Data center infrastructure market: AI-driven CapEx pushing IT and facility equipment spending toward $1 trillion by 2030

In short

• Data center equipment and infrastructure spending reached $290 billion in 2024, largely underpinned by hyperscaler CapEx, according to IoT Analytics’ 186-page Data Center Equipment & Infrastructure Market Report 2025–2030.
• 7 key IT and facility data center infrastructure segments are the main beneficiaries of this spending, with sustained double-digit growth expected for each segment until 2030 and a total estimated market of $1 trillion by 2030.
• Innovation intensity is high: 4 key developments discussed include 1) server design changes, 2) new storage recording technologies, 3) AI-ready electrical architectures, and 4) innovative cooling techniques.

Why it matters

• For data center infrastructure vendors: With unprecedented spending in data centers underway, vendors need to understand where new opportunities lie and steer their product roadmaps accordingly.
• For data center operators: Operators should understand distinct vendor offerings as well as recent data center infrastructure and equipment innovations to aid in selecting the right data center build-out strategy.

Data center spending to surpass $1 trillion by end of decade. Fueled by the race for AI supremacy, the global data center infrastructure market is on course to surpass $1 trillion in annual spending by 2030, according to IoT Analytics’ 186-page Data Center Equipment & Infrastructure Market Report 2025–2030 (published November 2025). This growth is being propelled by the world’s largest technology companies. In 2024, spending reached $290 billion, with growth set to accelerate in 2025. Of the $290 billion, Alphabet, Microsoft, Amazon, and Meta invested nearly $200 billion in CapEx, a figure expected to climb by over 40% in 2025 as they rush to build the computational power needed to train and deploy next-generation AI models.

Investments in data centers create a cascading effect across the data center value chain. While the world watches the hyperscalers (as well as large colocation providers and other data center owners), the real story is the trickle-down effect of these investments. The investments in data centers drive infrastructure and equipment spending (e.g., servers, networking equipment, and electrical systems), which, in turn, drive investments in subcomponents and so on down the data center value chain. Overall, the biggest beneficiary of this investment currently is US-based chipmaker NVIDIA. From a data center infrastructure point of view, the biggest beneficiary is Taiwan-based electronics contract manufacturer Foxconn, with its server equipment, while France-based energy technology and automation solutions provider Schneider Electric also benefits as a market leader in helping distribute electricity to power these servers. Even companies not traditionally associated with data centers are seeing a boom in their portfolios, like US-based heavy equipment manufacturer Caterpillar and its backup generators.

Below is a look at how the ever-growing data center CapEx is being distributed, followed by a deep dive into key innovations in data center infrastructure and equipment technology.

Insights from this article are derived from

Data Center Equipment & Infrastructure Market Report 2025-2030

A 186-page report detailing the market for data center equipment incl. market size, competitive landscape, key trends and challenges and more.

Request sample

Already a subscriber? View your reports and trackers here →

Data center capital allocation: Where the $1 trillion is going

3 areas of data center spending: IT infrastructure, facility infrastructure, and “other”—with IT infrastructure the largest. Based on the research, CAPEX distribution is heavily skewed toward the direct engines of computation: IT infrastructure. The remainder is shared between facility infrastructure and other services. The following is a breakdown of IT and facility infrastructure spending.

IT infrastructure

IT infrastructure spending dominates. In 2024, 78% of data center spending went to IT infrastructure equipment, the core hardware that processes and moves data. Below is a breakdown of equipment in this area and the share of total data center spending for each type:

• Servers (61%): These systems are core to data processing. Leading vendors in this space include Taiwan-based electronics contract manufacturer Foxconn and US-based IT hardware manufacturers Dell, Super Micro Computer (Supermicro), and Hewlett Packard Enterprise (HPE).
• Networking (10%): Routers, switches, and firewalls connect and protect servers, storage, and external networks, helping to keep data flowing securely. US-based digital communications company Cisco is a leading vendor in this market.
• Storage (6.5%): These systems hold the vast datasets for AI models. Dell and US-based enterprise-grade storage solutions company NetApp are leaders in this area.

