While Nvidia's name has become closely associated with the AI surge, a less visible but increasingly important group of beneficiaries are companies that supply the power and cooling equipment for datacenters. Demand for AI infrastructure now strains not only high‑performance chips, but also transformers, generators, power distribution systems and cooling technologies.
Scale and timelines
McKinsey estimates nearly $7 trillion of datacenter investment could occur worldwide by 2030. A growing constraint on deployments is not compute capacity itself but the energy infrastructure needed to run it: developers want to commission facilities faster, but grid connections and supporting infrastructure often delay projects. Pivotale AI told Reuters that in some emerging markets grid connection can take up to two years, while in many developed markets the wait can exceed eight years.
Early market signals
South Korea's HD Hyundai Electric saw its order backlog rise 23 percent over six months to $8.5 billion by the end of June. Chinese Jinpan Smart Technology reported that new datacenter‑related orders in the first half of the year grew more than fourfold year‑on‑year, and its associated order backlog nearly tripled. Many suppliers say they now have multi‑year order books — in some cases more than three years — and are even negotiating deliveries scheduled for 2030.
More power, more cooling
AI chips consume substantially more electricity. Bank of America estimates an AI rack could consume over 1.5 megawatts by the end of 2030, roughly a hundred times a traditional rack. That increases both power delivery needs and heat to be removed. Consequently, liquid cooling is gaining share alongside traditional air cooling: Bank of America projects liquid cooling could be used in about 70 percent of new AI datacenters by 2030, up from roughly 30 percent today. McKinsey estimates liquid cooling can reduce energy use by more than 27 percent.
Suppliers positioned to benefit within the Nvidia ecosystem include Taiwan's Delta Electronics, Asia Vital Components and Auras Technology, as well as China's Envicool, among others.
New technologies and competitive dynamics
The industry is also exploring new technical approaches to cope with rising power needs. One example is the solid‑state transformer (SST), which uses power electronics instead of large magnetic cores and windings to transform and transmit electricity. UBS estimates SSTs could improve efficiency and reduce costs by about 4 percent and suggests the technology could reach up to 40 percent market penetration by 2030. Chinese manufacturers may gain share on cost and scale grounds.
Developers are also increasingly considering solutions partially or fully independent of the grid. Some unconventional concepts have surfaced — from onsite generation using aircraft engines to floating or underwater datacenters and servers sited in caves or tunnels — but the core issue remains securing reliable high‑power and cooling capacity.
Not every supplier will win equally
Rapidly rising demand does not guarantee uniform gains across suppliers. Some stocks have already jumped: Delta Electronics' share price has risen more than 90 percent this year, while HD Hyundai Electric's performance has been more muted. Competition, component shortages and potential project delays could pressure margins.
Bank of America analysts warn against treating the sector as a single, guaranteed winner; success will likely accrue to suppliers that actually secure contracts from the largest datacenter developers.
Conclusion
The spread of AI datacenters increases demand not only for chips but for megawatts of reliable power and substantially greater cooling capacity. In the coming years the critical question will be not only who makes the best processors, but who can deliver the energy and thermal infrastructure necessary to run next‑generation AI facilities.
This article is not investment advice or a recommendation.
