As hyperscalers look to recoup their massive investments in artificial intelligence (AI), IT infrastructure and operations (I&O) leaders must urgently adopt a platform-centric model and prepare for significant price increases for traditional cloud services.
Analysts speaking at the recent Gartner IT Infrastructure, Operations, and Cloud Strategy Conference in Sydney highlighted the key tension facing today’s IT leaders: managing the hype surrounding AI while maintaining core business operations and reducing costs.
“AI agents are right at the height of the hype, but as always, new and hyped technologies emerge all the time,” said Autumn Stanish, Director Analyst at Gartner. “And as always, we have to keep the lights on and go about our daily tasks.”
Paul Delory, research vice president at Gartner, said that while fundamental practices such as infrastructure automation and DevOps remain important, pressure to innovate is increasing. Cost reduction remains a top priority for CIOs as we head into 2026, and many companies are looking to AI to help realize those savings.
However, business demand for AI is currently outpacing I&O support. Stanish said half of I&O leaders believe integrating AI into their current infrastructure is their biggest challenge, warning that teams could lose relevance. To avoid a repeat of the loss of control experienced in the early days of cloud migration, Delory argued that I&O must evolve into a value-driving capability that can deliver AI agents, continuous operations, and a platform-centric model.
Over the next year, Gartner advised I&O teams to form dedicated AI centers of excellence and build fully automated delivery pipelines with tight cost controls. Delory noted that this could likely be accomplished within 90 days, as most of the tools needed are free and open source. Practical applications for AI agents in I&O today include automatically responding to infrastructure changes by updating scripts and playbooks, training AI to act as quality assurance engineers, and deploying compliance agents trained on human-readable policy documents.
Moving to this platform-centric model requires organizational changes. Stanish proposed creating a dedicated platform team that absorbed traditional server and storage engineers, led by a product owner who would directly align the technology with end-user needs. This will require new performance metrics, shifting the focus from basic baselines such as uptime to business outcomes such as revenue growth and customer satisfaction.
In addition to the management structure, the conference also highlighted the crisis in technology procurement. Luke Ellery, vice president analyst at Gartner, said a 2024 study found that 79% of buyers regret their technology purchase because it didn’t meet their expectations or settled for a lower-quality solution.
To address this, Ellery asked organizations to keep senior-level sponsors engaged throughout the purchasing process to ensure the final purchase is aligned with business needs, rather than simply allowing the procurement team to find a cheaper but unsuitable alternative. Buyers should adopt agile, lean procurement methods that focus on measurable business outcomes rather than detailed functional specifications and enable iterative improvements rather than rigid waterfall processes.
Additionally, Ellery advised leaders to use data to understand risks rather than blindly avoiding them, adopt risk tolerances carefully, and build confidence in vendor negotiations through targeted training and market knowledge.
The risks of inadequate investment are particularly acute in IT support. Gartner predicts that by 2027, half of all AI projects designed for service desks will be abandoned due to unanticipated costs, risks, or failure to meet expected return on investment.
To avoid being included in this statistic, Joe Rogus, director analyst at Gartner, suggested focusing on easily accessible features within existing software, especially virtual support agents (VSAs) that can move incidents away from human staff.
AI can also significantly support human agents through discovering unique knowledge using search augmented generation (RAG), converting chat logs into new knowledge base articles, and using machine learning to respond to endpoint anomalies. Additionally, AI can help automate ticket classification and routing, as well as case summarization, if organizations make the effort to clean up existing data first.
Logas also cautioned against the growing hype around agent AI, noting that many vendors are simply putting a label on basic automation tools. True agent AI requires giving systems the autonomy to perform their own actions, which requires high reliability and strict data hygiene.
Looking further into the future of cloud computing, Logas noted that public cloud spending is expected to exceed $1 trillion by 2027, largely driven by AI. But as hyperscalers pump hundreds of billions of dollars into AI infrastructure, they are expected to recoup those costs by raising the prices of traditional cloud services.
To prove the continued business value of cloud migration, Logas pointed to the rise of AI-infused industry-specific composable solutions that move from siled infrastructure to a core layer that supports data fabrics and packaged business functions.
This has major implications for IT strategies for 2030. On the sovereignty front, IT leaders must carefully distinguish between data sovereignty, operational sovereignty, and technology sovereignty, and balance the tradeoffs between global hyperscalers and local providers. Multi-cloud strategies will also need to be rethought, with Gartner predicting that most enterprises will eventually run intensive AI model activities in one cloud while leveraging data in another cloud.
Sustainability becomes a bottleneck as AI-optimized data center racks require significantly more power than traditional servers, potentially tripling energy demand by 2030. Meanwhile, as AI agents effectively act as digital workers within networks, security frameworks must evolve from static policies to dynamic, real-time approaches.
Finally, cloud financial management becomes non-negotiable. Gartner has warned that because the majority of AI workloads now run in containers, and those containers are significantly over-provisioned, companies that fail to optimize their computing environments could end up paying up to 50% more than their more efficient competitors.
