I wrote an article on cloud computing a year ago and at the time the industry forecasts put the global cloud computing market to reach $1 trillion market by 2030 [1]. Now, those estimates have doubled. This overview revisits the cloud computing state with fresh 2025 insights. We’ll examine its market growth, the AI boom’s impact on data centers, the push for high-density AI-ready infrastructure, emerging trends (from power consumption to capacity constraints), and shifts in energy strategy and regulation.

The cloud computing industry’s growth has accelerated beyond expectations, prompting analysts to raise their forecasts. Current projections foresee $2 trillion in cloud revenues by 2030 [2], a milestone originally expected later. Cloud computing is on track to become a multi-trillion-dollar industry within this decade, underscoring its role in the digital economy. For context, Gartner estimates end-user spending on public cloud services will reach $723 billion in 2025 (up from ~$596B in 2024), highlighting enterprises’ reliance on cloud platforms for IT services. Several factors underpin this growth:

1. Cloud adoption is broadening across industries and geographies. Despite early adopters, only “30% of workloads” have moved to the cloud, leaving room for expansion as businesses modernize legacy systems. Companies are investing in digital transformation, migrating on-premises applications to cloud-based infrastructure and software-as-a-service (SaaS) solutions. The pay-as-you-go economics and limitless scalability of cloud providers remain a compelling.

2. AI is boosting cloud demand. Goldman Sachs research finds generative AI could account for 10–15% of cloud spending by 2030 [2] (roughly $200–$300 billion of that $2T market). In 2024–2025, we witnessed an AI arms race, with organizations striving to develop and deploy AI-driven applications on cloud platforms.

Advanced data analytics, machine learning services, and AI-powered SaaS workloads are cloud-intensive, requiring massive computing power, specialized chips, and storage that cloud providers can supply. They also offer pre-trained AI models and AI-as-a-service platforms, lowering the barrier for companies to infuse AI into their products. The rise of generative AI has become a catalyst for cloud growth, expanding the total addressable market.

1. Market composition. All cloud segments (IaaS, PaaS, and SaaS) are benefiting. While SaaS currently holds the largest revenue share, infrastructure and platform services are growing the fastest. By 2030, the cloud market is expected to split as 40% SaaS (~$780B), 30% PaaS (~$600B), and 30% IaaS (~$580B), reflecting the importance of raw compute power and development platforms due to AI and data analytics needs.

The key players remain familiar: Amazon Web Services, Microsoft Azure, and Google Cloud, which account for the majority of market share, complemented by a rich ecosystem of SaaS providers and specialized cloud services. The competitive landscape is evolving: industry-specific cloud offerings (vertical clouds for healthcare, finance, etc.) are emerging, and on-premises IT is adapting via hybrid cloud strategies.

The cloud’s role as the digital backbone of business looks more solid than ever, with growth outpacing broader IT spending.

If cloud computing is the backbone of the digital economy, AI is its new muscle that needs room to grow. The growth of generative AI is driving an expansion in data center investment and cloud infrastructure [3]. Since OpenAI’s ChatGPT launched in late 2022, hyperscale cloud providers and enterprises have raced to scale up the computing power for AI model training and inference. This has led to significant capital expenditures on data centers.

Cloud giants are investing record sums into AI-oriented infrastructure. Major hyperscalers (Amazon, Microsoft, Google, and Meta) are projected to spend $318 billion in capital expenditures this year (2025), half of all market capex growth. This includes outlays for new data center campuses, expanding existing server farms, and equipping them with advanced hardware (like NVIDIA GPUs and custom AI chips).

A consortium led by OpenAI, SoftBank, Oracle, and others announced a proposed “Stargate” project, a $500 billion initiative to build next-gen AI supercomputing data centers (and dedicated energy infrastructure) over the coming years. The first phase envisions a $100B investment in Texas. While not all plans may materialize, they signal the ambition around AI. We are seeing the construction of a new class of AI-purpose-built data centers.

AI-ready infrastructure. What sets these new investments apart is their focus on high-density, AI-optimized infrastructure. Traditional enterprise data centers or cloud regions were designed for general-purpose computing and storage; AI workloads demand concentrated computing power and specialized cooling and networking. An MIT researcher said, “a generative AI training cluster might consume seven or eight times more energy than a typical workload” [4]. This means racks filled with tens of thousands of GPUs or AI accelerators, drawing far more wattage per square foot than a conventional server rack.

These “AI clusters” require advanced cooling techniques (like liquid or immersion cooling) to dissipate heat from densely packed chips. Data center operators call this trend rack densification, cramming more computing power into the same space, pushing power and cooling demands to new levels. Industry research notes that higher power-density racks and liquid cooling are becoming standard in AI-centric facilities. Suppliers of high-capacity power and cooling equipment (like industrial UPS and coolant system vendors like Vertiv) are seeing booming demand, as hyperscalers outfit data halls for 30–40+ kilowatt per rack densities, unprecedented levels a few years ago.

The capital expenditures by hyperscalers reflect this shift. Evercore ISI reports that in Q1 2025, U.S. hyperscale companies spent $81 billion on capex (a 71% year-on-year jump), primarily on AI and cloud infrastructure. They’ve raised their forecast for 2025 hyperscaler capex growth to +44% (up from 38% prior) as companies “continue to increase investments in AI infrastructure.” All major players significantly hiked their capital spending plans to build AI capacity:

• Google’s capex is expected to reach $75B in 2025 (a 43% YoY increase, mainly for servers and data centers).

• Microsoft and Amazon saw capex jump over 50–70% YoY in early 2025, funneling billions into data center expansion.

• Meta doubled its capex and plans to spend $64–$75B in 2025 (up from $60–65B) mainly on AI-related infrastructure.

• Oracle is increasing its data center investments significantly as well.

These investments aren’t just about spending; they herald a new generation of cloud data centers. We’re seeing the rise of “AI superhubs” with specialized designs: heavy power draw, novel cooling, low-latency networking (e.g. high-bandwidth optical interconnects), and enormous scale. Evercore analysts highlight that generative AI is creating “meaningful tailwinds for the broader data center ecosystem”, benefiting cloud operators and suppliers of GPUs, high-speed switches, fiber optics, power gear, and cooling systems needed for AI-ready facilities. An entire supply chain is ramping up to build the infrastructure for AI.

