Hybrid cloud

Hyperscalers including Google, Microsoft and AWS will account for 67% of data center capacity by 2031, driven by significant AI infrastructure investments, according to a report from Synergy Research Group. Enterprise on-prem data centers will drop from 56% of total data center capacity in 2018 to 19% by 2031, the firm said.

“Almost 60% of that hyperscale capacity is now in own-built, owned data centers with the balance being in leased facilities,” according to the report. “With non-hyperscale colocation capacity accounting for another 20% of total capacity, that leaves on-premise data centers with just 32% of the total.” 

On-prem data center’s share of total capacity remained relatively constant over the last eight years, even seeing a boost from enterprise AI deployments and GPU servers, according to the report. However, by 2031, hyperscalers will have 14 times the amount of data center capacity than 2018 as AI accelerates expansion.  

As enterprises rush to adopt AI, hyperscalers are building data centers to keep up with demand and navigate capacity crunches while facing rising concerns about electricity costs and community pushback. 

The top three hyperscalers, Microsoft, Google and AWS, plan to invest more than $500 billion in capital expenditures for infrastructure supporting AI deployment in fiscal year 2026. Hyperscalers operated 1,297 large data centers globally in Q3 2025, nearly triple the amount from 2018, according to a separate SRG report. That number rose to 1,360 by the end of 2025.

“We will increase our total AI capacity by over 80% this year and roughly double our total data center footprint over the next two years, reflecting the demand signals we see,” Microsoft CEO Satya Nadella said during the company’s Q1 2026 earnings call in October. 

It’s not just hyperscalers vying for compute power. Large language model provider Anthropic earlier this week announced an expanded partnership with Google to secure its TPU chips, supplied by Broadcom, adding multiple gigawatts of compute capacity starting in 2027. Meanwhile, LLM provider OpenAI in October partnered with AMD to add six gigawatts of AI compute capacity. 

Despite massive investments to expand compute capacity, the tech industry is already facing shortages. During Microsoft's Q1 2026 earnings call, EVP and CFO Amy Hood said Microsoft was struggling with a capacity shortage expected to extend throughout the fiscal year. 

Naveen Chhabra, principal analyst at Forrester, previously told CIO Dive in an email that CIOs will need to prepare for compute capacity being a constrained resource. 

In addition to capacity shortages, hyperscalers are grappling with concerns about how data centers will raise electricity costs. The U.S. Energy Information Administration said data center electricity demand could drive price hikes of up to 79% in areas like Texas by 2027. 

The concern has not gone unnoticed. OpenAI in January committed to paying its own way on energy across its Stargate projects, a $500 billion AI infrastructure push announced by President Donald Trump in January 2025.

Last month, the White House announced the Ratepayer Protection Pledge, a voluntary commitment from tech providers including Google, Oracle, xAI, Meta, Microsoft, OpenAI and Amazon  to help offset electricity costs as data center buildouts continue. 

The tech industry is also facing pushback on data center builds from local and state governments. A city in Wisconsin recently passed an anti-data center referendum, according to Politico reporting. Meanwhile, the Wall Street Journal reported that Maine could freeze data center construction until November 2027.

Soaring automation demand will triple AI infrastructure budgets for most enterprises, increasing investment and ownership in the physical infrastructure needed to power AI by 2028, according to a Deloitte report. The company surveyed 515 U.S. enterprise companies in December for the report.

Almost half of respondents said they have more than 30 AI pilots in the works. By 2028, Deloitte projects that nearly 70% of companies will be running that many AI proofs of concept.

Respondents said they use a mix of closed, open, derivative and packaged SaaS models in their AI toolkit, with closed models leading slightly. There’s no clear direction on which model or mix of models will win out in the next few years, the report found.

As AI demand accelerates, companies are working to build a mixed portfolio of AI architecture. The recipe includes hyperscaler models and services, and a combination of public cloud with their own on-premises infrastructure.

