A majority of companies are currently failing to successfully implement artificial intelligence (AI). According to DDN’s 2026 AI Infrastructure Report, 65% of organizations find their AI environment too complex to manage, leading to delays and lower returns on investment (ROI).
The report, conducted by Vanson Bourne and commissioned by DDN, leading AI data platform provider, with collaboration from Cognizant and Google Cloud, surveyed 600 U.S. IT and business leaders to uncover the hidden pressures slowing AI adoption.
The findings reveal a startling reality: Organizations are seeking to adopt AI, but most lack the foundation to sustain it. Rising complexity associated with infrastructure, underutilized cloud environments, unplanned energy requirements, and persistent skills gaps are quietly stalling projects and eating into ROI.
Key Findings Highlight the Stakes:
- Two-thirds of organizations (65%) say their AI environments are too complex to manage.
- Over half (54%) have delayed or canceled AI initiatives in the past two years.
- A majority (97%) overwhelmingly agree that cloud infrastructure is essential to scaling AI.
- Nearly all (93%) are actively seeking to reduce AI’s energy impact.
- Most (83%) say their internal teams are struggling with AI workloads today.
“The AI boom has hit an infrastructure wall,” said Alex Bouzari, CEO and Co-Founder at DDN. “Companies are chasing models and GPUs, but the real bottleneck is the data layer underneath. Without modern, unified infrastructure, AI can’t scale.”
Infrastructure Complexity Is Stifling AI ROI
The study exposes that AI infrastructure complexity, not capability, is the silent killer of ROI. Sixty-five percent of respondents say their AI environments are already too complex to manage, causing 54% of them to delay or cancel AI projects. With AI workloads projected to grow 110% in the next year, 76% of leaders still face fundamental data challenges, from legacy infrastructure and siloed datasets.
The reported complexity in managing AI environments, affecting two-thirds of organizations (65%), is fundamentally a consequence of infrastructure fragmentation; organizations deploy AI workloads across a patchwork of disconnected solutions, which are separate systems for data processing, training compute, and serving endpoints, none of which were inherently designed for the scale and demanding requirements of Gen AI. This fragmentation mandates continuous, complex data movement, necessitates manual, resource-intensive resource orchestration across disparate silos, and ultimately prevents the cohesive scaling of compute, storage, and networking necessary for efficiency.
Unified AI infrastructure, provided by leaders like DDN, purpose-built for scale, simplicity, and efficiency, is now the single biggest driver of success. Trying to retrofit traditional, fragmented systems to handle modern AI workloads rarely works and often guarantees failure. The organizations breaking through are those simplifying—not stacking.
“Enterprises are discovering that scaling AI isn’t a compute problem—it’s an integration problem. If your infrastructure isn’t unified, your AI can’t learn efficiently. Simplicity is the new scalability.” —Sven Oehme, DDN CTO
Cloud is the Smartest Place to Start AI
Ninety-seven percent of respondents say cloud is essential to scaling AI, and more than half cite it as their fastest path to production. Cloud-based deployments let teams experiment, onboard GPUs faster, adopt the latest technologies quickly, and reduce early-stage failure rates; making it the most common “launch zone” for successful AI adoption.
“The survey responses validate the investments we’ve made over the years to develop a robust cloud infrastructure that empowers organizations to easily scale AI workloads,” said Asad Khan, Sr Director, Product Management, Google Cloud. “With Google Cloud Managed Lustre, customers can leverage the latest GPUs and TPUs while reducing complexity and accelerating innovation.”
Tokens per Watt is the New Currency of AI
AI’s next constraint isn’t compute, it’s energy. AI’s rapid scale has created an unprecedented energy demand that has become an operating constraint in AI data centers. Most (93%) of respondents report they are actively seeking to reduce AI’s energy footprint, and nearly half (47%) cite power and cooling costs as their top infrastructure constraint.
The report highlights the importance of maximizing AI output per watt, a next-generation efficiency metric that quantifies how effectively AI workloads convert energy into usable compute. DDN is uniquely positioned to help organizations improve their energy usage situation by keeping GPUs fully saturated and moving data in parallel, enabling up to 70% reduction in power and cooling costs.
Massive AI Skills Gaps are Solved Through the Partner Ecosystem
Almost all (98%) organizations cite a skills shortage (in both IT and data science roles) as a major barrier to scaling AI. In fact, external research highlights that 65% of organizations have abandoned AI projects due to a lack of skills. However, the report shows that one of the ways leaders are closing this gap is through ecosystem collaboration, pairing internal expertise, proven reference architectures, and pre-tested solutions with partners like DDN, Cognizant, Google Cloud, NVIDIA, and others.
Most enterprises now recognize that external expertise is not a stopgap—it’s a strategic enabler. That’s why 72% rely on third-party expertise to build and manage their AI infrastructure, while just 12% depend solely on in-house talent. Organizations who make this connection are forming deeper, long-term partnerships that accelerate implementation while transferring knowledge and reducing operational friction.
