TradeRadar logo
    Home/AI Infrastructure Stack/Inside the Building
    Part 7 of 8
    Power & cooling
    3 Jun 2026

    Power and Heat: AI's Data Center Cooling Companies

    An AI rack draws the power of a small neighborhood and runs too hot for air. Meet the data center cooling companies — Vertiv and Eaton — keeping the racks alive.

    Key Takeaways

    • 1This article covers key developments in the crypto market
    • 2Always verify claims with official FCA and regulatory sources
    • 3Past performance does not guarantee future results
    • 4Consider speaking to a qualified financial adviser before acting
    • 5TradeRadarNews provides information only — not financial advice
    An AI rack can draw as much power as a small neighborhood and runs hotter than air can cool. Meet the data center cooling companies — led by Vertiv and Eaton — that keep the racks alive.

    Inside the buildings is a physical problem most coverage skips: a rack of modern AI accelerators concentrates an extraordinary amount of power and heat into a small space. Densities have climbed to the point where blowing cold air across the chips is no longer enough, and operators are piping liquid directly to them. Delivering that power cleanly and carrying that heat away is its own industry — the layer of pumps, plates, busbars, and backup power that sits between the building and the chips. It is unglamorous and absolutely load-bearing.

    Where this sits in the stack. This layer lives inside the buildings of the AI infrastructure stack, draws its power from the grid, and exists to keep the chips running safely.

    Why density changed everything

    For years data centers were cooled with air, like a very large server room. AI broke that. As power per rack has climbed, the heat produced has outrun what air can remove, so liquid cooling — circulating coolant directly to the hottest components — has moved from exotic to standard for the densest rows, running alongside traditional air cooling for the rest of the floor. At the same time, getting that much electricity to the rack reliably has made power distribution — uninterruptible supplies, switchgear, busbars — a growth business in its own right.

    Vertiv: the pure-play

    What it does. Vertiv (VRT) is the cleanest public proxy for data-center power and cooling, with roughly 80% of its revenue tied to data centers.

    The numbers. First-quarter 2026 revenue was about $2.65 billion, with organic sales up 23% and the Americas up 44% organically; adjusted earnings per share rose 83%, per its results. The backlog stood above $15 billion, up 109% year over year, at a 2.9-times book-to-bill — meaning Vertiv is booking nearly three dollars of new orders for every dollar shipped. It raised full-year 2026 guidance to roughly $13.5–14.0 billion in revenue.

    The edge. Leadership in liquid cooling, deepening ties to Nvidia's reference designs, and the credibility of joining the S&P 500 in March 2026.

    The risk. The stock has re-rated hard on the AI thesis; the case now rests on the order book converting on schedule, and project timing can slip.

    Liquid cooling pipework around server racks inside a data centre
    Liquid cooling has moved from exotic to standard in AI deployments. Image generated for editorial use.

    Eaton: the electrical layer

    What it does. Eaton (ETN) is a broad electrical-infrastructure supplier, with roughly 70% of revenue in electrical segments, increasingly pulled by data-center demand.

    The numbers. Eaton has cited unprecedented data-center demand, posted high-single-digit organic growth, and in March 2026 acquired Boyd Thermal to add liquid-cooling capability — a business projected to contribute on the order of $1.5–1.7 billion in 2026.

    The edge. A far broader electrical portfolio than a pure-play, now extended into cooling, with a planned spin-off of its vehicle unit to sharpen focus.

    The risk. Data center is one slice of a diversified industrial business, so even strong segment growth is diluted at the company level.

    The valuation question

    The bull case is the order book itself: backlogs, book-to-bill ratios above two, and a structural shift to liquid cooling that is still early. The bear case is that these names have run hard, trade at rich multiples, and depend on hyperscaler capex staying elevated and projects landing on time. Both are defensible; the physics of the heat problem is not in doubt, but the timing and the valuation are.

    Rows of glowing servers inside a modern AI data centre
    AI data centres now anchor a multi-trillion dollar infrastructure build-out. Image generated for editorial use.

    What this layer feeds

    This hardware keeps the chips alive inside the buildings, drawing on the grid to do it. For the full map, start with the series overview.

    This article is for information only and is not investment advice or a recommendation to buy or sell any security. TradeRadarNews is not a licensed financial adviser. Figures are accurate as of June 2026 and will change. Do your own research.

    Risk Warning: Trading and investing carries significant risk. Your investments can fall as well as rise. CFDs carry high risk of rapid loss due to leverage. Cryptocurrency is not FCA-regulated and not covered by FSCS. This is information only, not financial advice. Seek independent advice before investing.

    Written by

    TradeRadarNews Team

    Editorial Team

    Our editorial team covers markets, fintech, and regulatory developments across the UK and globally.

    Frequently Asked Questions

    Back to the series overview

    Risk Warning: Trading and investing carries significant risk. Your investments can fall as well as rise. CFDs carry high risk of rapid loss due to leverage. Cryptocurrency is not FCA-regulated and not covered by FSCS. This is information only, not financial advice. Seek independent advice before investing.

    We use cookies to improve your experience.