From a cloud services perspective, the payoff is that providers can offer customers the horsepower required for AI innovation, whether via renting GPU-based instances, offering machine learning platforms, or hosting large AI models accessible via API. As AI becomes embedded in finance, healthcare, and consumer apps, demand for these cloud-based AI services is expected to remain strong.

The rapid expansion of cloud infrastructure presents challenges in data center development. Three interrelated trends stand out in 2025:

1. Rising power density and consumption,

2. Strained power supplies and site selection.

3. A tight supply-demand balance for data center capacity (at least in the short term).

New AI-oriented data centers pack unprecedented computing power into each facility. This high-density compute trend means power consumption per data center is soaring. Recent studies show global data center power usage has tripled over the past decade and could triple again by 2028 if trends continue [5]. By 2026, global data center energy consumption is projected to approach ~1,000 TWh, making them the fifth-largest electricity consumer globally if they were a country.

In the U.S., the Department of Energy forecasts an additional 20 GW of data center load by 2030, and that U.S. data centers will consume 6.7%–12% of national power by 2028 (up from about 4.4% in 2023).

To accommodate high densities, operators are implementing new engineering solutions:

• Many hyperscale data centers now use liquid cooling systems (circulating coolant directly to servers or using immersive cooling baths) because traditional air conditioning is inadequate for racks drawing 30–50 kW each.

• AI training rooms often resemble supercomputer installations, with specialized cooling units and backup generators for large loads.

• The electrical infrastructure in modern facilities, like transformers to power distribution units, has been upgraded to handle higher amperages.

All of this increases construction costs and complexity, but is essential to run AI workloads reliably.

The grid is now a critical bottleneck. An industry regulator said, “as these data centers get bigger and consume more energy, the grid can’t withstand the loss of large loads” under current configurations. That comment referred to an incident in “Data Center Alley” in Virginia, where 60 data centers (about 1,500 MW) disconnected from the grid last year, forcing emergency measures by the grid operator [6]. This event, triggered by a voltage fault, highlighted the concentrated and critical power draws of data centers threatening grid stability.

Grid authorities (like NERC in North America) are studying these scenarios and considering new standards to prevent data center clusters from causing blackouts. Data centers have evolved from niche facilities to integral pieces of national infrastructure with corresponding responsibilities.

Supply-demand mismatches and capacity crunch. With high demand, is the data center space supply keeping up? In the short term, barely. The cloud build-out faces constraints like construction lead times, land availability, and power availability. Industry data indicates major data center markets have very tight supply.

A 2024 CBRE survey noted “record tenant demand, historically low supply, and continued positive rental growth” in the data center sector, with rents jumping 20–30% year-over-year in 2023 and many new facilities pre-leased before completion [13]. This is driven by hyperscalers. When a Microsoft or Amazon wants a new availability zone, they can absorb an entire 30 MW data center campus at once.

Smaller cloud and colocation providers are racing to add capacity but face bottlenecks. As a result, vacancy rates for quality data center space in top markets, like Northern Virginia, are at historic lows, and supply is struggling to keep pace with AI-fueled demand.

A critical bottleneck is power grid connections. It doesn’t matter how fast you can erect a server rack building if you can’t get sufficient megawatts from the utility. Across the U.S. and parts of Europe, utilities are reporting high request volumes from data center projects.

• A Reuters survey of U.S. electric utilities found nearly half had received inquiries from data center operators for more power than the utility’s entire current peak load [6].

• Oncor Electric in Texas received requests for an additional 119 GW of load, almost 4× its existing demand. Not all projects will be built, but the scale of requests shows companies’ willingness to secure power for future data centers.

• In Texas’s ERCOT grid, officials noted data center developers filed for 136 GW of new interconnections by mid-2025, up from 41 GW a year earlier [7]. Peak load forecasts in regions like Dallas are being revised for these new AI facilities.

The timing mismatch is an issue: a big data center can be built in less time than a new power plant or substation. This has led to a peak of uncertainty in the power sector. Planners haven’t dealt with demand growing this fast before.

In response, grid operators and governments are acting. Some utilities now require data centers to provide more upfront notice of their power needs, and regions like Northern Virginia have paused new data center approvals in maxed-out grid areas. In Ireland and the Netherlands, moratoriums or strict quotas on new data center construction were implemented to manage electricity constraints and land use concerns.

In 2023, President Biden signed an Executive Order directing federal support to meet the “massive energy needs for fast-growing advanced AI data centers”, recognizing that without coordination, power shortfalls could hinder tech growth. Meanwhile, data center design is adapting: there’s interest in on-site power generation (small natural gas or hydrogen turbines, fuel cells, and modular nuclear reactors). Providers are exploring demand response techniques, like temporarily throttling non-critical workloads or using batteries to smooth out peaks for better grid integration and secure priority electricity access.

Relief is on the horizon: industry projections suggest that by 2026–2027, new capacity will outpace demand growth, easing today’s crunch. Goldman Sachs forecasts the global data center occupancy rate (a measure of demand vs. capacity) will rise from ~85% in 2023 to ~95% by late 2026, then moderate as new builds are completed [5].

A wave of data centers under construction should meet AI-driven demand by 2027, leading to a more balanced market (and potential oversupply risks if AI adoption slows). For the next couple of years, demand is outstripping supply, driving up prices, spurring intense building activity, and challenging operators to innovate around power and cooling.

The cloud computing boom and the resulting data center proliferation have amplified sustainability concerns. Hyperscale cloud providers have touted their progress on green energy and carbon neutrality. Many have made strides in reducing Scope 1 and Scope 2 emissions (direct emissions from operations and purchased electricity). Tech giants like Google and Microsoft have invested heavily in renewable energy for their data centers, aiming for 100% carbon-free power (Google claims its data centers now match 100% renewable energy purchases annually).

