Hybrid cloud

Hyperscale technology providers supported multibillion-dollar data center buildouts by growing revenues last year, according to Synergy Research Group. The top 20 global cloud and internet service firms saw revenues increase by more than 10% year over year to $2.65 trillion in 2024, according to the report.

Revenues from digital services grew by 13%, led by cloud, which spiked 23% year over year, the analysis found. Generative AI-specific cloud consumption more than doubled in 2024, as the technology sparked an infrastructure building boom among the three largest providers, AWS, Microsoft and Google Cloud.

“Some question the logic behind hyperscale companies jacking up their CapEx numbers recently, and again planning to increase investments substantially in 2025,” John Dinsdale, SRG chief analyst, said in the report. “Once you understand the scale and growth of the digital services that the CapEx is supporting, the logic becomes much clearer.”

Cloud compute consumption was on a steady growth trajectory prior to the emergence of generative AI technologies, as enterprises leveraged infrastructure services to drive modernization. Appetite for AI compounded existing demand last year, triggering a capacity crunch and subsequent buildouts.

Amazon, Microsoft and Google, respectively, poured $79 billion, $76 billion and $53 billion into capital expenditures largely to expand their data center footprints and capacity, Dinsdale said in an email.

The spending aligned with each company’s cloud market share. AWS remained the dominant player, accounting for 30% of the $91 billion enterprises spent on cloud infrastructure services in the last three months of 2024, according to SRG research. Microsoft captured 21% of the market while Google Cloud held onto a 12% share in Q4.

The three providers each signaled their intent to maintain elevated infrastructure spending levels in 2025. Amazon CEO Andy Jassy defended the company’s $100 billion infrastructure investment plan for 2025 in a Thursday letter to shareholders.

Microsoft Vice Chair and President Brad Smith said the company pledged $80 billion to data center buildouts during its current fiscal year in a January blog post. 

Google isn’t far behind. The third largest hyperscaler plans to spend $75 billion to expand AI and cloud capacity in 2025, CEO Sundar Pichai said during a February earnings call.

The spending surge is broadly reflected in hyperscaler financials. The CapEx to revenue ratios rose to more than 12%, from roughly 10% in prior years, Dinsdale said.

The hyperscalers “operate in capital intensive infrastructure markets, where huge ongoing investments are essential in order to be leading players,” said Dinsdale. The investment levels are “eminently reasonable for infrastructure-oriented industries,” he added in an email. “This picture does not show an extravagant level of spending.”

Global spending on data center hardware and software hit an all-time high last year, growing 34% year over year to $282 billion, according to a Synergy Research report.

Hyperscalers accounted for more than half of all infrastructure investments, but CIOs and their surrogates helped drive the surge. Enterprise spending on data center gear increased 21% after five years of marginal growth, thanks largely to generative AI, according to SRG.

“GPUs and generative AI systems lit a fire under the market, resulting in record growth rates for the industry,” SRG Chief Analyst John Dinsdale said. “While the ongoing success of public cloud has been the main driving force behind data center investments for well over a decade now, no-one imagined a 2024 market for data center gear reaching over $280 billion.”

Enterprises leaned hard on cloud to pilot generative AI use cases and jumpstart adoption. But IT chiefs shored up internal systems, too, as large language model technologies spread to hybrid ecosystems and cost concerns rose.

Despite an enterprise spending spike, SRG expects big cloud providers to maintain and expand their dominance in data center capacity.

“It is good to see the enterprise side of the market growing again, though the long-term trend remains,” Dinsdale said in the report. Cloud providers accounted for one-fifth of global spend on data center hardware and software a decade ago. By 2030, SRG expects hyperscalers to consume nearly two-thirds of the market.

Last year, the amount of compute capacity in the hands of hyperscalers eclipsed what was left on-prem. Big cloud will control an estimated 60% of global compute before the end of the decade, according to SRG’s analysis. The remaining 40% will be split evenly between on-prem and colocation data centers.

The trend toward cloud is reflected in SRG’s latest round of projections.

Hyperscale data center capacity will triple by 2030, according to a separate SRG report published Thursday. The average capacity of new cloud data centers is expected to double in the next four years as compute density increases.

