Growing costs and increased cybersecurity complexity are driving some companies to repatriate workloads from the cloud to the data center.
One huge problem with expanding technology associated with today’s innovation is how to power it. Companies moved away from local storage to the cloud, but even with new tech like AI, we still have to figure out where to put infrastructure. In the latest report from CoreSite, “The 2025 State of the Data Center,” companies reveal just how complex their balancing act is to optimize workloads across a hybrid environment.
What are the key findings of the “2025 State of the Data Center” report?
At one point in pursuit of digital transformation, everything was moving to the cloud. However, after surprise costs and an increase in cybersecurity complexity, more than a few organizations have repatriated at least part of their workloads back to an on-premises location. CoreSite’s report finds that 98% of leaders plan to adopt a hybrid IT model or have already.
This trend leaves a gap. Workloads may be optimized across a hybrid system, but moving them to different locations reduces connectivity. 93% of respondents pointed to connectivity as a key feature for digital transformation, but only a small amount (19%) say their colocation provider offers it. That confidence seems to have dropped in the past few years based on the numbers from previous reports.
As companies continue to customize their tech stack to collect, store, and use data, connectivity is a powerful piece that determines how efficiently they can use it. The report shows that AI, a highly data-driven application, is fundamentally changing infrastructure strategies as companies try to manage its intensive workloads.
But they still lack the connectivity they need. The report theorizes that one reason for this is that many rely on colocation from their primary cloud provider, which may not offer native connections to third-party apps. This requires companies to engage in a complex game of building blocks to try to overcome connectivity issues. Unless they can get those builds just right, they’re looking at security loopholes, inefficiencies, and blind spots within their stack.
See also: What Is MLOps and How Will It Change the Data Center?
Why do genAI workloads work well in colocation?
The report lists the usual suspects for shifting to colocation: cost, compliance, and security. In all these cases, companies making the shift could see real support in protecting assets and maintaining budgets. But the number one reason companies are looking towards colocation is interconnection.
For some, this is because AI shares much of the same workflows as existing BI strategies. Direct interconnection can reduce transfer times as well as egress costs because these data-intensive applications share similar infrastructure.
AI is shifting the balance towards colocation, however. Colocation may offer more control over things like infrastructure placement or hardware configuration, so users can better manage the balance between performance and cost. In some cases, things that might be too cost-prohibitive solely in the cloud could slide within reach.
And there are practical concerns. Just the cooling and power solutions might be too much for a single organization to handle, so it’s not surprising that a majority of respondents cite this as a practical benefit of switching to a colocation provider. Additionally, support for infrastructure and expertise in deploying AI/ML solutions help bring these new technologies logically within an existing environment.
Is compliance an important part of AI implementation?
Traditional approaches to risk management may not address the full scope of danger in AI workflows. Data protection is a big part of new regulatory requirements, and data-hungry applications like genAI will need a new approach.
Additionally, inconsistencies in regulation for AI itself, as well as much talk about possible expansion or restriction within governing bodies, create a need for a more forward-thinking approach to security in the first place. Colocation facilities could help blend the need for physical security with accessibility and the opportunity to scale up or down as needed.
The report notes that colocation centers can provide higher levels of security and meet geographic specificity in ways that address more stringent compliance demands. This is an attractive feature of colocation for respondents, showing up as one of the top three business drivers for shifting those workloads to a colocation center.
Colocation centers can provide users with greater control as well as transparency in decision-making. The public cloud is still an attractive option for optimizing business processes that must flex on a moment’s notice, but going hybrid in some capacity could ease those security concerns.
Security is also ranked as the number one factor in choosing a colocation provider. With support and service quality and total cost of ownership coming in as second and third, respectively, we’re likely seeing the maturation of partnerships between service providers and users, rather than total reliance on service providers to make all the decisions.
Data centers are still in the mix
A blend of public, on-premises, and colocation options can give companies better control and better optimization over their IT investments. Connectivity from providers will determine what companies are able to do with their data and how much return they can get from AI deployments. Plus, the relationships with specialists in AI infrastructure can help organizations shorten the curve between AI investment and their ROI. This is an important relationship, and that isn’t likely to change over the coming years.
