Happy Monday. This week belongs to the Big Four earnings reports — and they deserve the attention. Amazon, Microsoft, Alphabet, and Meta all reported Q1 2026 results on April 29, collectively posting roughly 60% earnings growth year-over-year while announcing a combined $725 billion in 2026 capital expenditures. That’s a 77% increase from the record $410 billion they spent in 2025. To put it plainly: in one calendar year, four companies plan to spend more on AI infrastructure than the entire U.S. interstate highway system cost to build in inflation-adjusted dollars.
The headline that matters most for our industry isn’t the profits — it’s the supply constraint. Microsoft stated for the third consecutive quarter that demand is outpacing available capacity, and CFO Amy Hood confirmed the company expects to remain supply-constrained through at least the end of 2026. AWS told a nearly identical story. These are not companies short on capital. They are companies short on energized, operational data center space. That single dynamic — capital abundance meeting infrastructure scarcity — is what’s driving every deal, every site selection decision, and every power negotiation in this market right now.
Two other stories worth your attention this week: Kevin O’Leary’s Utah hyperscale project is approaching final approval with a tax incentive structure that could become a template for secondary market development, and the Middle East data center market is flashing a new and unusual risk signal — geopolitical instability around the Iran conflict is causing major operators to pause investment decisions and begin modeling hazard pay premiums for on-site workers. That’s a first for this industry.
The buildout continues at historic speed. The constraints are real but being worked around by operators with discipline and the right assets in place. As always, reach out if any of this week’s news connects to something you’re working through.
Dell’Oro Group’s January 2026 report declared it plainly: “Liquid cooling has crossed a critical threshold. What was once treated as an optional efficiency upgrade is now a functional requirement for large-scale AI deployments.” The data center liquid cooling market is projected to reach $7 billion in manufacturer revenue by 2029, growing from a base that was largely nonexistent five years ago. That trajectory is being driven by one unavoidable physics problem — AI GPU racks are generating heat loads that air cooling systems simply cannot handle.
The numbers tell the story. Average data center rack power density increased 38% from 2022 to 2024 alone. Today’s AI GPU clusters routinely operate at 80–120kW per rack; cutting-edge deployments are pushing past that. Traditional air-cooled data centers are designed for 5–15kW per rack. There is no retrofit path that bridges that gap with air alone. Direct-to-chip (D2C) cooling — where cold plates are mounted directly onto CPUs, GPUs, and memory modules — removes 70–80% of heat at the source. Combined with rear-door heat exchangers handling the remainder, D2C allows existing facility shells to support dramatically higher densities without full reconstruction.
The competitive landscape is consolidating quickly. Vertiv leads the market with roughly 3% global share, followed by CoolIT, nVent, and Boyd. Trane Technologies made a significant move in February 2026 by agreeing to acquire LiquidStack, signaling that industrial HVAC players see liquid cooling as the future of their data center business. Aaon has delivered rapid share gains through deep hyperscaler partnerships and highly customized solution delivery. The two-phase immersion segment — where servers are fully submerged in dielectric fluid that vaporizes to carry heat away — holds a projected 66% market share in 2026 and is gaining traction in the highest-density AI training deployments.
For operators and developers, the practical question is no longer “should we deploy liquid cooling?” It’s “how much of our new build capacity should be liquid-ready from day one, versus retrofitted later?” The consensus answer emerging from hyperscaler deployments: design for liquid cooling in all new AI-focused builds, because retrofitting is expensive, disruptive, and frequently results in underutilized capacity during the transition period.