“That installed base will continue to grow, but the most marked change in the market is the ever-increasing capacity of data centers that are being brought online,” Dinsdale said.

In late 2024, global cloud and internet service providers had 1,103 data centers operational and nearly 500 additional facilities in the pipeline, by SRG’s count.

The industry’s biggest players kicked off 2025 with major infrastructure investments. 

Since the start of January, AWS announced plans to invest $11 billion in Georgia-based data centers and Microsoft said it will spend $80 billion on AI infrastructure during its current fiscal year.

Sponsored content

By Splunk

In today’s rapidly evolving technological landscape, maintaining security and observability is paramount for ensuring customer trust and operational resilience. Companies are grappling with rapid technological advancements, rising customer expectations, and an interconnected global economy where even minor disruptions can have significant consequences.

$180M is the estimated total cost of downtime annually in the technology industry, according to Splunk’s Hidden Costs of Downtime report.

Observability provides the crucial visibility needed to navigate this complexity, enabling organizations to maintain resilient operations, reduce downtime, and fuel innovation.

Observability: Beyond Troubleshooting

Observability is more than a mere troubleshooting tool; it is a catalyst for innovation. By offering real-time feedback on how changes impact system performance, teams can swiftly identify and resolve issues during feature deployments or updates. This agility encourages faster iteration cycles and experimentation, allowing companies to innovate without fear of unintended consequences. As a result, organizations position themselves as industry leaders, delivering superior customer experiences and staying ahead of competitors.

As observability evolves, it offers deeper insights that drive smarter decision-making and enhance operational resilience. Predictive analytics can forecast potential failures or capacity constraints, enabling proactive measures like redistributing workloads or scaling infrastructure to prevent downtime. Observability tools optimize resource utilization across hybrid cloud environments and enhance threat detection by correlating performance anomalies with potential vulnerabilities. These advancements transform observability into a strategic enabler, allowing businesses to confidently navigate complexity and maintain a competitive edge.

Splunk provides end-to-end visibility across complex, hybrid environments to keep digital systems securely up and running. With our unified security and observability platform, organizations can proactively prevent major issues, remediate threats and disruptions faster, and adapt quickly to new opportunities.

Hybrid Cloud Monitoring: A Vital Use Case

Hybrid cloud environments present unique challenges due to their complexity and fragmented nature. Observability tools play a crucial role in monitoring these environments by providing comprehensive visibility and insights across both on-premises and cloud-based systems. This unified approach enables organizations to optimize resource allocation, ensure seamless performance, and swiftly address issues that may arise. With hybrid cloud monitoring, businesses can effectively manage workloads, scale infrastructure, and maintain operational continuity, ensuring that their digital operations remain resilient and efficient.

Splunk expands visibility to the cloud in minutes with hundreds of out-of-the-box integrations, prebuilt dashboards, automatic detectors and alerts for Kubernetes, containerized environments and other cloud services. Built-in AI/ML capabilities quickly pick up any incidents.

The Future of Observability

AI and AIOps enhance observability by improving real-time monitoring, reducing detection times, and identifying anomalies. This facilitates precise root cause analysis and predictive intelligence, ensuring business continuity. By automating tasks and boosting reliability, AIOps transforms IT into a growth engine, supporting resilience and strategic growth amid hybrid cloud expansion.

As the observability landscape matures, advancements will redefine how enterprises monitor, understand, and optimize their digital operations. Organizations must balance innovation with cost-effectiveness and strategic value. By harnessing the power of observability, hybrid cloud monitoring, and AI, businesses can achieve measurable outcomes, transforming IT into a driving force for innovation and growth.

Splunk’s unified platform enables SecOps, ITOps, and engineering teams to effectively collaborate and achieve their goals, regardless of scale. By providing shared data, language, and tooling, the platform ensures comprehensive visibility across operations, rapid detection and investigation of issues, and optimized response strategies. This integration keeps digital systems secure and reliable, enhancing organizational resilience and paving the way for more robust operations and innovation.

Enterprise cloud bills are rising despite a respite in price increases, according to technology expense-management software provider Tangoe. The company analyzed vendor market trends and customer spending habits for its report. 

