AI Chips Need Liquid Cooling. That's Why Vertiv's Stock Rally Can Continue.

The key market mechanism is not simply higher equipment sales, but the transition from one-time build revenue to an installed-base annuity stream. That should support a premium multiple for VRT versus broader industrials, but only if service attach rates stay high and customers do not squeeze pricing once the initial buildout wave normalizes. The second-order winner may be operators with secured power and land, such as IREN, because in a power-constrained market the scarce asset is grid access, not just cooling hardware.

The main near-term risk is that AI demand remains strong while deployment slows: permitting, transformer lead times, and utility interconnect queues can delay monetization even when capex budgets look healthy. That creates a gap between backlog optics and cash conversion over the next 1-3 quarters. Falsification would show up as order growth slowing, gross margin flattening, or management commentary shifting from “expansion” to “schedule slippage.”

Contrarian view: consensus is treating every planned megawatt as fully addressable revenue for the cooling stack, but a meaningful share of economics may accrue to electrical infrastructure, civil contractors, and vertically integrated data center owners rather than pure-play thermal vendors. If liquid cooling becomes standardized, the install margin can compress even as unit demand rises. For NVDA, this is a reminder that infrastructure bottlenecks can cap end-market shipments without damaging long-run AI demand.