They’ve improved energy efficiency through custom server designs and AI-driven cooling optimizations, and experimented with submerging servers underwater or building in colder climates for outside air cooling. These efforts have kept data center energy efficiency improving; many hyperscale facilities operate near a 1.1 PUE (meaning very little extra energy overhead beyond the servers).

However, the scale of cloud growth and AI workloads is raising new sustainability challenges.

A key emerging issue is Scope 3 emissions, the indirect emissions in a company’s supply chain and construction. According to data center engineering firms, “Scope 3…linked to building and supply chain activities, [is] the biggest slice of the carbon pie” for modern high-tech data centers.

• The problem: Even if a cloud data center runs on renewable electricity (zero Scope 2 emissions), the embodied carbon from manufacturing the servers, GPUs, cooling equipment, and constructing the concrete and steel building is substantial.

• The AI factor: AI-ready data centers are material- and carbon-intensive. They require vast quantities of steel and concrete (high carbon footprints) and energy-intensive cutting-edge chips whose fabrication is energy-intensive. An MIT analysis noted that producing a single advanced GPU incurs significantly more emissions than a standard server CPU, due to complex fabrication and emissions from sourcing and transporting raw materials for these chips [4]. Multiplied by tens of thousands of AI chips, plus backup batteries, diesel generators, miles of copper cabling, etc., the embedded carbon accumulates quickly.

• The response: As Manfred Engelhard of Exyte said supporting AI at scale forces us to “confront a hard truth: If AI is to be a force for good, we must address Scope 3 emissions with the same urgency” as direct emissions. Cloud providers are responding, some are using low-carbon concrete and recycled steel in new builds, demanding Environmental Product Declarations from suppliers, and optimizing construction processes to reduce waste.

Measuring and mitigating Scope 3 remains a challenge across the industry.

Another environmental concern is water usage. Large data centers typically rely on water-cooled chillers or cooling towers, which evaporate water to carry away heat. For each kilowatt-hour of energy a data center consumes, roughly 2 liters of water are needed for cooling. Thus, a single big facility can use millions of gallons of water per day, a concerning figure in water-scarce regions.

The surge in AI data centers has magnified this issue; local communities and governments are scrutinizing data center water permits. In response, operators are deploying alternatives like direct-to-chip liquid cooling (which can reduce or eliminate water evaporation) and using reclaimed wastewater instead of potable water for cooling.

In hot climates, avoiding water use without sacrificing efficiency is tough. The sustainability dialogue now includes water alongside carbon: some municipalities require water recycling systems or limit evaporative cooling operating hours on the hottest days.

The electricity powering cloud infrastructure must come from somewhere. Even though tech companies contract huge volumes of wind and solar power, the intermittency of renewables means fossil fuels often fill the gap, especially for 24/7 computing loads. As one researcher stated, “the pace at which companies are building new data centers means the bulk of the electricity must come from fossil fuel plants” (at least in the near term given grid realities) [4].

This reality has spurred cloud firms to explore alternative energy sources. The most notable development is the shift toward nuclear energy to support data center growth.

In recent years, big techs have announced initiatives to incorporate nuclear power (especially next-generation small modular reactors, SMRs) into their future energy mix:

• In October 2024, Amazon Web Services announced plans to invest in SMR projects that could provide over 5,000 MW of clean, continuous power in the 2030s [9].

• Microsoft signed a 20-year deal to buy energy from a revived nuclear plant (the infamous Three Mile Island site) and is working with startups on microreactors [10].

• Google agreed with SMR developer Kairos Power to supply a future data center region with nuclear electricity [9].

The motivation is clear: as cloud and AI loads grow, nuclear energy offers a steady, carbon-free baseload source to complement wind and solar, without hydroelectric limitations. Even AI thought leaders advocate this shift, Meta’s AI chief scientist Yann LeCun predicted “AI data centers will be built next to [nuclear] power plants” producing cheap, gigawatt-scale, low-emission energy continuously. While it will take years for these plans to materialize (and they face regulatory hurdles), it’s notable that Big Tech is directly investing in energy technology, becoming a catalyst for power industry innovation.

Besides nuclear, cloud companies are exploring energy storage and advanced grid solutions. Massive battery installations at data centers could provide backup power (replacing diesel generators) and stabilize the grid by storing excess renewable energy. Some data centers are implementing heat reuse programs, capturing server waste heat to warm nearby buildings or greenhouses to enhance energy efficiency.

New regulations in parts of Europe mandate this: Germany passed laws requiring large data centers to feed waste heat into district heating networks where feasible. This approach can turn data centers into energy contributors for communities, not just consumers.

Finally, we must address the environmental benefits of cloud computing. While the expansion of data centers raises sustainability worries, cloud computing can be more efficient and greener than the alternative. Centralized cloud data centers tend to be more energy-efficient than dispersed enterprise server rooms. They operate at higher utilization and can be optimized at scale.

Cloud-enabled AI and software might help solve environmental challenges. AI optimizes energy grids, designs better materials, and reduces supply chain waste. The hope is that AI and cloud productivity and efficiency gains will outweigh their environmental costs over time. But realizing that net positive outcome will require conscious effort, hence the emphasis on sustainable design, renewable energy, and innovation to decouple digital growth from carbon footprint.

The cloud story doesn't end with massive data centers. Three trends are reshaping cloud computing: the rise of edge infrastructure, evolving government oversight, and shifting investor expectations around AI's promises.

As cloud grows, edge computing momentum pushes computation closer to end-users or devices. The rise of Internet of Things (IoT) devices, autonomous systems, and the need for ultra-low latency services have driven interest in edge data centers. These are smaller facilities or computing clusters located near population centers or on-premises at factories, retail stores, cell tower sites, etc.

The idea is to process data locally (at the “edge”) when milliseconds matter or bandwidth needs to be conserved, then integrate with the central cloud for heavy lifting. In 2024–2025, telecom providers and cloud companies partnered on edge infrastructure. For example, AWS’s Wavelength and Azure’s Edge Zones allow cloud services to run inside telecom networks for 5G applications.

