Data center job postings have more than doubled over the past two years, even as overall US job postings declined approximately 12%, according to a July 2026 analysis from\u00a0Indeed Hiring Lab<\/a>:\u00a0\u201cSix in every 1,000 US job postings are now data center-related, up from 2 per 1,000 in May 2023.\u201d<\/p>\n Why enterprises are going on-prem again <\/p>\n It\u2019s not just the hyperscalers or massive data centers that have many communities up in arms that are driving this demand \u2014 enterprises are also expanding on the centers within their walls, Ammar Maraqa,<\/a> senior VP and chief strategy officer for Cisco, told ZDNET. <\/p>\n In his role, he is seeing firsthand the resurgence in demand for hardware and data center-centric roles, such as hardware engineers, network specialists, and security experts. Enterprises are increasingly looking inward \u2014 to manage their own data and infrastructure for security and cost reasons. \u201cAI inferencing and training blew all of the assumptions about smaller data centers out of the water because you needed new infrastructure, you needed GPUs, not just CPUs,\u201d he said.<\/p>\n In addition, there\u2019s a shift toward enterprises owning their own proprietary data and keeping it in-house. \u201cOn-premises infrastructure also helps enterprises not only manage and control IT security and token costs, but also enables greater creativity,\u201d Maraqa said. Importantly, sovereignty and data governance also play into the trend toward on-premises. \u201cThat\u2019s becoming the new [intellectual property].\u201d<\/p>\n The expertise in demand now <\/p>\n That means more career opportunities in the infrastructure and systems management space, especially with the push to modernize infrastructure capable of handling massive AI workloads. Demand is especially rising for expertise in \u201cdeep-tech\u201d fields such as hardware, networking, and cybersecurity, Maraqa said. \u201cRacks and GPUs are becoming so large that specialized skills are needed to network them effectively, as well as network across data centers.\u201d<\/p>\n Sure, coding and building AI-driven applications may be getting easier and easier thanks to vibe coding<\/a>. \u201cBut these can\u2019t just be built out of thin air,\u201d he said. \u201cThere needs to be infrastructure, support, and security.\u201d Data center professionals are needed to observe and manage the performance of these systems. \u201cAre your GPUs running hot? Is your storage failing?\u201d<\/p>\n Also: Could AI really destroy us all, or are humans still the bigger threat?<\/a><\/p>\n Expertise at connectivity and networking is becoming very important, he said. \u201cIn the past, you had to understand the workloads that were running on your infrastructure. Now, a lot of workloads are AI workloads. You need to understand the stack that\u2019s being created, whether it\u2019s the models, the interface of the models, or the prompt.\u201d<\/p>\n There is a shortage of people \u201cwho understand the deep-tech managing of data centers, and understanding how to optimize these really complicated large-scale systems,\u201d said Maraqa. \u201cWe probably won\u2019t have enough expertise relative to the demand.\u201d Addressing these concerns requires \u201can entirely new set of security and observability,\u201d he added. \u201cIf you don\u2019t solve these problems, it\u2019s very difficult to get the full benefit of AI.\u201d<\/p>\n AI and associated models add a new set of priorities for those running data centers. \u201cYou need to monitor the behavior of agents that are in your environment,\u201d he said. \u201cIt requires a very different type of security and observability where you\u2019re testing the behavior against guardrails that you\u2019ve set up.\u201d<\/p>\n The problem with outdated infrastructure <\/p>\n Cisco\u2019s own research shows that an AI agent requires four and a half times the load on the network as a human for an individual task. \u201cIf you think about it, it makes sense; it\u2019s constantly going; it keeps sort of pinging the network,\u201d he said. An infrastructure absorbing agentic AI workloads, then, needs to be modernized. \u201cThere\u2019s a lot of aging infrastructure, all up in a lot of enterprises and public sector customers. Some of the infrastructure is so old that you can\u2019t even patch it.\u201d<\/p>\n Also: Who\u2019s responsible for catching rogue AI agents? You are<\/a><\/p>\n Outdated infrastructure poses major security risks, he continued. \u201cOur enterprise customers are worried about the security of all of these models \u2014 roles and agents and all of that,\u201d Maraqa said. \u201cThey\u2019re also worried about how to make sure they have the right data access. How do they make sure they keep the data access within the enterprise? How do they expose them in the right way?\u201d<\/p>\n Fine-tuning your AI strategy <\/p>\n Employing in-house models also avoids the overkill inherent in tapping into the large frontier models. \u201cThere are tasks that only need cheaper small language models,\u201d he explained. For example, Cisco has a resident deep-networking model to help its network administrators and customers troubleshoot networking. \u201cYou don\u2019t need a frontier model for that. You need a very specific local language model that is trained for networking expertise or security expertise. That needs to run on infrastructure that sits on-premises.\u201d<\/p>\n The advent of AI feels \u201cvery different\u201d from previous technology incarnations, he said. \u201cI\u2019m old enough to have gone through the e-commerce transition, the cloud transition, and the mobile transition,\u201d he recounted. \u201cI think all of those were monumental, but this one feels fundamentally different. Everything feels very demand driven and everything feels constrained all the same time. There are all kinds of use cases being discovered every day. It feels like opportunities are pretty limitless in terms of how this technology can help and transform.\u201d<\/p>\n \t\t\t\t\t\t\t \n\t\t\t\t\t\t\tJoe McKendrick\n\t\t\t\t\t\t<\/p>\n \t\t\t\t\t\t\tContributing Writer\t\t\t\t\t\t<\/p>\n \t\t\t\t\t\t\t\t\t\t\tJoe McKendrick is an author and independent analyst who tracks the impact of information technology on management and markets. He has also served as co-chair of the annual AI Summit in New York since 2021. He has authored numerous research reports in partnership with Forbes Insights and Unisphere Research, a division of Information Today, Inc. Joe speaks frequently on AI, cloud, data, and enterprise computing topics at industry events and Webcasts. In a previous life, Joe served as director of the Administrative Management Society (AMS), an international professional association advancing knowledge within the IT and business management fields. He is a graduate of Temple University.![\"Joe](\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2026\/09\/joe-mckendrick.jpg\")
\t\t\t\t\t\t<\/p>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\tSee full bio\t\t\t\t\t<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"ZDNET\u2019s key takeaways Data centers were once consolidating. Now they\u2019re proliferating. AI is driving data center expansion within…\n","protected":false},"author":2,"featured_media":913424,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[49,48,61],"class_list":["post-913423","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology","tag-ca","tag-canada","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/913423","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/comments?post=913423"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/913423\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media\/913424"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media?parent=913423"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/categories?post=913423"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/tags?post=913423"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}