Facility infrastructure

Facility infrastructure accounts for 12% of data center CAPEX. Facility infrastructure includes systems critical to keeping data center facilities operational. The top 2 areas of this segment and their share of total data center spending are:

• Electrical systems (6.5%): These systems are responsible for power delivery and distribution. Schneider Electric and US/Ireland-based intelligent power management company Eaton are 2 of the leading vendors offering complete data center electrical solutions, while US-based construction, mining, and engine equipment manufacturer Caterpillar is a leading provider of backup power solutions.
• Cooling systems (3.2%): These systems are essential for managing the immense heat generated by AI hardware. US-based digital infrastructure solutions company Vertiv and Schneider Electric are leading providers of data center cooling systems.

Deep dive: 4 key technical innovations

The Data Center Equipment & Infrastructure Market Report 2025–2030 report identifies several trends and profiles the strategies of over 20 data center infrastructure and equipment suppliers. The following are 4 areas seeing innovation as a result of the heavy CapEx going into data centers.

1. Server design changes, driven by AI

AI demand is driving surge in server spending. Servers represent the largest segment of data center spending, accounting for approximately 63% of total investment in 2024. Unprecedented investment in AI workloads underpins this share and is driving a significant change in server design.

Key design changes include:

1. Rack-scale and composable/disaggregated architecture. To address the complexity of high-density deployment, vendors are offering pre-configured rack-scale systems. This modular approach, seen in Supermicro’s Data Center Building Block Solutions, helps deliver validated, integrated racks quickly to accelerate the deployment of complete AI factories.
2. High-density computing and accelerator-ready chassis. The emergence of highly specialized, GPU-optimized systems is key for modern AI training and inference workloads. GPUs (such as those from NVIDIA, the leading data center GPU vendor and the de facto standard-setter for AI at scale) are the basis for most of today’s AI servers. For example, in March 2025, Taiwan-based electronics manufacturer Pegatron announced its high-density GPU rack specifically built on the NVIDIA GB300 NVL72 platform.
3. Power delivery and electrical architecture at rack level. With high-density computing comes high power demand, and as performance increases each year, high-performance AI racks are expected to consume around 1,000kW by 2029, necessitating changes to electrical architecture. While data center electrical systems are adapting (discussed further below), at the rack level, servers are adopting rack power delivery units (PDUs) and busway trunking systems (such as Siemens’ SIVACON 8PS) for flexible, robust power distribution within the rack.

Nearly 5x data center server market growth by end of decade. As hyperscalers and major tech firms continue to accelerate their expansion of generative AI (GenAI) capacity, the global data center server market is projected to grow rapidly, nearly quintupling from $204 billion in 2024 to $987 billion by 2030.

2. New storage recording technologies

Storage technological maturity helps keep costs low. The fact that storage spending greatly trails server spending does not necessarily reflect prioritization; rather, it speaks to technological advancements in storage density. While solid-state drives (SSDs) handle high-performance needs, hard disk drives (HDDs) remain the preferred medium for bulk capacity, archival, and nearline storage due to their lower cost per terabyte.

New recording techniques increase storage capacity. To support ever-scaling capacity needs, HDD manufacturers have worked to overcome physical limitations of established data recording techniques. For over 15 years, the standard has been perpendicular magnetic recording (PMR), which uses conventional magnetic writing to store data bits but is limited in areal density, maxing out around 2 TB per platter.

However, the development of and transition to heat-assisted magnetic recording (HAMR) has enabled greater capacity. This technology fundamentally changes how data is written to the platter by incorporating a tiny laser diode on the read/write head that briefly heats each bit location before writing. This brief heating lowers the magnetic coercivity of the high-stability media, enabling the write head to store much smaller, denser bits. This allows HAMR drives to achieve significantly higher areal density, reaching approximately 3.6 TB per platter and enabling overall drive capacities of over 30 TB.