The convergence of edge and cloud is expanding the cloud model beyond centralized regions. As one tech exec observed, “enterprises are prioritizing architectures that bring compute to data,” blending public cloud, private infrastructure, and edge nodes to meet different latency and sovereignty needs [12]. While edge deployments are still nascent, they’re expected to grow with core cloud, not in competition.

Government policy and regulatory oversight increasingly shape the cloud industry. Energy-related actions (like the U.S. executive order and European efficiency laws) are part of the picture. Other regulatory pressures include:

1. Data privacy and sovereignty laws are influencing cloud development. Regulations like Europe’s GDPR and national data residency laws require cloud providers to build local regions or special “sovereign cloud” offerings to keep data within certain borders. This has led to new cloud regions in countries worldwide and partnerships with local providers.

2. Antitrust scrutiny is another factor, as large cloud providers are closely examined regarding market power, prompting discussions of fair pricing, interoperability, and potential regulatory intervention to ensure a competitive ecosystem.

3. There’s an emerging focus on operational resilience: governments worry about systemic risk from a major cloud provider outage or cyberattacks targeting cloud infrastructure. This could lead to guidelines for cloud providers on uptime, incident reporting, and possibly requirements for multi-cloud portability for critical services.

Governments view cloud and AI as strategic assets. We see public sector investment in cloud and AI research, and cloud companies vying for large government contracts (e.g. U.S. DoD’s JWCC cloud procurement). Tensions between nations (such as U.S.-China tech competition) also play a role; export controls on advanced chips affect where and how cloud AI infrastructure can be built or accessed.

The regulatory environment is evolving: facilitating growth (support for infrastructure and R&D) but also constraining it (rules on privacy, competition, and sustainability). Cloud providers in 2025 must navigate a more complex political landscape than in the past.

Consider how investors and the market view the cloud computing boom. In 2023 and 2024, AI excitement surged stock valuations for cloud and chip companies (NVIDIA’s market cap soared, and Big Tech stocks hit new AI-optimism highs). However, as 2025 unfolds, there’s a mix of optimism and caution. Some on Wall Street warn of a potential “capex bubble”, echoing the late 90s telecom boom where over-investment led to a glut.

Questions arise about whether AI spending will yield returns. When hyperscalers reported earnings in late 2024, their stocks underperformed despite solid results, as investors worried that AI investments hadn’t translated into revenue growth [8]. J.P. Morgan analysts observed the market’s enthusiasm for AI has become tempered, oscillating between “excitement… and skepticism” regarding its near-term profitability [11]. Concerns about overbuilding are voiced; Goldman Sachs noted the risk of “long-term market oversupply” of data centers by 2027+ if AI adoption or efficiency improvements reduce demand.

Most analysts remain bullish on cloud and AI in the long run. We are in the early stages of an AI-driven productivity boom that will justify current investments. Kash Rangan of Goldman Sachs noted a high probability that “the generative AI opportunity is real” and that software and platform companies will reinvent themselves to capitalize on it, delivering attractive returns over the next few years. Evercore’s team emphasized the upward revision in hyperscaler capex forecasts and continued strong demand should mitigate fears of a near-term slowdown in the data center space. Even if growth rates fluctuate, the trend of increasing cloud and AI adoption appears intact.

For investors, the cloud sector now spans a broad value chain, including providers (Amazon, Microsoft, Google, etc.), chipmakers, data center REITs, network equipment firms, and energy companies. This diversification means investment opportunities (and risks) are spread out. Data center real estate is in high demand (near zero vacancy and rising rents), though interest rates and construction costs are factors to monitor.

Meanwhile, companies enabling AI infrastructure from semiconductor firms to power and cooling specialists have seen strong momentum. The market will watch how efficiently cloud providers can convert their capex into cash flow (through new AI cloud services, higher pricing for premium workloads, etc.), and how well supply and demand stay in balance. But given that cloud spending as a portion of IT budgets is still growing, and new use cases emerge regularly, the sentiment is that cloud computing remains a solid long-term growth story.

In revisiting the state of cloud computing in 2025, one theme resonates: scale. The scale of the industry (heading to $2 trillion), technological ambition (AI models and global cloud networks), and infrastructure deployed (mega data centers to novel energy plants) are far beyond a decade ago. Cloud computing has evolved from a convenient IT outsourcing model to the foundational fabric of the digital era enabling streaming services, enterprise software, and the latest AI breakthroughs. This evolution brings opportunities for innovation and efficiency across the economy, but also responsibilities and challenges.

Businesses leveraging the cloud today must navigate a complex environment: they have unparalleled computing power, yet must be mindful of costs (enter FinOps for cloud cost management) and sustainability goals. Cloud providers are focusing on R&D, not just in computing tech, but in green energy, edge devices, and more to ensure sustainable and resilient growth. Collaboration is increasing: between industries (finance, healthcare, etc. on cloud), between tech firms (open-source models and cross-cloud integrations), and between public and private sectors (to align on infrastructure and regulations).

In summary, cloud computing in 2025 is at the intersection of transformative currents. Continued innovation in designing, powering, and utilizing cloud infrastructure will determine how far this backbone can flex to support a digital and intelligent world. The journey to the $2 trillion cloud era is underway, impacting industries, communities, and individuals as we embrace the next phase of cloud-enabled transformation.

1. Cloud computing: Present & Future States | Capital One
   https://www.capitalone.com/tech/cloud/cloud-computing-evolution/

2. Cloud revenues poised to reach $2 trillion by 2030 amid AI rollout | Goldman Sachs
   https://www.goldmansachs.com/insights/articles/cloud-revenues-poised-to-reach-2-trillion-by-2030-amid-ai-rollout

3. J.P.Morgan forecasts spending on data centers could boost US GDP by 20 basis points in 2025-26 | Reuters
   https://www.reuters.com/markets/us/jpmorgan-forecasts-spending-data-centers-could-boost-us-gdp-by-20-basis-points-2025-01-16/

4. Explained: Generative AI’s environmental impact | MIT News | Massachusetts Institute of Technology
   https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117

5. to drive 165% increase in data center power demand by 2030 | Goldman Sachs
   https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030

6. Big Tech’s data center boom poses new risk to US grid operators | Reuters
   

I wrote an article on cloud computing a year ago and at the time the industry forecasts put the global cloud computing market to reach $1 trillion market by 2030 [1]. Now, those estimates have doubled. This overview revisits the cloud computing state with fresh 2025 insights. We’ll examine its market growth, the AI boom’s impact on data centers, the push for high-density AI-ready infrastructure, emerging trends (from power consumption to capacity constraints), and shifts in energy strategy and regulation.