Demand for AI infrastructure is universal — the neocloud market is set to reach $400 billion in revenues by 2031, as tech giants such as Meta strike deals to gain more compute. This week, Anthropic signed new agreements with Google and Broadcom to add multiple gigawatts of TPU capacity starting in 2027 as customer demand accelerates; the AI model maker also signed a multi-year agreement with CoreWeave for compute access.

Providing their own infrastructure is the only way some enterprises can scale AI at their desired pace in a crowded resource marketplace, Chris Thomas, U.S. hybrid cloud infastructure leader at Deloitte, told CIO Dive. 

“That's nearly doubling across the board in three years, and it's largely driven by token volumes that are doubling and tripling,” he said. “They are creating the workloads that the public cloud alone can’t serve cost effectively at this point in time.”

The investment into these projects is significant, with some enterprises projecting they’ll spend almost four times more on AI infrastructure by 2028. It’s a shift from traditional IT spending, which was planned for one-off modernization efforts. IT departments are shifting to more sustained, year-over-year spending, the report said. 

CIOs are also likely having to work more closely with other financial decision-makers in their enterprises to show the value in infrastructure spending, Thomas said. Companies will have to make strategic choices to prioritize AI spending. Some might choose to shift operating expenses to capital expenditures by owning infrastructure instead of renting cloud capacity. 

“The line is blurring between business and tech, no matter what industry you're in,” Thomas said. “Technology decisions have to be made together with the business. And given the size of these budgets that are required, you almost have no choice.”

Sponsored content

By Nerdio

As organizations embrace cloud computing, migrating virtual desktop infrastructure (VDI) to the cloud has become increasingly compelling. But for many businesses, a full public cloud migration isn't the right move — at least not yet. A hybrid cloud approach, which keeps portions of VDI on-premises while integrating with public cloud services, offers the flexibility to transition workloads gradually while balancing cost, control, and performance.

Here are the critical steps to ensure a smooth hybrid VDI migration.

1. Evaluate Your Cloud Readiness First

Before making any architecture decisions, assess where you actually stand. Start by auditing your current on-premises VDI environment — hardware age, capacity, and application dependencies all factor into how much of your infrastructure is ready to move.

Key areas to evaluate:

• Application compatibility: Identify legacy apps or OS instances that may not virtualize cleanly in a cloud environment.
• Network readiness: Ensure you have adequate internet bandwidth for cloud desktop access. For hybrid setups, consider migrating to a software-defined wide area network (SD-WAN) to optimize performance across on-premises and cloud segments. [TS1] Also, ensure network latency between your local and remote services is minimized. Some workloads can tolerate high latency, but some are extremely latency sensitive. Identify your latency requirements and plan accordingly.
• Security and compliance: Review data sovereignty regulations and confirm your cloud provider's certifications align with your requirements (e.g., HIPAA, GDPR).
• Team skills: Determine whether your IT staff has the expertise to manage a hybrid environment or whether training is needed.

This assessment will surface roadblocks before they become costly mid-migration surprises.

2. Define a Strategy That Goes Beyond Cost Savings

Many organizations pursue cloud migration for straightforward reasons — avoiding a hardware refresh, cutting costs, or responding to contract renewals. These are valid motivations, but they can lead to short-sighted decisions.

In a hybrid model especially, it's important to define what you want the cloud to enable long-term. The most durable hybrid strategies typically focus on:

• Cloud bursting: Using public cloud capacity to handle demand spikes while keeping baseline workloads on-premises.
• Business continuity: Leveraging geo-redundant cloud infrastructure and automated failover to complement on-premises resilience.
• Gradual modernization: Migrating applications and desktop images incrementally, using the hybrid phase as a transition period rather than a permanent state.

Take time to map out your application portfolio and identify which workloads are best suited for the cloud now versus later. This roadmap is the foundation of a well-executed hybrid strategy.

3. Architect for Operational Excellence in a Hybrid Environment

Hybrid VDI introduces management complexity — you're now running desktops across two or more environments simultaneously. Designing for operational efficiency from day one is critical.

