Private cloud providers may be among the key beneficiaries of today’s generative AI gold rush as, once seemingly passé in favor of public cloud, CIOs are giving private clouds — either on-premises or hosted by a partner — a second look.

At the center of this shift is increasing acknowledgement that to support AI workloads and to contain costs, enterprises long-term will land on a hybrid mix of public and private cloud.

“With how fast things are changing in the data and cloud space, we believe in a hybrid model of cloud and data center strategy,” says Jim Stathopoulos, SVP and CIO of Sun Country Airlines, who came to the regional airliner from United Airlines in early 2023 and inherited a Microsoft Azure cloud infrastructure and Databricks AI platform but is open minded about future IT decisions.

Controlling escalating cloud and AI costs and preventing data leakage are the top reasons why enterprises are eying hybrid infrastructure as their target AI solution. Most observers agree that a hybrid approach that involves on-premises or co-located private clouds will be required for most IT leaders to ensure cost control and data integrity in the face of AI’s resource needs and key business concerns around its use.

Here, private cloud platforms such as Dell APEX and HPE GreenLake, adorned with generative AI support, could provide enterprise customers an answer, as could co-locating with partners such as Equinix to host workloads in private clouds, says IDC’s top cloud analyst, Dave McCarthy.

“The excitement and related fears surrounding AI only reinforces the need for private clouds. Enterprises need to ensure that private corporate data does not find itself inside a public AI model,” McCarthy says. “CIOs are working through how to leverage the most of what LLMs can provide in the public cloud while retaining sensitive data in private clouds that they control.”

Somerset Capital Group is one organization that has opted to go private to run its ERP applications and pave the way for generative AI. The Milford, Conn.-based financial services company migrated data to the public cloud more than a decade ago and will continue to add workloads there, especially for customer-centric applications. But the company’s critical data — and any future generative AI data — will likely run on its new hosted private cloud, says Andrew Cotter, Somerset’s EVP and CIO.

“As we are testing and dipping our toes in the water with AI, we are choosing to keep that as private as possible,” he says, noting that the public cloud has the horsepower needed for many LLMs of today but his company has the option of adding GPUs if needed via its privately owned Dell equipment. “You don’t want a mistake to happen and have it end up ingested or part of someone else’s model. We’re keeping that tight control and keeping it in the private cloud.”

Todd Scott, senior vice president for Kyndryl US, acknowledges that AI and cost are among the key factors driving enterprises toward private clouds.

“Most enterprises are currently exploring AI on the public cloud, but we expect clients will ultimately bring the app to their data and run AI where the data is, in private environments and at the edge,” he says.

“Another factor that’s driving a move back to private cloud is predictability of cost,” Scott says. “Agile enterprises, by definition, make frequent changes to their applications, so they sometimes see big fluctuations in the cost of having their data on public clouds. Private clouds provide more predictability because the infrastructure is dedicated.”

According to Forrester’s Infrastructure Cloud Survey in 2023, 79% of roughly 1,300 enterprise cloud decision-makers surveyed said their firms are implementing internal private clouds, which will use virtualization and private cloud management. Nearly a third (31%) of respondents said they are building internal private clouds using hybrid cloud management solutions such as software-defined storage and API-consistent hardware to make the private cloud more like public cloud, Forrester adds.

IDC forecasts that global spending on private, dedicated cloud services — which includes hosted private cloud and dedicated cloud infrastructure as a service — will hit $20.4 billion in 2024, and more than double by 2027. 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.

While those numbers pale compared to public cloud’s expected $815.7 billion in 2024, IDC’s McCarthy sees hybrid cloud infrastructure as the future for most enterprises in this area. The emergence of turnkey private cloud offerings from HPE and Dell, McCarthy says, gives customers a private cloud they can run on premises or in a co-location facility that provides managed services.

Private clouds may also help enterprises better control their overall cloud costs, but there will be pros and cons for both, he points out.

“Enterprises are in a bit of a pickle with this,” McCarthy says. “Security concerns are what is driving them to private cloud, but the specialized hardware required to do large-scale AI is expensive and requires extensive power and cooling. This is a problem that companies like Equinix believe they can help solve, by allowing enterprises to build a private cloud in Equinix datacenters that are already equipped to handle this type of infrastructure.”

Tom Richer, founder of startup advisory Cloudbench, and a former Microsoft, Deloitte, and IBM executive, agrees private cloud providers and co-location partners will enjoy a bump from AI.

“There is a huge upside for the private cloud industry, and they need to embrace genAI for growth. GenAI models can be trained to learn patterns from normal data distributions and detect anomalies or outliers in real-time data streams. Private cloud platforms can leverage generative AI for anomaly detection applications in various domains, including cybersecurity, fraud detection, and predictive maintenance,” he says. “Private cloud providers can offer generative AI services to help businesses develop predictive models, conduct simulations, or perform ‘what-if’ analyses without the need for large datasets or extensive computational resources on premises.”

Still, some IT leaders remain comfortable running all workloads on the public cloud, even with the data privacy concerns generative AI imposes.

Steve Randich, CIO of the Financial Industry Regulatory Authority, an IT services arm that supports the US Securities and Exchange Commission, agrees the evolution of AI may open doors to private clouds. But he maintains the security of AWS public cloud is rock-solid.

“That is a valid concern, but I think it pertains to some cloud providers more than others,” Randich says, noting that FINRA has one of the biggest data sets on AWS. “Our position is to continue using AWS public cloud services and are not considering private or hybrid.”

Whatever the future holds for IT leaders’ decisions on enterprise AI workloads, private clouds, once declared DOA, are primed for the AI moment, vendors contend.

HPE GreenLake, for example, is an as-a-service offering that brings cloud-like flexibility to data centers and other locations with higher levels of data security, according to HPE. Such private cloud solutions eliminate the risks of multitenancy data leakage, for example, a key CIO concern with AI. Home Depot and R&D lab Ofino, which claims to have slashed cloud costs up to 90% with HPE, has partnered with HPE co-location partner CyrusOne for their private cloud needs.

“AI is like the ultimate hybrid workload,” says Hang Tan, hybrid cloud COO of HPE. “You’re going to have some very bursty things, but then for large model training and even for inference, there’s going be a base load that, if you can predict [that load] pretty confidently, that should all sit in private cloud.”

Voya Financial, which opted for a public-private mix when launching its digital transformation in 2018, is one organization taking a measured approach to workload placement. Although EVP and CIO Santhosh Keshavan’s chief concern is possibly missing out on innovation on the public cloud, the decision to retain a private cloud footprint has been best for his company in all respects.

“Not every application is ready to be run on the cloud. And for those applications that do not need to be run on Amazon or AWS or any of the clouds, we tend to keep them in a private cloud,” he says.

Just as not every application is best served by the public cloud, not every organization is as well.

The City of Williamson, Texas, which sits near Austin and a major Samsung semiconductor facility under construction, has been following a private cloud path since 2017 due primarily to the legal liability of courts’ digital evidence and the bandwidth needs of government agencies in a rapidly growing area, says CIO Richard Semple.

“We have no choice. At the end of the day, even if you’ve done everything perfectly, we are still responsible for the security and integrity of that data, and we’re going to transfer the risk to someone else?” Semple says.

Still, the CIO is not ruling out a migration to the public cloud if the variables change.

For all CIOs rethinking their cloud strategies, there are few variables that can change the equation as fast as AI.

By admin