The cloud computing industry’s growth has accelerated beyond expectations, prompting analysts to raise their forecasts. Current projections foresee $2 trillion in cloud revenues by 2030 [2], a milestone originally expected later. Cloud computing is on track to become a multi-trillion-dollar industry within this decade, underscoring its role in the digital economy. For context, Gartner estimates end-user spending on public cloud services will reach $723 billion in 2025 (up from ~$596B in 2024), highlighting enterprises’ reliance on cloud platforms for IT services. Several factors underpin this growth:

1. Cloud adoption is broadening across industries and geographies. Despite early adopters, only “30% of workloads” have moved to the cloud, leaving room for expansion as businesses modernize legacy systems. Companies are investing in digital transformation, migrating on-premises applications to cloud-based infrastructure and software-as-a-service (SaaS) solutions. The pay-as-you-go economics and limitless scalability of cloud providers remain a compelling.

2. AI is boosting cloud demand. Goldman Sachs research finds generative AI could account for 10–15% of cloud spending by 2030 [2] (roughly $200–$300 billion of that $2T market). In 2024–2025, we witnessed an AI arms race, with organizations striving to develop and deploy AI-driven applications on cloud platforms.

Advanced data analytics, machine learning services, and AI-powered SaaS workloads are cloud-intensive, requiring massive computing power, specialized chips, and storage that cloud providers can supply. They also offer pre-trained AI models and AI-as-a-service platforms, lowering the barrier for companies to infuse AI into their products. The rise of generative AI has become a catalyst for cloud growth, expanding the total addressable market.

1. Market composition. All cloud segments (IaaS, PaaS, and SaaS) are benefiting. While SaaS currently holds the largest revenue share, infrastructure and platform services are growing the fastest. By 2030, the cloud market is expected to split as 40% SaaS (~$780B), 30% PaaS (~$600B), and 30% IaaS (~$580B), reflecting the importance of raw compute power and development platforms due to AI and data analytics needs.

The key players remain familiar: Amazon Web Services, Microsoft Azure, and Google Cloud, which account for the majority of market share, complemented by a rich ecosystem of SaaS providers and specialized cloud services. The competitive landscape is evolving: industry-specific cloud offerings (vertical clouds for healthcare, finance, etc.) are emerging, and on-premises IT is adapting via hybrid cloud strategies.

The cloud’s role as the digital backbone of business looks more solid than ever, with growth outpacing broader IT spending.

If cloud computing is the backbone of the digital economy, AI is its new muscle that needs room to grow. The growth of generative AI is driving an expansion in data center investment and cloud infrastructure [3]. Since OpenAI’s ChatGPT launched in late 2022, hyperscale cloud providers and enterprises have raced to scale up the computing power for AI model training and inference. This has led to significant capital expenditures on data centers.

Cloud giants are investing record sums into AI-oriented infrastructure. Major hyperscalers (Amazon, Microsoft, Google, and Meta) are projected to spend $318 billion in capital expenditures this year (2025), half of all market capex growth. This includes outlays for new data center campuses, expanding existing server farms, and equipping them with advanced hardware (like NVIDIA GPUs and custom AI chips).

A consortium led by OpenAI, SoftBank, Oracle, and others announced a proposed “Stargate” project, a $500 billion initiative to build next-gen AI supercomputing data centers (and dedicated energy infrastructure) over the coming years. The first phase envisions a $100B investment in Texas. While not all plans may materialize, they signal the ambition around AI. We are seeing the construction of a new class of AI-purpose-built data centers.

AI-ready infrastructure. What sets these new investments apart is their focus on high-density, AI-optimized infrastructure. Traditional enterprise data centers or cloud regions were designed for general-purpose computing and storage; AI workloads demand concentrated computing power and specialized cooling and networking. An MIT researcher said, “a generative AI training cluster might consume seven or eight times more energy than a typical workload” [4]. This means racks filled with tens of thousands of GPUs or AI accelerators, drawing far more wattage per square foot than a conventional server rack.

These “AI clusters” require advanced cooling techniques (like liquid or immersion cooling) to dissipate heat from densely packed chips. Data center operators call this trend rack densification, cramming more computing power into the same space, pushing power and cooling demands to new levels. Industry research notes that higher power-density racks and liquid cooling are becoming standard in AI-centric facilities. Suppliers of high-capacity power and cooling equipment (like industrial UPS and coolant system vendors like Vertiv) are seeing booming demand, as hyperscalers outfit data halls for 30–40+ kilowatt per rack densities, unprecedented levels a few years ago.

The capital expenditures by hyperscalers reflect this shift. Evercore ISI reports that in Q1 2025, U.S. hyperscale companies spent $81 billion on capex (a 71% year-on-year jump), primarily on AI and cloud infrastructure. They’ve raised their forecast for 2025 hyperscaler capex growth to +44% (up from 38% prior) as companies “continue to increase investments in AI infrastructure.” All major players significantly hiked their capital spending plans to build AI capacity:

• Google’s capex is expected to reach $75B in 2025 (a 43% YoY increase, mainly for servers and data centers).

• Microsoft and Amazon saw capex jump over 50–70% YoY in early 2025, funneling billions into data center expansion.

• Meta doubled its capex and plans to spend $64–$75B in 2025 (up from $60–65B) mainly on AI-related infrastructure.

• Oracle is increasing its data center investments significantly as well.

These investments aren’t just about spending; they herald a new generation of cloud data centers. We’re seeing the rise of “AI superhubs” with specialized designs: heavy power draw, novel cooling, low-latency networking (e.g. high-bandwidth optical interconnects), and enormous scale. Evercore analysts highlight that generative AI is creating “meaningful tailwinds for the broader data center ecosystem”, benefiting cloud operators and suppliers of GPUs, high-speed switches, fiber optics, power gear, and cooling systems needed for AI-ready facilities. An entire supply chain is ramping up to build the infrastructure for AI.

