcoin stalled as Clarity Act odds slipped. This is the environment our five live in: no help from macro, all alpha has to come from the companies themselves. Our names mostly sat it out, with one loud exception. Nebius fell 5.5% to $212 after Bernstein flagged that neoclouds — CoreWeave explicitly, with Nebius named in the basket — are the most exposed names if frontier model development slows. I read that as priced-in paranoia at a 46x sales multiple rather than a thesis-breaker: the thesis continues, but the stance demands respect for the drawdown mechanics. TTD bounced 4.4% off the floor — a dead-cat until revenue growth says otherwise, and my neutral stance is unchanged. FOUR, RKLB, and OKLO were essentially flat and news-quiet; their theses stand exactly as written.
AI capex conviction vs. valuation anxietyTech selloff goes global overnightRates and oil reassert macro pressureChina demand data softens again
Market pulse
Management conviction vs. analyst anxiety — when the builders say the order book is fine, the 'AI slowdown' thesis needs better evidence than multiple compression.
Capital is still chasing AI silicon alternatives — good for the ecosystem's breadth, but more competition for the neoclouds' GPU stacks to route around.
Soft demand in the world's second economy is the excuse Asia needed to keep selling; watch whether it bleeds into the ad-spend recovery TTD is waiting on.
Elevated oil is a stealth tax on consumers and airlines — relevant context for FOUR's venue/hospitality volume, though no direct hit yet.
Duration risk is back on the agenda — for names like NBIS and OKLO that trade on cash flows years out, the discount rate is the chart to watch.
The capex super-cycle isn't just GPUs — industrial megaprojects argue the build-out thesis has legs beyond the hyperscaler quarter-to-quarter debate.
Risk-off ripples — crypto stalling on regulatory odds is a sentiment read, not a fundamentals one, for the broader growth complex.
Tracked companies
TTD
neutral
$14.97 +4.4%
The Trade Desk, Inc.
No take change: still neutral, and a 4% bounce off a 48%-style drawdown is a dead cat until revenue growth proves otherwise. The thesis is explicit that cheap isn't a thesis while the next quarter guides to a revenue decline. The watch item is unchanged: stabilization of growth and resolution of the agency fee disputes. A green day doesn't rewrite that.
FOUR
bullish
$45.32 +0.6%
Shift4 Payments, Inc.
Thesis CONTINUES with nothing to react to. Profitable payments at 10x earnings remains the pitch; the oil bid and China softness are worth a glance since hospitality volume is the end market, but there's no company-specific signal today. Bullish stance holds.
NBIS
bullish
$212.19 -5.5%
Nebius Group N.V.
Thesis CONTINUES, but this is exactly the drawdown mechanic the risk section warned about: at 46x sales there is no room for a demand scare, and today we got one. Bernstein's point is about pace-of-development risk for pure-play neoclouds — fair, and worth watching. But the counter is right there in the feed: the Palantir partnership is precisely the kind of demand-base broadening that de-risks the 'hyperscaler whim' story. Bullish stance holds; the thesis is working as designed, which includes being volatile.
RKLB
bullish
$62.55 -0.6%
Rocket Lab Corp
Thesis CONTINUES, nothing to update. The drawdown (58% off highs) is the position; the open variable remains Neutron execution and launch cadence, neither of which moved today. Bullish stance holds.
OKLO
bullish
$36.21 -0.0%
Oklo Inc.
Thesis CONTINUES, nothing to update. Pre-revenue nuclear options don't trade on daily news; the demand side (datacenter power deals across the industry) is the slow variable and it didn't change today. Bullish stance holds at small size.
Spotlight · NBIS
Bernstein vs. Broadcom: the AI demand debate lands on the neoclouds
Today was a proxy fight over the single question that matters for Nebius: does AI compute demand keep growing fast enough to justify 46x sales? Bernstein put the neoclouds — CoreWeave first, Nebius in the basket — at the top of the exposure list if frontier model development slows. Broadcom's CEO, on the same day, said the buildout is fine and stood by long-term revenue goals. Both can't be the whole story, and the market split the difference: NBIS -5.5%, the complex uneasy. My read: Bernstein's risk is real but it's a pacing risk, not a demand-destruction risk — and Nebius's answer is diversification of the customer base, with the Palantir partnership as the latest evidence. The bullish thesis explicitly priced this volatility in; it just arrived on schedule. No stance change, but I'll be watching whether the neocloud basket keeps underperforming the hyperscalers — that's the tell for whether this is healthy derisking or something structural.