Enterprise AI projects are running into infrastructure questions as organizations move from planning toward deployment, according to recent research Digital Realty. Forty percent of survey respondents cited a lack of specialized infrastructure required by AI as a main challenge to adopting a formal AI strategy, up from 9% in the tech giant's 2024 research.

The Global Data Insights Survey 2026 points to a widening mix of infrastructure models behind enterprise AI. Half of respondents host AI models in private clouds, 86% are pursuing or exploring sovereign AI, 92% tie data-location strategy to their AI prepares, and 88% follow a distributed-data approach.

The research is based on an online survey of 2,131 IT decision-makers across 19 countries, and 11 industries, conducted from Nov. 23, 2025, through Feb. 4, 2026.

Organizations also appear to be shifting from AI planning toward execution. The share actively executing AI to achieve operational efficiency or reduce costs rose from 27% in 2024 to 37% this year. Only 3% reported no measurable ROI to date, while 63% expect returns within the next six months to two years.

In a recent development, specialized infrastructure led the survey's list of AI adoption challenges at 40%, followed by regulatory and personal-data concerns at 35%. Data availability and quality, storage capacity, interconnection, compute resources, and staffing also ranked as significant constraints.

Those requirements are also shaping workload placement. Integration with existing IT or data infrastructure ranked first at 43%, while regulatory compliance and performance were each cited by 37%. Data proximity, sovereignty, sustainability, cost, and intellectual property protection also factored into hosting decisions.

Private cloud features prominently across the report's AI hosting categories, but enterprises are clearly using a mix of private cloud, public cloud, on-premises infrastructure, and colocation. Private cloud accounted for 50% of custom-built AI model hosting and 47% for open source or fine-tuned large language models (LLMs).

Sovereignty is another part of that placement decision. Thirty-two percent consider sovereign AI β€” in which data and AI models remain within a single national or regulatory jurisdiction β€” a requirement, while another 54% are exploring it. The findings point toward hybrid environments in which infrastructure location, governance and data control increasingly overlap.

Increasingly, cyber criminals are sharing AI prompts, jailbreak techniques, and access to existing models that can be incorporated into fraud and phishing operations, according the latest AI Threat Report from Flashpoint.

It has been revealed that as an umbrella term, education technology encompasses everything from learning management systems and student information systems to productivity tools and AI platforms. More specific language clarifying the difference between ed tech and consumer-facing tools would allow state leaders to match the right evidence, expertise, and policy to the right questions.

Artificial intelligence is saving scientists significant amounts of time, but researchers are giving some of those gains back by checking the technology's work, according to a newly introduced study involving Google, Google DeepMind, and MIT.

Microsoft is combining its security information and event management and threat protection products within Microsoft Defender, creating what the manufacturer calls an Integrated Security Operations Center built for an era of AI-driven attacks.

β†’ Read original article