From a cloud services perspective, the payoff is that providers can offer customers the horsepower required for AI innovation, whether via renting GPU-based instances, offering machine learning platforms, or hosting large AI models accessible via API. As AI becomes embedded in finance, healthcare, and consumer apps, demand for these cloud-based AI services is expected to remain strong.

The rapid expansion of cloud infrastructure presents challenges in data center development. Three interrelated trends stand out in 2025:

1. Rising power density and consumption,

2. Strained power supplies and site selection.

3. A tight supply-demand balance for data center capacity (at least in the short term).

New AI-oriented data centers pack unprecedented computing power into each facility. This high-density compute trend means power consumption per data center is soaring. Recent studies show global data center power usage has tripled over the past decade and could triple again by 2028 if trends continue [5]. By 2026, global data center energy consumption is projected to approach ~1,000 TWh, making them the fifth-largest electricity consumer globally if they were a country.

In the U.S., the Department of Energy forecasts an additional 20 GW of data center load by 2030, and that U.S. data centers will consume 6.7%–12% of national power by 2028 (up from about 4.4% in 2023).

To accommodate high densities, operators are implementing new engineering solutions:

• Many hyperscale data centers now use liquid cooling systems (circulating coolant directly to servers or using immersive cooling baths) because traditional air conditioning is inadequate for racks drawing 30–50 kW each.

• AI training rooms often resemble supercomputer installations, with specialized cooling units and backup generators for large loads.

• The electrical infrastructure in modern facilities, like transformers to power distribution units, has been upgraded to handle higher amperages.

All of this increases construction costs and complexity, but is essential to run AI workloads reliably.

The grid is now a critical bottleneck. An industry regulator said, “as these data centers get bigger and consume more energy, the grid can’t withstand the loss of large loads” under current configurations. That comment referred to an incident in “Data Center Alley” in Virginia, where 60 data centers (about 1,500 MW) disconnected from the grid last year, forcing emergency measures by the grid operator [6]. This event, triggered by a voltage fault, highlighted the concentrated and critical power draws of data centers threatening grid stability.

Grid authorities (like NERC in North America) are studying these scenarios and considering new standards to prevent data center clusters from causing blackouts. Data centers have evolved from niche facilities to integral pieces of national infrastructure with corresponding responsibilities.

Supply-demand mismatches and capacity crunch. With high demand, is the data center space supply keeping up? In the short term, barely. The cloud build-out faces constraints like construction lead times, land availability, and power availability. Industry data indicates major data center markets have very tight supply.

A 2024 CBRE survey noted “record tenant demand, historically low supply, and continued positive rental growth” in the data center sector, with rents jumping 20–30% year-over-year in 2023 and many new facilities pre-leased before completion [13]. This is driven by hyperscalers. When a Microsoft or Amazon wants a new availability zone, they can absorb an entire 30 MW data center campus at once.

Smaller cloud and colocation providers are racing to add capacity but face bottlenecks. As a result, vacancy rates for quality data center space in top markets, like Northern Virginia, are at historic lows, and supply is struggling to keep pace with AI-fueled demand.

A critical bottleneck is power grid connections. It doesn’t matter how fast you can erect a server rack building if you can’t get sufficient megawatts from the utility. Across the U.S. and parts of Europe, utilities are reporting high request volumes from data center projects.

• A Reuters survey of U.S. electric utilities found nearly half had received inquiries from data center operators for more power than the utility’s entire current peak load [6].

• Oncor Electric in Texas received requests for an additional 119 GW of load, almost 4× its existing demand. Not all projects will be built, but the scale of requests shows companies’ willingness to secure power for future data centers.

• In Texas’s ERCOT grid, officials noted data center developers filed for 136 GW of new interconnections by mid-2025, up from 41 GW a year earlier [7]. Peak load forecasts in regions like Dallas are being revised for these new AI facilities.

The timing mismatch is an issue: a big data center can be built in less time than a new power plant or substation. This has led to a peak of uncertainty in the power sector. Planners haven’t dealt with demand growing this fast before.

In response, grid operators and governments are acting. Some utilities now require data centers to provide more upfront notice of their power needs, and regions like Northern Virginia have paused new data center approvals in maxed-out grid areas. In Ireland and the Netherlands, moratoriums or strict quotas on new data center construction were implemented to manage electricity constraints and land use concerns.

In 2023, President Biden signed an Executive Order directing federal support to meet the “massive energy needs for fast-growing advanced AI data centers”, recognizing that without coordination, power shortfalls could hinder tech growth. Meanwhile, data center design is adapting: there’s interest in on-site power generation (small natural gas or hydrogen turbines, fuel cells, and modular nuclear reactors). Providers are exploring demand response techniques, like temporarily throttling non-critical workloads or using batteries to smooth out peaks for better grid integration and secure priority electricity access.

Relief is on the horizon: industry projections suggest that by 2026–2027, new capacity will outpace demand growth, easing today’s crunch. Goldman Sachs forecasts the global data center occupancy rate (a measure of demand vs. capacity) will rise from ~85% in 2023 to ~95% by late 2026, then moderate as new builds are completed [5].

A wave of data centers under construction should meet AI-driven demand by 2027, leading to a more balanced market (and potential oversupply risks if AI adoption slows). For the next couple of years, demand is outstripping supply, driving up prices, spurring intense building activity, and challenging operators to innovate around power and cooling.

The cloud computing boom and the resulting data center proliferation have amplified sustainability concerns. Hyperscale cloud providers have touted their progress on green energy and carbon neutrality. Many have made strides in reducing Scope 1 and Scope 2 emissions (direct emissions from operations and purchased electricity). Tech giants like Google and Microsoft have invested heavily in renewable energy for their data centers, aiming for 100% carbon-free power (Google claims its data centers now match 100% renewable energy purchases annually).

