Nicholson: NBIS Best Positioned Against CRWD & APLD, Data Center Risks Remain

The market is likely underestimating how much this theme is really a cost-of-capital trade, not a pure demand trade. In a capacity race where customers want optionality and uptime, the operator with the cleanest balance sheet can sign better deals, survive pricing noise, and keep bidding for scarce power/GPU inventory without needing to tap capital markets at a punitive time. That makes NBIS a relative winner versus more levered peers because equity holders are effectively long the spread between AI infrastructure demand and the financing terms required to fund it.

The second-order loser is not just the most levered stock; it is any model that assumes growth can be monetized before refinancing risk becomes visible. If the market’s risk appetite cools even modestly over the next 1-3 months, debt-heavy operators can be forced into slower buildouts, sale-leasebacks, or dilution, which tends to compress multiples faster than fundamentals change. That also creates a feedback loop: weaker balance sheets can lose competitive bids for enterprise contracts even when headline demand remains strong.

Contrarian view: the consensus may be too focused on demand durability and too casual about the bottleneck stack. If power, interconnects, or GPU availability are the real constraints, then revenue timing slips and balance-sheet strength matters less than execution timing, which can narrow the spread between names. The thesis is most vulnerable if capital markets reopen aggressively, rates fall, or one of the levered peers secures long-dated, inexpensive financing; that would reduce the premium for financial prudence and flatten the relative performance gap over the next 1-3 months.