Liquid Cooling TCO: A Vendor-Neutral Financial Model for Direct-to-Chip vs. Immersion

The Datacenter Economist

28 Jun 2025 — 4 min read

The era of air cooling for high-density AI is over. As operators race to support 100kW+ racks, the critical decision is no longer if they will adopt liquid cooling, but how. The two dominant technologies—Direct-to-Chip (D2C) and Single-Phase Immersion—present radically different cost structures, infrastructure requirements, and operational risks.

Choosing the wrong path can lead to millions in stranded capital and a competitive disadvantage. This is the definitive, vendor-neutral 5-year Total Cost of Ownership (TCO) analysis that breaks down the true cost of survival.

(To access the full financial model and our detailed verdict, please subscribe to The Datacenter Economist.)

The AI Land Crunch: Why U.S. Data Centers Are Running Out of Places to Build

AI megacampuses are accelerating across the U.S., but industrial land, zoning approvals, and community support are running out. A growing “AI Land Crunch” is emerging as the next major constraint on data-center expansion.

America’s Next Energy Shortage Isn’t Oil — It’s Transformers

We used to fear running out of oil. Now developers fear running out of transformers. As AI and electrification push America’s grid to its limits, the quiet crisis isn’t megawatts — it’s the hardware that delivers them. Here’s why transformer scarcity is becoming the new energy bottleneck.

The Transmission Upgrade Arbitrage: How Policy, Permitting, and Equipment Scarcity Are Repricing Time-to-Power

🕒 The New Commodity: Time-to-Power In 2025, the most valuable resource in digital infrastructure isn’t megawatts — it’s months. Developers and investors once modeled their projects around access to land, capital, and power purchase agreements. But a convergence of regulatory reform, grid congestion, and manufacturing scarcity has created something new:

The Hidden Schedule Premium: Why Lead Time Is the New Risk Class in Data Center Investing

Executive Summary In a sector where speed to market is as critical as cost control, the traditional focus on capex and latency is now being overshadowed by schedule risk—specifically, the risk of long, volatile, and opaque lead times. Equipment deliveries, key long-lead packages, and module kits are no longer

Read more

The AI Land Crunch: Why U.S. Data Centers Are Running Out of Places to Build

America’s Next Energy Shortage Isn’t Oil — It’s Transformers

The Transmission Upgrade Arbitrage: How Policy, Permitting, and Equipment Scarcity Are Repricing Time-to-Power

The Hidden Schedule Premium: Why Lead Time Is the New Risk Class in Data Center Investing