Practical tips:

• Centralize management: Use cloud-native tools that provide unified visibility across both your on-premises and cloud VDI estate. Avoid relying on separate, siloed monitoring tools for each environment.
• Automate scaling: Implement auto-scaling policies so cloud capacity expands and contracts based on real-time demand, reducing manual intervention.
• Use infrastructure as code: Automate VDI deployments, patching, and configuration management across both environments using policy-driven pipelines. This reduces drift between your on-premises and cloud configurations.
• Break down silos: In a hybrid model, endpoint, cloud, and app teams must work in a coordinated way. Establish clear Service Level Objectives (SLOs) and escalation paths that apply across the full environment.

4. Implement Zero-Trust Security Across Both Environments

Security in a hybrid VDI environment requires consistent enforcement regardless of where workloads run. A zero-trust framework — where no user or device is trusted by default — is the right foundation.

Focus on:

• Identity and access management: Centralize authentication via cloud identity providers and enforce multi-factor authentication for all virtual desktop access, whether users connect to on-premises or cloud desktops.
• Data protection: Encrypt data at rest and in transit using cloud-native tools, and ensure data residency policies are respected across both environments.
• Continuous monitoring: Implement centralized logging and user behavior analytics across your hybrid estate to detect anomalies regardless of where they originate.

5. Manage Costs Proactively

Hybrid environments can be cost-effective, but only with disciplined governance. Without proper controls, cloud consumption can quickly exceed budget.

Effective tactics include: using reserved instances for predictable workloads while leveraging on-demand pricing for bursting, enabling auto-suspend and shutdown policies for idle cloud desktops, and using cost allocation tools to track spending by workload, department, or host pool. This level of visibility creates accountability and makes it easier to demonstrate ROI.

The Bottom Line

A hybrid cloud VDI approach offers a pragmatic path forward — it lets organizations capture cloud benefits without abandoning existing investments overnight. The key is treating the hybrid phase as a deliberate strategy, not just a waypoint. With the right readiness assessment, architecture decisions, and operational discipline, a hybrid VDI environment can deliver scalability, resilience, and cost control for years to come.

Are you looking for ways to smoothly transition and reduce risk in upcoming cloud migration projects? Learn how Nerdio Manager for Enterprise can help.

Despite lengthy cloud adoption efforts, just 14% of enterprises have reached the highest level of cloud maturity, according to a survey published by NTT Data. The technology services company surveyed 2,300 senior decision-makers in 33 countries for its report.

With most businesses still in the earliest stages of cloud maturity, where cloud is used for isolated workloads or infrastructure hosting, fewer than half of surveyed leaders said they were fully satisfied with cloud's role in innovation efforts, according to the survey.

The disconnect between cloud readiness and AI deployment plans is leading most businesses to rethink their cloud spend. Three-quarters of companies expect to increase their cloud spending significantly in the next two years, NTT Data found.

Cloud is key to scaling AI adoption, as quality results from advanced models require the right infrastructure and data. But most enterprises have been unable to fully modernize legacy cloud estates, preventing rich datasets from delivering value. 

In an effort to attract enterprise customers, hyperscalers have expanded their menu of AI offerings and pitched themselves as the natural providers of AI services. 

The focus on enterprise spend is already showing up on their balance sheets. The market for cloud infrastructure services — led by AWS, Microsoft and Google — reached $419 billion last year, according to Synergy Research Group data. The analyst firm credits generative AI services as a major driver for growth.

“Cloud has moved well beyond infrastructure and is now the execution layer for AI,” Charlie Li, president and global head of cloud and security at NTT DATA, said in a release accompanying the report. “Organizations that fail to evolve their cloud foundations risk constraining the growth and value of their AI investments.”

But cloud immaturity is not the only obstacle to enterprise AI. Nearly half of cloud leaders surveyed by NTT Data listed a lack of AI skills as a key blocker to hindering their cloud strategies over the coming year. The skills gaps worsen another problem in cloud: stubborn legacy implementations.

“Many organizations manage complex application estates that are difficult to refactor, integrate or retire,” NTT Data said in its report. “Skills gaps in cloud-native development, automation and DevOps compound the challenge, slowing progress and increasing execution risk as AI places new demands on application architecture.”