Further, because more data can be packed onto fewer drives, HAMR systems provide substantially lower watts per terabyte stored (approximately 60% less compared to baseline PMR). For large-scale cloud deployments, adopting HAMR can reduce the physical space required and cut power consumption by nearly 40% for the same capacity. The shift away from PMR represents one of the biggest jumps the industry has taken in the last decade.

Example of HAMR storage systems

• Seagate: In May 2025, US/Ireland-based data storage company Seagate announced that it had been sampling its latest Mozaic 4 platform (an HAMR-based, 4TB-per-disk HDD) to data center customers, with full production expected in 1H 2026. Seagate’s current mass-produced HAMR-based HDD, Mozaic 3+ (released in early 2024), currently handles 3TB per disk, and its flagship Exos data center HDDs hold up to 30TB.

3. AI-ready electrical architectures prepared to meet electrical demands

AI server racks are straining power infrastructure. The rise of high-density, power-hungry GPU-based servers has created exceptional energy demands for data centers. Traditional data centers were largely built around low-voltage distribution, which was adequate for general-purpose CPU racks drawing 5–15 kW. However, modern AI racks, particularly those utilizing powerful GPU architectures, are already exceeding 100 kW per rack, with peak densities projected to exceed 1,000 kW by 2029.

Vendors shift to scalable power architectures. In response and preparation for the future of AI computing, vendors are transitioning from low-voltage power distribution to more efficient, scalable solutions, such as medium-voltage (MV) architectures. Switzerland-based electrification and automation company ABB was an early innovator of the data center MV power architecture with its HiPerGuard UPS (uninterruptible power supply; released in 2021), which enables data centers to increase power density, reduce their physical footprint, and scale more effectively to handle the power requirements of “AI factories.” In October 2025, ABB announced a partnership with NVIDIA to develop gigawatt-scale data centers to support the demand of future AI workloads, combining MV UPSs with direct current power distribution to the server rooms using solid-state power electronics devices. Other market players are also leveraging MV architectures. For example, in March 2024, Schneider Electric announced it would invest $140 million into its US manufacturing operations to support rising data center demand, with a focus on manufacturing custom electrical switchgear and MV power distribution products.

Vendors are also developing high-capacity UPS units and power distribution systems, with AI-focused facilities expected to increase the “dollar per megawatt” content for electrical system providers. Ultimately, this architectural shift helps ensure the reliability and robustness needed to manage the amplified power-quality issues—such as voltage deviations and transients—introduced by GPU-based model-training clusters.

Examples of AI-ready electrical systems

• Schneider Electric’s high-density solutions: In April 2025, Schneider Electric released its Galaxy VXL UPS, a modular, high-density 3-phase UPS system providing 500–1,250 kW per unit at 400V. Specifically designed for AI-ready data centers and large electrical loads, it features TÜV-verified Live Swap technology, enabling power module replacement while online with no downtime.
• ABB’s AI-ready power architectures: In June 2025, Switzerland-based electrification and automation company ABB launched a UL-market-focused version of its MegaFlex UPS, providing 415V at 60Hz and engineered to support AI workloads in large-scale data centers. A month later, it also introduced its SACE Emax 3 air circuit breaker to assist facilities with surging power needs.

4. New cooling techniques to manage server components and electrical heat

Liquid cooling became essential for performance compliance. The concentration of high-wattage server components has created a critical thermal management challenge. As noted, peak rack density is expected to exceed 1,000 kW by 2029, and this technical escalation, often driven by suppliers like NVIDIA (whose latest chips can drive racks up to 132 kW, with future generations projected to reach 240 kW per rack), makes traditional air-cooling methods (e.g., computer room air conditioners and handlers, chillers, and heat exchangers) insufficient.

Further, maximizing cooling efficiency has moved beyond internal operational needs into environmental mandates. Governmental regulations, such as Germany’s Energy Efficiency Act (EnEfG), require new data centers to operate under strict power usage targets and to reuse waste heat. The capacity for efficient liquid cooling is thus inextricably linked to both performance and sustainability goals.