“While competition among hyperscalers like AWS, Microsoft Azure and Google Cloud has led to some incremental cost reductions — particularly in storage and compute — these decreases are often offset by enterprises increasing their consumption,” Chris Ortbals, chief product officer at Tangoe, said.

The pause in pricing increases has created an opening for CIOs to lock in lower rates. “Hyperscalers are constantly adjusting pricing models, discount structures and bundled offerings,” Ortbals said. “If an organization signed a three-year agreement in 2022, there’s a good chance that they’re now paying above-market rates for certain services.”

Cloud’s pay-as-you-go billing model led many organizations down a costly path as overprovisioning increased costs. But scalability provided IT leaders with levers to dial back consumption and optimize spend.

As strategies matured, CIOs charted a hybrid modernization approach that continues to gain momentum, according to the Tangoe report. Post-migration sticker shock has put cost on par with other critical workload placement factors like scale and security.

“We’re seeing more enterprises repatriate specific workloads to private cloud or on-premises environments to optimize costs, particularly for predictable workloads,” Ortbals said.

AWS and Microsoft, the two largest hyperscaler by market share, said their public cloud offerings faced on-prem competition in separate filings with the U.K’s Competition and Markets Authority last year.

The shift to a more balanced hybrid cloud strategy can give customers a stronger hand in hyperscaler negotiations. Enterprises that demonstrate they have viable alternatives to public cloud can push for better rates, reserved instance discounts and more flexible contract terms, Ortbals said.

Hyperscaler services nevertheless remain a center of gravity for enterprise IT.

Global spending on cloud increased $60 billion year over year to more than $330 billion in 2024, according to Synergy Research Group. Public cloud infrastructure and platform services accounted for the bulk of the boost, the firm found.

Generative AI has been a major source of hyperscaler growth in the last two years, SRG said. LLM technologies are also driving up the cost of specific cloud services, as the big three cloud providers pour billions into infrastructure upgrades to power AI workloads.

“Specialized services, such as AI and high-performance computing, continue to carry premium pricing,” Ortbals said. “Cloud and SaaS pricing may be coming down, but more usage may raise related expenses overall.”

Digital networks are the plumbing that undergirds IT systems, pumping data to and from the cloud, connecting user workstations and feeding enterprise applications. In complex hybrid ecosystems, networking can also be a costly headache for tech teams.

Most companies have figured out how to get data to the cloud without incurring major fees, according to a Forrester report. Cloud providers are happy to foot the bill for initial migrations and charge customers for usage later. Data transport costs, however, can mount, driving up enterprise cloud spend.

“A lot of companies are surprised by networking costs,” said Andre Kindness, Forrester principal analyst and the author of the report. “Unless you’re 100% in cloud and don’t have a private data center, you’re going to have data moving in and out of the cloud.”

Egress fees — charges levied when enterprises move data out of the public cloud — can be a major pain point when applications aren’t designed with networking in mind.

Chatty applications that generate frequent network requests are the leaky faucets of cloud networking, adding unexpected costs in drips and drabs. AWS, Microsoft and Google Cloud moved to eliminate some egress charges for customers who opt to shutter workloads last year. Minimizing egress charges for data that remains in the public cloud is an ongoing struggle, Kindness said.

Network misconfigurations are an even more pernicious problem, according to a Stacklet cloud usage optimization report published last year. Nearly half of the 315 cloud and FinOps practitioners surveyed by Centiment for the report cited misconfigured networking as a common source of wasted cloud spend.

Weaknesses in cloud network configurations pose cyber risks, too.

Misconfigurations provided the initial access vector for nearly one-third of malicious cloud incursions during the first half of 2024, according to a Google Cloud threat horizons report.

Communication breakdowns

While there are a number of technical factors that contribute to networking misfires, broader organizational issues within IT exacerbate the problem. Networking professionals don’t communicate with developers and cloud engineers regularly enough, according to Kindness.

“Networking teams don’t usually get invited to the cloud party,” Kindness said. “The disconnect drives security issues and operational inefficiencies, which means there’s a large cost behind it.”