Hybrid cloud strategies continued to trend in 2025, leading to a rise in interest around specialized AI cloud or neocloud providers that can support the growth of AI workloads. 

Neocloud revenues grew 205% year-over-year in Q2 2025 and were on track to exceed $23 billion for the full year, according to an October report from Synergy Research Group. The market researcher predicted that neocloud providers will reach almost $180 billion in revenue by 2030 and will grow an average 69% annually. 

The segment of specialized AI cloud providers faces fierce market competition in 2026, needing to double down on their entry into the realm of enterprise buyers — a group that’s also increasingly interested in what neoclouds have to offer. More than 80% of enterprise cloud buyers want to modernize their cloud strategies, according to the International Data Corporation’s Cloud Pulse survey released in February 2025. 

Neoclouds can undercut hyperscalers with aggressive pricing and simplified services, providing an attractive alternative for enterprise cloud buyers, said Dave McCarthy, research vice president at IDC.  

Additionally, enterprises are becoming more comfortable with picking the right cloud for the job and understanding that each provider has strengths and weaknesses, McCarthy said. The strategy shift gives emerging providers an opening. 

“These neoclouds, as long as they can continue to provide some differentiated services, will find a piece of the business they can keep long term,” McCarthy said. 

AI adoption accelerated compute demand

Neoclouds began as independent GPU-as-a-service providers, coming into force in the last two years when there was a “real scarcity of GPU resources,” said Pankaj Sachdeva, a senior partner at McKinsey & Company. 

“There is a meaningful amount of upfront capital that’s needed in ownership of GPUs,” Sachdeva said. “Many enterprises didn’t really have a clear roadmap and predictive model of how the usage of AI compute scaled, so leasing was a better option than buying at that point.” 

As agentic and advanced reasoning AI model adoption accelerated demand for compute, neocloud providers stepped in to fill a supply gap. The neocloud industry has evolved in terms of offers, customers, contract duration and the overall market structure, Sachdeva said. 

While neocloud providers might have started by simply leasing GPUs, their offerings and capabilities became more relevant to enterprises, especially with CIOs working on AI-related pilot projects, Sachdeva said. 

Scaling AI is “quite nascent,” with only one to three projects out of five to 20 reaching deployment for the average enterprise, Sachdeva said. 

“It’s at that inflection point of going from pilots to really scale up that they’re having to think, ‘Now I have to think strategically about my AI workloads,’ and they’re evaluating all kinds of options,” Sachdeva said. 

Options could include continuing to grow with a public cloud provider or adopting a hybrid approach to keep some AI workloads on premises and some in the cloud, Sachdeva said. Or a company might want to consider working with a new cloud provider. 

“With neoclouds having a value proposition — sometimes price — that is also coming up as an attractive alternative for enterprises,” Sachdeva said. 

The next wave

CoreWeave is the top hyperscaler competitor among neoclouds, according to Synergy’s market analysis. 

However, it has a ways to go toward greater enterprise adoption, as its top customers currently include companies like Microsoft and OpenAI. Microsoft has spent billions for access to CoreWeave’s compute, while OpenAI entered into a more than $22 billion contract for the neocloud’s services. 

CoreWeave and other neocloud providers such as Vultr, FluidStack and DataCrunch are feeding a need for AI infrastructure that will continue gaining momentum in 2026, McCarthy said. 

“These types of companies represent the next wave of cloud providers that are capturing attention,” McCarthy said. “This is a different approach than perhaps the original clouds that were very general purpose in nature.” 

CoreWeave’s focus is on providing and supporting AI capabilities, said Corey Sanders, SVP of product management at CoreWeave. While the company serves AI researchers and developers, Sanders said interest from enterprises and financial services has been slowly growing. 

“Customers who are doing AI-focused efforts — training, research, inferencing — both the importance to their business and the cost to their business in doing that work is very high,” Sanders said, adding that they’re willing to look beyond their existing cloud providers for other options. 

