Approximately 25 percent of enterprises are already achieving consistent annual ROI from their AI investments, while another 35 percent expect to see returns within the next year, according to a new report published by DataBank, a provider of enterprise-class colocation, connectivity and managed services.
The report, “Accelerating AI: Navigating the Future of Enterprise Infrastructure,” documents how organizations are evolving from generic AI applications to customized models. The sponsored research indicates enterprises are quickly adopting hybrid infrastructure approaches and geographic distribution strategies as AI is targeted at business-critical applications.
“While most enterprises start AI initiatives in the cloud, they’re quickly adopting hybrid approaches that combine cloud, on-premises and colocation for different workloads,” said Raul Martynek, CEO. “Success requires infrastructure flexibility that can support both centralized training and distributed inference while meeting security and compliance requirements.”
Among major obstacles for implementation, integration challenges, scaling difficulties and talent shortages were identified as primary barriers. Just 20 percent of those surveyed cited poor data quality as a major concern.
In all, 76 percent of enterprises surveyed plan to expand AI infrastructure closer to data sources and end users for latency reduction and compliance. While AI training is being centralized, inference workloads are distributed globally, according to the report.
In addition, enterprises indicated they’re moving from generic, third-party AI and large language models toward customized or proprietary solutions. Infrastructure approaches are blending off-the-shelf applications, custom solutions and tailored deployments models, DataBank said.
The report can be downloaded for free from the DataBank website.
