AI Data Centre Boom Fuels $7 Trillion Rush for Power and Cooling Infrastructure

Massive Infrastructure Investment Reshapes Global Power Markets

The AI data centre boom is triggering one of the largest infrastructure buildouts in modern history. McKinsey1

1

estimates nearly $7 trillion could be invested in data centres globally by 2030, creating unprecedented opportunities for power and cooling equipment suppliers racing to overcome critical infrastructure bottlenecks. PwC projects cumulative global data centre capital expenditure could reach $31.6 trillion through 20502

2

, with annual spending climbing from $800 billion in 2026 to $1.8 trillion by 2050.

While Nvidia dominates headlines around AI workloads, a lesser-known group of power and cooling equipment suppliers is emerging as critical enablers of this expansion. Energy-intensive AI data centre infrastructure is driving demand for specialized equipment ranging from transformers and power-management systems to advanced liquid cooling technologies5

5

. The shift is creating winners across Asia's supply chain, though high valuations and supply constraints remain key risks for investors.

Grid Connection Delays Create Major Deployment Challenges

Expanding data centre capacity is becoming increasingly difficult as developers face severe delays in securing power and connecting new facilities to electricity grids. Consultancy Pivotale AI estimates grid connection delays can reach 24 months in some emerging markets and more than 8 years in major developed economies1

1

. These lengthy timelines are putting greater focus on critical equipment such as transformers, which convert high-voltage electricity from the grid into levels suitable for servers, cooling systems and power distribution units.

Hyperscalers typically want facilities delivered within 6 months, creating intense pressure on supply chains5

5

. Wing Kin Cheung, CEO of digital infrastructure service provider BodaData, noted that within the industry circle, people question lead times for generators and transformers rather than focusing solely on GPU clusters. This infrastructure constraint is fundamentally reshaping how the AI data centre expansion unfolds globally.

Transformer Manufacturers Report Surging Order Backlogs

Leading transformer manufacturers are seeing sharp increases in orders linked to AI infrastructure projects, particularly in North America. South Korea's HD Hyundai Electric reported strong demand during the first half of 2026, with the company's order backlog increasing 23% to $8.5 billion at the end of June from six months earlier1

1

. The company expects data centre demand to remain robust and has production capacity for major power equipment largely committed for the next three years.

HD Hyundai Electric told Reuters it currently has an order backlog covering more than three years and is in discussions with key customers for volumes scheduled for delivery as far out as 20305

5

. The company is also seeing increased interest from Europe as U.S. hyperscalers expand investments in Finland, Germany and Britain, while demand from the Middle East has remained strong.

China's Hainan Jinpan Smart Technology has similarly benefited from this trend. New data centre orders in the first half of 2026 more than quadrupled from a year earlier, while its related backlog nearly tripled1

1

.

Power Consumption Per Rack to Jump 100-Fold by 2030

The rapid growth in AI computing is dramatically increasing power demand at the rack level. Bank of America estimates power consumption per AI rack could rise to more than 1.5 megawatts by the end of 2030 based on Nvidia's technology roadmap1

1

. That would be nearly 100 times the power consumption of a conventional rack, fundamentally changing infrastructure requirements.

Traditional data centre facilities were built for standard server racks requiring 5 kW to 10 kW of power4

4

. Today, advanced AI workloads processing massive language models and real-time analytics rely on dense GPU clusters forcing rack densities up to 30 kW, 50 kW, and beyond 80 kW per rack3

3

. When power density reaches these levels, off-the-shelf power and cooling equipment hits hard limits, demanding custom-built hardware engineered at the factory level.

Solid-State Transformers Emerge as Efficiency Solution

One technology attracting increasing attention is the solid-state transformer, or SST. Unlike traditional transformers that rely on bulky magnetic components and copper windings, SSTs use semiconductor technology to transform and route electricity. UBS estimates SSTs could improve power efficiency by around 4% and reduce costs1

1

. Commercial adoption remains at an early stage, but the bank expects penetration to reach 40% by 2030 and sees Chinese manufacturers gaining market share because of their technological capabilities and cost advantages.

HD Hyundai Electric and Jinpan are expanding their work on SST technology5

5

. Taiwan's Delta Electronics, another major power infrastructure supplier, has said a small data centre is already using its SSTs. Delta told Reuters that demand for AI power, cooling and data centre infrastructure solutions remains a key growth driver, and the company is expanding its production footprint in Thailand, the United States and China to meet rising demand.