The downside to neoclouds is overspecialization, McCarthy said. Some neoclouds lack the breadth for enterprise needs. However, neocloud providers such as Vultr ride the line between being general purpose cloud and AI specific, resulting in a more diverse customer base. 

The neocloud market faces an “existential question” heading into 2026, Sachdeva said.

Neocloud providers can maintain market growth by continuing to build their business around one or two large customers such as hyperscalers or AI labs, Sachdeva said. However, future growth and success will be rooted in enterprises. 

While the value proposition of neoclouds exists for enterprises, the industry will need to overcome barriers to greater adoption, including adjusting their go-to-market and distribution strategies, creating more out-of-the-box functionalities for enterprises and tailoring service level agreements for business critical applications, Sachdeva said. 

“In the end, the long-term sustainability of the neocloud model is actually premised in some sort of enterprise adoption,” he said.

AWS is expanding application modernization options in collaboration with IBM as enterprises pursue hybrid cloud strategies with interoperability between AWS cloud computing services and IBM Z mainframes, AWS said in a blog post.

Alongside IBM’s Think 2026 conference, AWS identified five patterns supporting enterprises taking a hybrid cloud approach, including real-time business data exchange between core mainframe applications and AWS following IBM’s $11 billion acquisition of Confluent, a data streaming platform. Enterprises can also synchronize mainframe data with AWS, enabling cloud-based analytics and AI model access across trusted mainframe data, the blog said. 

“A hybrid strategy that includes IBM Z and AWS allows for rapid AI-[led] innovation that makes it easier to access applications and data without significant changes, and optimizes the costs of running or extending applications,” AWS said in the blog post. “Together, this approach maximizes business agility and return on investment.” 

Costly legacy mainframe modernization projects are taking a turn as the latest AI developments spur vendors to roll out tools and initiatives simplifying the process. 

AWS’ expanded collaboration with IBM is the latest example of vendors attempting to bring critical legacy infrastructure into the modern era, making it easier for enterprises to speed modernization and gain more value from core platforms. 

Capgemini last year released a code conversion toolkit powered by generative and agentic AI for refactoring COBOL applications. Rocket Software also last year launched a suite of AI-driven modernization services that includes mainframe anomaly detection automation. 

Hyperscalers AWS, Microsoft and Google have their own products aimed at mainframe application modernization, and enterprise customers are taking advantage. Toyota Motor North America tapped AWS Transform, an AI agent service that became generally available a year ago, to help automate its mainframe migration, leading to the modernization of more than 40 million lines of COBOL code to Java in half the time it would’ve taken without AI assistance, according to AWS. 

Even IBM, the biggest player in the mainframe space, has made strategic investments to help accelerate mainframe application modernization, including AI-assisted modernization through watsonx Code Assistant for Z or IBM Bob, CEO Arvind Krishna said during the company’s Q1 2026 earnings call in April. IBM Bob became generally available last month.

“Enterprises are investing in capabilities that increase resiliency, productivity and accelerate growth. They are modernizing core systems,” Krishna said during the call. “They are scaling AI and they’re making deliberate choices about where workloads should run and who controls the infrastructure underneath them.” 

AI will run across public, private and sovereign clouds, as well as on premises, with mainframes remaining a critical enterprise infrastructure component, which is why IBM has taken steps to position itself as a hybrid cloud and AI platform company, Krishna said.

Google is handing tech leaders new tools to support adoption of AI agents, including the introduction of Agentic Data Cloud at Google Cloud Next ‘26. The AI-native architecture aims to turn legacy enterprise data platforms into reasoning engines. 

The Agentic Data Cloud powers a universal business context engine for AI agents and a cross-cloud lakehouse to connect an organization’s data estate and eliminate silos, the announcement said.

The new cloud product targets boundaries between data platforms and AI agent functionality, according to Dion Hinchcliffe, VP and practice lead, CIO and technology buyers, at The Futurum Group. “For CIOs, this can be compelling because it directly addresses multicloud data sprawl and governance gaps,” he said in an email to CIO Dive. 