There are technical fixes that can improve individual aspects of cloud networking. But the scope of the problem calls for a more holistic solution. CIOs should bring their networking teams into the tent with developers and cloud architects to help optimize application design and bolster hybrid ecosystems, Kindness said.

Most network professionals could also use training in cloud and coding, while developers need to cultivate a better understanding of networking protocols.

“Networking doesn’t speak the same language as software development,” said Kindness. “If your developers don’t understand networking, they’re going to create chatty applications. If your networking team can’t code, they’re not going to understand what developers are doing.”

At a higher level, enterprises simply need to focus more intently on the plumbing by documenting best practices for hybrid cloud networking, Kindness said.

Gartner expects end-user cloud spending to increase 21.5% globally to $723 billion in 2025, according to a report. The market will reach nearly $600 billion this year, the analyst firm said.

Cloud-based application and software services continue to be the largest spend segment, the analysis shows. Spending on SaaS applications will approach $300 billion next year, up from just over $250 billion in 2024. Infrastructure and platform services, the second and third largest segments, are the fastest growing markets with expected year-over-year spending increases of 25% and 22% respectively.

Enterprise AI adoption is helping to drive cloud growth, according to Sid Nag, VP analyst at Gartner. “The use of AI technologies in IT and business operations is unabatedly accelerating the role of cloud computing in supporting business operations and outcomes,” he said in the report.

Cloud infrastructure delivered large language models to the enterprise, and hyperscaler platforms now serve as marketplaces for a growing number of generative AI-powered applications. 

Combined, cloud infrastructure and platform services — a segment Gartner calls CIPS — will comprise the largest category of enterprise cloud spend in 2025, increasing by 24% to over $300 billion, according to the forecast.

“Organizations are choosing CIPS because workloads of today are complex, and organizations are seeking integrated platforms to simplify development, deployment and operations,” Nag said. “Organizations deploying a multicloud adoption model, which is still recording growth, are also driving spend on CIPS.”

As enterprises modernize data operations and adopt cloud-based generative AI tools tailored to industry verticals, hybrid ecosystems that integrate on-premises systems with hyperscaler infrastructure will also expand. Gartner expects 90% of organizations to have hybrid cloud deployments by 2027.

Enterprises are shifting workloads to container solutions to drive generative AI adoption, according to a Vanson Bourne survey of 1,500 IT and platform engineering professionals commissioned by software provider Nutanix.

Nearly 9 in 10 respondents reported that at least some of their organization’s applications are containerized and more than half leverage containers for all applications, the report found.

“We probably had 15% of our data center customers deploying containers in the past,” Lee Caswell, SVP of product and solutions marketing at Nutanix, said. “Now, we’re seeing an early indicator of a wave of containerization coming into on-prem and across hybrid cloud.”

There’s no shortage of hurdles on the path to scaling generative AI. As enterprises confront data privacy concerns, model reliability issues and an uncertain regulatory environment, IT teams are recalibrating infrastructure needs.

More than half of respondents said their organization needs to increase investments in IT to support generative AI workloads and applications. Integrating generative AI workloads with existing infrastructure was the biggest obstacle to moving use cases from development into production, the report found.

“Organizations have reached an inflection point with GenAI implementation and deployment,” Caswell said, pointing to the rise in container adoption. More than two-thirds of respondents — 7 in 10 — plan to containerize generative AI applications.

Gartner is seeing a similar dynamic. The analyst firm expects more than three-quarters of AI deployments to use container technology by 2027, up from 50% currently.

Containers and Kubernetes orchestration solutions provide the scalability, reliability and adaptability needed to support AI workloads, Gartner said in a January report. Containerization is becoming “the de facto standard” in the space, according to the firm.

While containers grew up in public cloud, the technology has transferred well to on-prem environments and private cloud estates. Workload portability across ecosystems coupled with the spread of large language models are dual drivers of the shift.

“Generative AI, by its very nature, is a hybrid cloud workflow,” Caswell said. “You’re going to be developing LLMs in the public cloud, deploying them, tuning them and maybe retraining them in your private data center, and then doing inferencing right at the edge.”