Market snapshot, not real-time
S&P 500 (SPY) $748.62 +0.88% Nasdaq-100 (QQQ) $708.90 +1.85% Dow (DIA) $522.55 +0.89% Russell 2000 (IWM) $295.82 +1.20% 10-Year Treasury (IEF) $93.32 −0.23% Crude Oil (USO) $128.37 +2.28% Gold (GLD) $373.85 +1.70% US Dollar Index (UUP) $28.45 +0.23% Volatility (VXX) $21.29 −3.77% Semiconductors (SMH) $583.24 +4.37% Silver (SLV) $53.36 +4.67% Emerging Markets (EEM) $65.45 +2.97% Bitcoin (BTC) $66,470.13 +1.90% Ethereum (ETH) $1,922.88 +1.00% S&P 500 (SPY) $748.62 +0.88% Nasdaq-100 (QQQ) $708.90 +1.85% Dow (DIA) $522.55 +0.89% Russell 2000 (IWM) $295.82 +1.20% 10-Year Treasury (IEF) $93.32 −0.23% Crude Oil (USO) $128.37 +2.28% Gold (GLD) $373.85 +1.70% US Dollar Index (UUP) $28.45 +0.23% Volatility (VXX) $21.29 −3.77% Semiconductors (SMH) $583.24 +4.37% Silver (SLV) $53.36 +4.67% Emerging Markets (EEM) $65.45 +2.97% Bitcoin (BTC) $66,470.13 +1.90% Ethereum (ETH) $1,922.88 +1.00%

Opinion · July 2026

The 0.79 Correlation Between Nebius and CoreWeave Wasn't Academic Last Week

July 26, 2026 · Weekend analysis, built on data already published this week

The correlation matrix on the Holdings page has shown NBIS and CRWV moving together at 0.79 for weeks — a number I've cited more than once as the reason I treat NBIS, NBIL, and CRWV as one correlated cluster for sizing purposes, not three independent bets. Friday gave that number a live test. Nebius fell 15.02% to $187.77. CoreWeave fell 11.37% to $71.88. Same day, same underlying worry — capex plans against a revenue base the market wasn't willing to fund on faith anymore — and CoreWeave was actually the one with better-looking fundamentals that day, up 112% year over year in revenue with a $99.4 billion backlog. It fell hardest anyway. A 0.79 correlation isn't a hedge against a bad week; it's a promise that a bad week for one is a bad week for the other, and that's exactly what happened.

What a correlation number is actually for

It's tempting to read a correlation matrix as a risk-reduction tool — diversify across enough names and the whole book smooths out. That's true across most of this book's 32 positions, where most pairwise correlations sit under 0.2. It is not true inside the NBIS/NBIL/CRWV cluster, and the matrix has said so plainly the entire time. The number's actual job is narrower and less comforting: it tells you, in advance, which of your positions are going to have their worst days on the same calendar day, for the same reason, regardless of how differently their individual businesses are performing. CoreWeave's 112% revenue growth and Nebius's own AI-infrastructure buildout are genuinely different stories at the company level. At the correlation level, on a day the market decided to stop trusting neocloud capex plans generally, they were the same trade.

Where this connects to my own sizing

The reason NBIS, NBIL, and CRWV get aggregated as one position for concentration purposes — a rule stated on the Investment Policy Statement and applied on the Compliance Ledger — isn't a technicality. It's this exact scenario. Individually, NBIS is 1.31% of the book, NBIL is 2.81%, and CRWV is 1.25% — each comfortably small on its own. Priced through their real correlation instead of their labels, Friday showed what they actually are: one bet on neocloud-capex sentiment, sized at roughly 5.4% combined, that had a rough day all at once. The IPS rule existed before Friday happened; Friday is just the first time this cycle it was easy to see why.

This is an opinion piece reflecting my own interpretation of data already published on the Holdings page and in Friday's interim note. No new prices or correlation figures were introduced to write it. It is not investment advice.