By Ryan Mallory, CEO, Flexential
Conversations around AI infrastructure over the last year or so have focused mainly on power, and for good reason. Location, power, and permits decide who gets to build, and that will continue to be the case in 2027. But when we asked enterprise leaders what is slowing down their AI projects, one issue stood out the most. In Flexential’s 2026 State of AI Infrastructure Report, 96% of respondents reported at least one network-related performance issue affecting their AI workloads over the past 12 months. Here is what that means for next year.
Two years ago, application latency ranked last among the network concerns in our survey. This year it ranks first, and the amount of respondents citing it has more than doubled. I expect site selection to follow that trend. Access to fiber and multiple carriers matters just as much as power and space, and connectivity has limited where facilities can operate, even where power is available. In 2027, buyers will be asking about the network as early as they ask about power: which carriers are available, how quickly their data can get where it needs to go, etc.
Demand for capacity, infrastructure, and skilled labor is going to stay high, with costs continuing to change rapidly. The operators who select their sites early, do their due diligence, and build to a consistent design will be able to deliver capacity more predictably than those starting each project from scratch. Capacity planning still needs to be a year or more out in order to work through the demand that has already been sold.
This also means that multi-tenant colocation is getting more and more interest from enterprises and investors because of steady growth. Change is coming in 2027, and buyers want partners who have been doing this for a long time and know how to deliver. When capacity is limited, many enterprises start with a smaller footprint, a managed private cloud, or interconnection to the providers they already use. Nobody should be left behind because they are not buying at hyperscale.
More companies are now running their AI compute in one place and storing their data in another. That likely won’t change in 2027. Companies want to keep their data somewhere they can control, and they often rent computing power. That means that the connection between those two points will be discussed early, and data protection needs to be part of that. Organizations need to get their data back fast in the event of an outage, so recovery plans need to be in place from the start. I expect more companies to plan for that up front. Immutable backups, tested recovery plans, and a separate recovery location will all need to become part of the conversation in 2027.
Additionally, as inference becomes more distributed, regional, network-dense data centers become more useful. More organizations are turning to third-party colocation for edge processing and fewer deploying AI hardware on their own premises. Both of these trends will continue, and second-tier markets will get more attention than they have previously.
Last but certainly not least, data centers will continue to be a big topic of conversation in local communities. Residents want to know what a data center does, how much power it uses, how much water it uses, and what it means for their town. Not all data centers are created equal, and the industry hasn’t always made that clear. In 2027, I expect operators to be a larger part in those conversations. We’ll show up at community meetings, answer questions plainly, and explain what we’re building and why. The people who work in our data centers live in these communities, too.
Power will stay at the center of the AI infrastructure conversation in 2027, but network, data protection, recovery plans, and community relations will all play a big role in development as well.
##
ABOUT THE AUTHOR

As Chief Executive Officer of Flexential, Ryan Mallory is responsible for setting the company’s strategic direction and leading its external-facing priorities—including customer and partner relationships, investor engagement, and long-term growth planning. He works closely with the leadership team to ensure continued alignment across brand, financial strategy, and platform development.






