Don’t give up on the private cloud just yet.

That’s one message from my interview with Red Hat CEO Jim Whitehurst last week. While public cloud providers like Amazon Web Services (AWS) and Microsoft Azure are riding high, Whitehurst insists that he keeps hearing that for certain kinds of apps, private cloud offers better performance and lower costs.

He may be right. Particularly at scale.

Public all the way down

Clearly, there’s something to the public cloud. As I’ve written, public cloud makes it easy for enterprises to experiment with big data and other critical workloads. As developers wield rising levels of influence, traditional software licensing schemes slow them down.

So, they abandon on-premise software to go directly to the cloud.

Not surprisingly, 83% of enterprises report unauthorized (public) cloud use. Developers, and the lines of business they support, can’t help it. They need to get stuff done.

Private cloud, by contrast, feels like an attempt by old-school enterprise IT to dress itself up in new school cloud clothes. Or like the “cool mom” in Mean Girls. (“I’m not regular IT. I’m cool IT.”)

And so, public cloud growth dramatically outpaces private cloud growth (20X vs. 3X), and private cloud deployments continue to underwhelm.

But this isn’t the whole picture.

When private cloud makes sense

Whatever the public cloud adoption numbers, the private cloud is paying off for Red Hat (and other vendors). As Red Hat executives touted on a recent earnings call, several of Red Hat’s top deals for the quarter included OpenStack, and “the number of times [a] top-30 deal included OpenStack or OpenShift this quarter tripled from Q4 a year ago.”

The question is, why?

According to Whitehurst, “The public cloud gets to become obscenely expensive at scale.”

For workloads that “don’t vary a lot in usage,” Whitehurst told me in an interview last week, the private cloud makes a lot of sense:

“A number of enterprise customers tell us that apps that don’t vary a lot in usage are significantly cheaper to run in their own data center than on the public cloud. Looking at AWS margins, you can see there’s some arbitrage between actual price charged and what it costs them.”

When I prodded him on comments from AWS data science chief Matt Wood (“You need an environment that is flexible and allows you to quickly respond to changing big data requirements”), Whitehurst remained unmoved. “If you’re at scale, and your workload isn’t varying a lot, once you get beyond dev and test experimentation, you’re going to want to do it in your data center.”

In other words, start in AWS–but it’s going to make sense to finish in the data center.

While it’s absolutely the case that AWS adoption is (far) more than dev and test workloads, it’s also the case that some enterprises have embraced Whitehurst’s pattern for cloud adoption. As I’ve profiled before, Etsy is a “cool kid” company that moved off the public cloud to run within its own data center. Surely, there will be more.

After all, as Whitehurst points out, “Look at AWS margins and they’re pretty high,” making plenty of room for companies to run workloads at lower cost.

Room for improvement

To get there, however, the private cloud vendors have a lot of work to do. Though OpenStack, the open-source darling of the private cloud movement, has roughly 1.2 gazillion community members, it remains a rounding error in terms of broad deployments. Gartner can only find a few hundred of them.

Some of this comes down to the complexity that continues to plague OpenStack, as OpenStack chief Randy Bias argued in an excellent keynote at the recent OpenStack Summit. The community needs to dramatically simplify the OpenStack experience:

While one OpenStack pioneer bemoans OpenStack for having “lost its heart” to money-grubbing corporate interests (apparently, he’s unfamiliar with Linux and every other broadly adopted open-source project), the reality is that corporate interests like Red Hat are precisely what OpenStack needs to improve and become more enterprise ready.

It may be the case, as Whitehurst argues, that private clouds make sense for predictable workloads at scale. There are good reasons to believe enterprises must move at such a frenetic pace now that “predictable” and “enterprise” may never sit easily together in the same sentence again. And it’s probable that AWS will always run its infrastructure more efficiently than an enterprise can.

Even so, as OpenStack matures, it will become a safer choice for enterprises still reluctant to entrust their workloads to the public cloud. Red Hat has a big part to play in maturing OpenStack. May the best cloud win.