The industry consensus is that liquid cooling is now essential to meet these goals, with 2 solution types currently being adopted: liquid-to-air and liquid-to-liquid. Liquid-to-liquid cooling solutions have become dominant because they require half the physical space as liquid-to-air cooling.

“I believe, next year, everything will be liquid cooling, and liquid-to-liquid is dominant because liquid-to-air needs more space. Liquid cooling vs. air cooling cuts the space needed in half.”

 Representative from WiWynn at Computex 2025

As vendors focus on boosting heat transfer, space efficiency, and reliability, liquid cooling continues to evolve, and the report explores several emerging innovations in this space.

Example of innovative cooling solutions:

• Johnson Controls’ Silent-Aire CDU liquid-cooling platform. In September 2025, Ireland-based building technology conglomerate Johnson Controls launched the Silent-Aire Coolant Distribution Unit platform, which aims to help high-density data centers transition to liquid cooling as rack power rises. The CDUs span capacities from 500 kW to over 10 MW with flexible designs tailored to different deployment needs. Units can be installed in-row or at the whitespace perimeter and support multiple liquid-cooling configurations and hybrid designs to deliver precise cooling from edge inference clusters to large AI factories. Johnson Controls says its solutions can cut non-IT energy use by more than 50% in most North American data-center hubs, improving overall facility efficiency.

Key takeaways & opinion: The path to a $1 trillion data center infrastructure market

AI investment growth to remain heavily concentrated among hyperscalers. As shared above, AI-driven demand for compute capacity has triggered record levels of CapEx in data center infrastructure, with hyperscalers accounting for most new buildouts. IoT Analytics expects hyperscalers will continue to lead data center spending and account for more than 60% of all data centers by 2030.

AI boom or bubble? Why $1 trillion by 2030 could be both too bullish and too bearish. These companies are accelerating investment cycles, committing tens of billions annually to expand AI-optimized capacity; Microsoft alone dedicated $80B in FY2025. The focus is on scaling GPU-rich facilities capable of supporting industrial-scale large model training and inference. While enterprise and colocation providers are increasing AI-related investment, their share remains small compared to the hyperscaler-led expansion. However, hyperscalers are now deploying nearly all their free cash flow into AI data centers, and questions are emerging over the sustainability of this pace, particularly if shareholders lose confidence in the expected returns or if popular AI applications, such as ChatGPT, fail to convert user enthusiasm into sufficient paying demand to justify the build-out. On the flip side, if agentic AI delivers on its promise to automate workflows at scale, we may look back in 2030 and realize that today’s projections were, in fact, an underestimation of what was required.

AI workloads accelerating liquid-cooling market expansion. The IT infrastructure segment’s high market share, particularly servers’ lion’s share, reflects the direct link between AI adoption and demand for high-performance computing hardware. Cooling systems, while a smaller category in absolute terms, are one of the fastest-growing market segments. The shift towards high-density AI workloads, particularly GPU-based clusters, is driving the adoption of advanced liquid-cooling solutions and accelerating replacement cycles, making cooling a key beneficiary of AI-led infrastructure change.

NVIDIA’s dominant position makes entire ecosystem align with its release cycles. Although technically a component supplier, NVIDIA has become a pivotal force shaping data center buildouts. As the de facto standards for AI training and inference at scale, its GPUs heavily influence facility design, power and cooling requirements, and overall investment priorities. The company’s rapid 1-year release cadence for chips and control over a critical part of the AI compute stack creates a dependency that cascades through the supply chain. As a result, hyperscaler strategies, cooling technology choices, and even energy planning are increasingly aligned to NVIDIA’s product roadmap. Its product roadmap essentially dictates when hyperscalers expand capacity, how facilities are designed, and which cooling technologies are adopted.

How NVIDIA impacts the data center landscape (Insights+)