As vendors bet big on autonomous agents executing tasks on businesses’ behalf, they’re recognizing a growing need for agents to operate across clouds and gain access to critical data and context. 

Businesses are leaning toward hybrid cloud architectures, with 84% of cloud leaders intentionally selecting multiple clouds, according to Kyndryl’s 2025 Cloud Readiness Report. Yet agentic AI success hinges on connected infrastructure that allows data and intelligence to move, the report said. 

Google and AWS launched a collaboration in December to address this challenge and simplify multicloud connections, with Microsoft Azure planning to join the initiative later in 2026. More recently, CoreWeave collaborated with Google Cloud to enable AI training and inference across their cloud products, even in some instances across AWS, Microsoft Azure and on premises.

Cross-platform integrations are happening across vendors. Software company Databricks expanded cloud partnerships with Microsoft and Google Cloud last year. Meanwhile, Snowflake, a cloud-based data platform, restructured its data cloud to support enterprise AI agent adoption, connecting its cloud platform to Google Drive and Salesforce Data Cloud. 

With Agentic Data Cloud, Google is taking aim at latency and fragmentation challenges between data insights and actions — a core enterprise pain point, Hinchcliffe said. Through this approach, Google appears to be making the data layer actionable by enabling cross-cloud access, continually enriching context and allowing agents to operate directly on the data, he added.  

The move helps position Google as the most AI-native cloud offering, while AWS focuses on infrastructure and Microsoft targets applications and workflows, Hinchcliffe said. 

“Google is betting that whoever owns the data context layer for agents will control enterprise automation outcomes,” Hinchcliffe said. “If Google succeeds, it doesn’t need to win the entire cloud stack, it just needs to become the system where enterprise data is understood well enough for their agents to act better than other agents, which is in fact the highest-value layer in an agentic enterprise.” 

Along with Agentic Data Cloud, Google also launched Gemini Enterprise Agent Platform, allowing companies to build, scale and govern AI agents, according to a Wednesday announcement. The Agent Platform marks the next stage of Vertex AI, the announcement said.

Moving forward, Vertex AI will be delivered through the Agent Platform rather than as a standalone service to power agent development.

Three-fourths of business leaders are concerned with the geopolitical risks of storing and managing their data in global cloud environments, according to the Kyndryl’s Cloud Readiness Report. 

Local requirements such as digital sovereignty regulations have pushed 65% of business leaders to change their cloud strategies as a result, the report found. 

“With AI demanding seamless data access and governance requirements rapidly evolving, a hybrid cloud model is the differentiator that enables successful AI adoption,” Nicolas Sekkaki, global cloud practice leader at Kyndryl, said in a press release. 

Hybrid cloud strategies can help businesses support AI functionality while adhering to local regulations for data storage and management. 

Indeed, 84% of cloud leaders intentionally opt for multiple clouds, the Kyndryl report found. Another 41% reported repatriating some of their data from public cloud to on-premises servers. Integration between public and private clouds has improved, allowing enterprises to use the best option for each workload, especially in countries with strict data rules, according to the report. 

Hyperscalers are investing billions to build out cloud servers for customers to store data locally, Kyndryl found. 

Google plans to invest 5.5 billion euros (around $6.4 billion) over the next four years in cloud and AI infrastructure in Germany, according to a Tuesday announcement. The hyperscaler said it would continue to offer sovereign cloud products to let companies adopt advanced cloud and AI capabilities while adhering to local regulations. 

Microsoft also expanded its sovereign cloud offerings in June with the launch of Microsoft Sovereign Cloud, a suite of offerings spanning public cloud and private digital infrastructure, according to an announcement. 

While hyperscalers lead the market for cloud and AI services, the rise of specialty cloud providers such as Akamai, Vultr and Expedient, enable businesses to meet certain use cases that might fall between the cracks, Gartner Distinguished VP Analyst Ed Anderson said during the Gartner IT Symposium/Xpo in October. That includes sovereign clouds to meet location-specific requirements, he added. 

“Hybrid is going to be critical no matter what,” Anderson said.