Advanced thermal management
Cooling and workload control increasingly decide how much computing a fixed power connection can run. NVIDIA describes 45°C liquid-cooling inlets and dynamic power allocation for Rubin systems. Higher inlet temperatures may reduce chiller use, depending on climate and facility design. The capacity gains are projections that have not been measured at every site, and direct liquid cooling is different from immersion cooling.
Research snapshot · updated Nov 12, 2026
Core thesis
Each Nvidia B200 rack draws 120kW — beyond air cooling capacity. The data center industry is transitioning from HVAC to precision liquid/immersion cooling. Submer, GRC (immersion), CoolIT (direct liquid), and Vertiv (rear-door heat exchangers) lead. Cooling represents 30-40% of data center energy. The companies that solve AI thermal management unlock GPU density that defines who can train frontier models.
State of the art (2026)
Direct-to-chip liquid cooling is now the default for frontier AI racks, not an option. NVIDIA's GB300 NVL72, shipping through 2026, removes roughly 90% of heat via cold plates with B300 GPUs at 1,400W TDP and racks near 135kW – beyond any air-cooled envelope. The Vera Rubin VR200 NVL72, due in H2 2026, is 100% liquid-cooled at 190–230kW per rack on an 800V DC architecture, with Rubin Ultra's Kyber rack targeting ~600kW in 2027. Vertiv, Schneider, CoolIT, Boyd and Motivair supply cold plates and CDUs; immersion players Submer, LiquidStack and GRC face a fluids reset after 3M ceased Novec manufacturing at end-2025, pushing two-phase vendors onto Syensqo Galden and HFO-based alternatives.
Explore the deeper analysis in CanaryIQ
Explore what is covered below. These previews show the structure; log in to read the full analysis.
Signal stack
Evidence stacked leading → lagging
9 signals
Signal stack
Evidence stacked leading → lagging
Technology-native KPIs
Metrics that predict trajectory, tracked over time
3 tracked
Technology-native KPIs
Metrics that predict trajectory, tracked over time
Landscape map
Who builds what, and who depends on whom
208 players · 6 layers
Landscape map
Who builds what, and who depends on whom
Catalyst calendar
Dated events that will move the position
3 ahead
Catalyst calendar
Dated events that will move the position
Technology roadmap
Milestones on the path to maturity
8 milestones
Technology roadmap
Milestones on the path to maturity
Watchlists
Companies, people and papers, each with a condition for removing it
20 · 20
Watchlists
Companies, people and papers, each with a condition for removing it
Decision frameworks
What our view means for your job
Locked
Decision frameworks
What our view means for your job
Thesis changelog
When our view changed, and why
8 updates
Thesis changelog
When our view changed, and why
Change our mind
2 disconfirming conditions
You've read the verdict. The file is much deeper.
The full signal stack, technology-native KPIs tracked over time, the landscape of who depends on whom, the dated catalyst calendar, decision frameworks for every desk, live watchlists and a record of every time our view on Advanced thermal management has changed. All of it is in CanaryIQ.