Liquid Cooling Technologies Gain Rapid Market Share

Thermal management represents another critical area where specialized manufacturing is essential. Bank of America expects liquid cooling to account for 70% of new AI data centre installations by 2030, compared with around 30% currently1

1

. McKinsey estimates liquid cooling can reduce energy consumption by more than 27%, making it increasingly attractive as power demand escalates.

Matty Zhao, Bank of America's Asia-Pacific head of research for basic materials, oil and gas, explained that power and cooling go hand in hand—the more power used, the more cooling needed because of heat generation5

5

. Standard air conditioning simply cannot handle modern GPU clusters, and the risk of thermal throttling or total power failure becomes a daily operational challenge when rack densities exceed 30 kW3

3

.

Direct-to-chip liquid cooling systems are becoming critical for high-density compute environments. Companies like Delta Electronics are deploying specialized Liquid-to-Liquid and Liquid-to-Air Cooling Distribution Units that route cooling directly to high-TDP processors3

3

. CoolIT Systems builds custom cold plates, fluid distribution manifolds, and heat-exchange units that sit directly on top of high-performance chips, keeping processors running cool under continuous parallel processing loads.

India Emerges as Major Beneficiary Despite Power Challenges

India could emerge as one of the biggest beneficiaries of the global AI data centre boom, driven by demand for sovereign AI, local cloud infrastructure and AI-powered services. The Asia-Pacific region could attract $8.2 trillion in cumulative data centre capital expenditure through 2050, with India and China emerging as the biggest sources of incremental demand2

2

. Large populations, rapidly growing digital economies and significant headroom for AI adoption make both markets critical to the next phase of buildout.

India's data centre capacity has expanded dramatically. In 2020, total capacity stood at 375 megawatts. By 2025, that figure skyrocketed to roughly 1,500 MW, and the country is on track to reach between 1,700 and 2,000 MW through 20264

4

. S&P Global says India's data centre IT load stood at 1.4 GW during April-June 2025, with another 1.4 GW under construction2

2

. More than 5 GW of additional IT load is expected to come online by 2030.

Technology companies have announced close to $300 billion of investments in India's data centre infrastructure over the past two years2

2

. According to BloombergNEF, Reliance Industries and the Adani Group have collectively committed $210 billion to data centre development in India, while Google, Amazon and Microsoft together have announced $84 billion of investment.

Power Infrastructure Expansion Becomes Critical Constraint

Power and grid infrastructure expansion pose major challenges for India's AI ambitions. India's data centre electricity consumption was around 13 TWh in 2024, accounting for 0.8% of the country's total electricity demand2

2

. By 2030, S&P Global Commodity Insights expects data centre power demand to rise almost fivefold to 57 TWh, taking the sector's share of India's electricity demand to around 2.6%.

A March 2026 parliamentary reply citing the Ministry of Power put electricity demand from data centres at 13.56 GW, while an August reply citing the Central Electricity Authority put it at around 17 GW2

2

. A July 2026 report said states had submitted AI data centre projects that could add 26.3 GW of grid load, highlighting the scale of capacity the power system may eventually need to accommodate.

S&P Global estimates 15-30 GW of additional renewable capacity will be needed over the next five years to meet projected data centre power demand2

2

. While this represents less than 10% of expected renewable capacity additions, the challenge lies in timely grid expansion to move power where needed. Data centre operators have already announced more than 3 GW of renewable procurement deals for live and pipeline projects in India.

Specialized Manufacturing Capabilities Drive Competitive Edge

Building AI-ready data centres requires robust Original Design Manufacturing capabilities to design, build and scale complex power and thermal hardware4

4

. Companies like Delta Electronics India are manufacturing megawatt-scale power shelves and high-efficiency UPS systems specifically for compute-dense environments3

3

. Eaton India provides high-density Power Distribution Units and energy-storage-ready UPS setups designed to buffer heavy load fluctuations.

Stulz India specializes in custom direct-to-chip cooling infrastructure and high-capacity chillers engineered for dense server rooms3

3

. Rittal India manufactures industrial-grade modular rack systems and climate-controlled enclosures engineered to bear heavy structural loads, with sealed rack designs integrating liquid lines directly into enclosure frames. These specialized capabilities are becoming essential as the industry transitions toward high-voltage direct current systems and 800VDC architecture products to support high-power demands efficiently.

The supply chain bottlenecks and infrastructure challenges facing the AI data centre boom highlight a critical reality: sustained AI growth depends not just on cutting-edge chips, but on the unglamorous hardware that keeps those chips powered and cool. As deployment timelines stretch and power demand multiplies, the companies solving these fundamental infrastructure problems are positioning themselves at the center of a multi-trillion-dollar transformation.