They’ve improved energy efficiency through custom server designs and AI-driven cooling optimizations, and experimented with submerging servers underwater or building in colder climates for outside air cooling. These efforts have kept data center energy efficiency improving; many hyperscale facilities operate near a 1.1 PUE (meaning very little extra energy overhead beyond the servers).

However, the scale of cloud growth and AI workloads is raising new sustainability challenges.

A key emerging issue is Scope 3 emissions, the indirect emissions in a company’s supply chain and construction. According to data center engineering firms, “Scope 3…linked to building and supply chain activities, [is] the biggest slice of the carbon pie” for modern high-tech data centers.

• The problem: Even if a cloud data center runs on renewable electricity (zero Scope 2 emissions), the embodied carbon from manufacturing the servers, GPUs, cooling equipment, and constructing the concrete and steel building is substantial.

• The AI factor: AI-ready data centers are material- and carbon-intensive. They require vast quantities of steel and concrete (high carbon footprints) and energy-intensive cutting-edge chips whose fabrication is energy-intensive. An MIT analysis noted that producing a single advanced GPU incurs significantly more emissions than a standard server CPU, due to complex fabrication and emissions from sourcing and transporting raw materials for these chips [4]. Multiplied by tens of thousands of AI chips, plus backup batteries, diesel generators, miles of copper cabling, etc., the embedded carbon accumulates quickly.

• The response: As Manfred Engelhard of Exyte said supporting AI at scale forces us to “confront a hard truth: If AI is to be a force for good, we must address Scope 3 emissions with the same urgency” as direct emissions. Cloud providers are responding, some are using low-carbon concrete and recycled steel in new builds, demanding Environmental Product Declarations from suppliers, and optimizing construction processes to reduce waste.

Measuring and mitigating Scope 3 remains a challenge across the industry.

Another environmental concern is water usage. Large data centers typically rely on water-cooled chillers or cooling towers, which evaporate water to carry away heat. For each kilowatt-hour of energy a data center consumes, roughly 2 liters of water are needed for cooling. Thus, a single big facility can use millions of gallons of water per day, a concerning figure in water-scarce regions.

The surge in AI data centers has magnified this issue; local communities and governments are scrutinizing data center water permits. In response, operators are deploying alternatives like direct-to-chip liquid cooling (which can reduce or eliminate water evaporation) and using reclaimed wastewater instead of potable water for cooling.

In hot climates, avoiding water use without sacrificing efficiency is tough. The sustainability dialogue now includes water alongside carbon: some municipalities require water recycling systems or limit evaporative cooling operating hours on the hottest days.

The electricity powering cloud infrastructure must come from somewhere. Even though tech companies contract huge volumes of wind and solar power, the intermittency of renewables means fossil fuels often fill the gap, especially for 24/7 computing loads. As one researcher stated, “the pace at which companies are building new data centers means the bulk of the electricity must come from fossil fuel plants” (at least in the near term given grid realities) [4].

This reality has spurred cloud firms to explore alternative energy sources. The most notable development is the shift toward nuclear energy to support data center growth.

In recent years, big techs have announced initiatives to incorporate nuclear power (especially next-generation small modular reactors, SMRs) into their future energy mix:

• In October 2024, Amazon Web Services announced plans to invest in SMR projects that could provide over 5,000 MW of clean, continuous power in the 2030s [9].

• Microsoft signed a 20-year deal to buy energy from a revived nuclear plant (the infamous Three Mile Island site) and is working with startups on microreactors [10].

• Google agreed with SMR developer Kairos Power to supply a future data center region with nuclear electricity [9].

The motivation is clear: as cloud and AI loads grow, nuclear energy offers a steady, carbon-free baseload source to complement wind and solar, without hydroelectric limitations. Even AI thought leaders advocate this shift, Meta’s AI chief scientist Yann LeCun predicted “AI data centers will be built next to [nuclear] power plants” producing cheap, gigawatt-scale, low-emission energy continuously. While it will take years for these plans to materialize (and they face regulatory hurdles), it’s notable that Big Tech is directly investing in energy technology, becoming a catalyst for power industry innovation.

Besides nuclear, cloud companies are exploring energy storage and advanced grid solutions. Massive battery installations at data centers could provide backup power (replacing diesel generators) and stabilize the grid by storing excess renewable energy. Some data centers are implementing heat reuse programs, capturing server waste heat to warm nearby buildings or greenhouses to enhance energy efficiency.

New regulations in parts of Europe mandate this: Germany passed laws requiring large data centers to feed waste heat into district heating networks where feasible. This approach can turn data centers into energy contributors for communities, not just consumers.

Finally, we must address the environmental benefits of cloud computing. While the expansion of data centers raises sustainability worries, cloud computing can be more efficient and greener than the alternative. Centralized cloud data centers tend to be more energy-efficient than dispersed enterprise server rooms. They operate at higher utilization and can be optimized at scale.

Cloud-enabled AI and software might help solve environmental challenges. AI optimizes energy grids, designs better materials, and reduces supply chain waste. The hope is that AI and cloud productivity and efficiency gains will outweigh their environmental costs over time. But realizing that net positive outcome will require conscious effort, hence the emphasis on sustainable design, renewable energy, and innovation to decouple digital growth from carbon footprint.

The cloud story doesn't end with massive data centers. Three trends are reshaping cloud computing: the rise of edge infrastructure, evolving government oversight, and shifting investor expectations around AI's promises.

As cloud grows, edge computing momentum pushes computation closer to end-users or devices. The rise of Internet of Things (IoT) devices, autonomous systems, and the need for ultra-low latency services have driven interest in edge data centers. These are smaller facilities or computing clusters located near population centers or on-premises at factories, retail stores, cell tower sites, etc.

The idea is to process data locally (at the “edge”) when milliseconds matter or bandwidth needs to be conserved, then integrate with the central cloud for heavy lifting. In 2024–2025, telecom providers and cloud companies partnered on edge infrastructure. For example, AWS’s Wavelength and Azure’s Edge Zones allow cloud services to run inside telecom networks for 5G applications.

