Is generative AI bringing back private clouds?

According to Forrester’s Infrastructure Cloud Survey in 2023, 79% of about 1,300 enterprise cloud decision-makers surveyed said their organizations are implementing private clouds. Additionally, IDC forecasts that global spending on private, dedicated cloud services, including hosted private clouds, will hit $20.4 billion in 2024 and will at least double by 2027.

In addition, global spending on enterprise private cloud infrastructure, including hardware, software, and support services, will be $51.8 billion in 2024 and grow to $66.4 billion in 2027, according to IDC. Of course, public cloud providers are still the 800-pound gorilla in the room. Public clouds, including the big three AWS, Microsoft, and Google, are expected to rake in $815.7 billion in 2024.

The AI bump

Clearly, AI is driving this reemergence of private clouds, which have little more value than purchasing hardware and sticking it in a data center. Indeed, private clouds have dropped in popularity due to the features that public cloud providers offer, which far exceed what you get in an open-source private cloud system or those offered by enterprise hardware players today.

As AI workloads become more prevalent and complex, many organizations are reevaluating their cloud strategies. These days, the consensus among enterprise architects suggests a shift towards a hybrid cloud infrastructure.

One primary factor propelling this trend is the increasing need to control escalating costs associated with cloud and AI technologies. Public cloud providers are proving to be more expensive than their on-premises counterparts, and this fact is finally sinking in, as CIOs find their CFO at the door, looking for an explanation about the cloud resources costing them about 2.5 times the original estimate.

Key to this transition are private cloud platforms such as Dell APEX and HPE GreenLake (now equipped with generative AI support). However, I understand that there are many other “private cloud providers,” so hold your suggestions, PR people. Indeed, most on-premises systems can enter the private cloud market just by declaring it so. This private cloud washing was a huge part of the early days of cloud computing but has largely fallen by the wayside as the market matured, thank goodness. But it could be coming back.

False sense of security

These platforms provide the computational power, even GPUs, and the flexibility necessary for handling AI workloads. They also maintain strict control over data privacy and security, though this security is often more sentiment than reality. In many instances, public cloud providers provide better security due to their greater investments in their own solutions.

The rise of AI has heightened concerns about data security, specifically the risk of private corporate data being inadvertently fed into public AI models. Again, perception does not match reality, but I hear about this concern often enough that it’s worth exploring. I don’t see a world where public cloud providers not only accidentally access corporate data but also train their AI models on it. That would be a scandal of epic proportions. Nevertheless, many enterprises find the private cloud an attractive option as it allows them to retain “sensitive data” within a controlled environment.

Challenges arise

Despite their advantages, private clouds are not without their challenges. For instance, specialized hardware is required for large-scale AI operations, such as using GPU-powered servers. This can be cost-prohibitive and require extensive power and cooling systems, and enterprises have yet to understand the new upgrades and new costs that this will drive. In many cases, it will be more extensive than running these AI workloads on public cloud providers.

However, solutions are emerging, such as building private clouds within colocation data centers like those provided by Equinix. These guys are specifically equipped to handle these infrastructure requirements, and I consider them better options than building this stuff out DIY. After all, we need to get out of the data center business at some point, leaving it to public cloud providers, colocation providers, and managed service providers to provide better options.

So, is the private cloud a good option for enterprises? Sure, they have always been on the table for architects. They have their uses, and if they are more cost-effective or if they can return more value to the business, use them—AI or no AI.

My guess is that as AI technologies and their applications continue to evolve, the shifts in cloud strategies are expected to reflect an increasing preference for some private cloud alternatives. This trend suggests a promising future for private cloud solutions. I suspect those enterprise technology players who saw a diminishing interest in their private cloud offerings now have more pep in their step. Thanks, AI.