The convergence of edge and cloud is expanding the cloud model beyond centralized regions. As one tech exec observed, “enterprises are prioritizing architectures that bring compute to data,” blending public cloud, private infrastructure, and edge nodes to meet different latency and sovereignty needs [12]. While edge deployments are still nascent, they’re expected to grow with core cloud, not in competition.

Government policy and regulatory oversight increasingly shape the cloud industry. Energy-related actions (like the U.S. executive order and European efficiency laws) are part of the picture. Other regulatory pressures include:

1. Data privacy and sovereignty laws are influencing cloud development. Regulations like Europe’s GDPR and national data residency laws require cloud providers to build local regions or special “sovereign cloud” offerings to keep data within certain borders. This has led to new cloud regions in countries worldwide and partnerships with local providers.

2. Antitrust scrutiny is another factor, as large cloud providers are closely examined regarding market power, prompting discussions of fair pricing, interoperability, and potential regulatory intervention to ensure a competitive ecosystem.

3. There’s an emerging focus on operational resilience: governments worry about systemic risk from a major cloud provider outage or cyberattacks targeting cloud infrastructure. This could lead to guidelines for cloud providers on uptime, incident reporting, and possibly requirements for multi-cloud portability for critical services.

Governments view cloud and AI as strategic assets. We see public sector investment in cloud and AI research, and cloud companies vying for large government contracts (e.g. U.S. DoD’s JWCC cloud procurement). Tensions between nations (such as U.S.-China tech competition) also play a role; export controls on advanced chips affect where and how cloud AI infrastructure can be built or accessed.

The regulatory environment is evolving: facilitating growth (support for infrastructure and R&D) but also constraining it (rules on privacy, competition, and sustainability). Cloud providers in 2025 must navigate a more complex political landscape than in the past.

Consider how investors and the market view the cloud computing boom. In 2023 and 2024, AI excitement surged stock valuations for cloud and chip companies (NVIDIA’s market cap soared, and Big Tech stocks hit new AI-optimism highs). However, as 2025 unfolds, there’s a mix of optimism and caution. Some on Wall Street warn of a potential “capex bubble”, echoing the late 90s telecom boom where over-investment led to a glut.

Questions arise about whether AI spending will yield returns. When hyperscalers reported earnings in late 2024, their stocks underperformed despite solid results, as investors worried that AI investments hadn’t translated into revenue growth [8]. J.P. Morgan analysts observed the market’s enthusiasm for AI has become tempered, oscillating between “excitement… and skepticism” regarding its near-term profitability [11]. Concerns about overbuilding are voiced; Goldman Sachs noted the risk of “long-term market oversupply” of data centers by 2027+ if AI adoption or efficiency improvements reduce demand.

Most analysts remain bullish on cloud and AI in the long run. We are in the early stages of an AI-driven productivity boom that will justify current investments. Kash Rangan of Goldman Sachs noted a high probability that “the generative AI opportunity is real” and that software and platform companies will reinvent themselves to capitalize on it, delivering attractive returns over the next few years. Evercore’s team emphasized the upward revision in hyperscaler capex forecasts and continued strong demand should mitigate fears of a near-term slowdown in the data center space. Even if growth rates fluctuate, the trend of increasing cloud and AI adoption appears intact.

For investors, the cloud sector now spans a broad value chain, including providers (Amazon, Microsoft, Google, etc.), chipmakers, data center REITs, network equipment firms, and energy companies. This diversification means investment opportunities (and risks) are spread out. Data center real estate is in high demand (near zero vacancy and rising rents), though interest rates and construction costs are factors to monitor.

Meanwhile, companies enabling AI infrastructure from semiconductor firms to power and cooling specialists have seen strong momentum. The market will watch how efficiently cloud providers can convert their capex into cash flow (through new AI cloud services, higher pricing for premium workloads, etc.), and how well supply and demand stay in balance. But given that cloud spending as a portion of IT budgets is still growing, and new use cases emerge regularly, the sentiment is that cloud computing remains a solid long-term growth story.

In revisiting the state of cloud computing in 2025, one theme resonates: scale. The scale of the industry (heading to $2 trillion), technological ambition (AI models and global cloud networks), and infrastructure deployed (mega data centers to novel energy plants) are far beyond a decade ago. Cloud computing has evolved from a convenient IT outsourcing model to the foundational fabric of the digital era enabling streaming services, enterprise software, and the latest AI breakthroughs. This evolution brings opportunities for innovation and efficiency across the economy, but also responsibilities and challenges.

Businesses leveraging the cloud today must navigate a complex environment: they have unparalleled computing power, yet must be mindful of costs (enter FinOps for cloud cost management) and sustainability goals. Cloud providers are focusing on R&D, not just in computing tech, but in green energy, edge devices, and more to ensure sustainable and resilient growth. Collaboration is increasing: between industries (finance, healthcare, etc. on cloud), between tech firms (open-source models and cross-cloud integrations), and between public and private sectors (to align on infrastructure and regulations).

In summary, cloud computing in 2025 is at the intersection of transformative currents. Continued innovation in designing, powering, and utilizing cloud infrastructure will determine how far this backbone can flex to support a digital and intelligent world. The journey to the $2 trillion cloud era is underway, impacting industries, communities, and individuals as we embrace the next phase of cloud-enabled transformation.

1. Cloud computing: Present & Future States | Capital One
   https://www.capitalone.com/tech/cloud/cloud-computing-evolution/

2. Cloud revenues poised to reach $2 trillion by 2030 amid AI rollout | Goldman Sachs
   https://www.goldmansachs.com/insights/articles/cloud-revenues-poised-to-reach-2-trillion-by-2030-amid-ai-rollout

3. J.P.Morgan forecasts spending on data centers could boost US GDP by 20 basis points in 2025-26 | Reuters
   https://www.reuters.com/markets/us/jpmorgan-forecasts-spending-data-centers-could-boost-us-gdp-by-20-basis-points-2025-01-16/

4. Explained: Generative AI’s environmental impact | MIT News | Massachusetts Institute of Technology
   https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117

5. to drive 165% increase in data center power demand by 2030 | Goldman Sachs
   https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030

6. Big Tech’s data center boom poses new risk to US grid operators | Reuters