TL;DR
A survey commissioned by DDN with Google Cloud and Cognizant finds that over 50% of AI initiatives at large U.S. firms were delayed or canceled in the past two years, largely because infrastructure is hard to manage. Respondents pointed to complexity, underused hardware and rising power costs, and experts say workforce education and better-aligned use cases are needed.
What happened
DDN commissioned research, conducted with involvement from Google Cloud and Cognizant, polling 600 IT and business decision-makers at U.S. companies with at least 1,000 employees. The study reports that more than half of AI projects were postponed or terminated over the last two years, and roughly two-thirds of those surveyed call their AI environments too complex to operate. DDN’s CEO highlighted operational shortfalls—such as underutilized GPUs and higher power bills—that make many investments uneconomic. The report echoes other recent industry findings: large studies and analyst forecasts have flagged limited measurable returns, cancellations of agentic AI projects, and delayed AI spending. While nearly all surveyed decision-makers say scaling AI will require cloud environments, DDN’s leadership cautioned that cloud migration does not automatically remove needs for unified data and orchestration. The company and its partners work with major AI vendors to optimize data flow into compute infrastructure, and executives say integrators and better-defined use cases could help bridge the gap between pilots and revenue-generating deployments.
Why it matters
- High cancellation and delay rates suggest many organizations are failing to convert AI pilots into production, slowing digital transformation.
- Complex infrastructure drives inefficiency: hardware like GPUs can sit idle while power and operational costs rise, undermining project economics.
- Widespread belief that cloud is required for scale may lead to migrations that still leave unresolved orchestration and data-unification challenges.
- Education of IT staff and guidance from systems integrators could be decisive in helping enterprises pick realistic, value-driven AI use cases.
Key facts
- Survey sample: 600 IT and business decision-makers at U.S. enterprises with 1,000+ employees.
- More than half of AI projects were delayed or canceled within the past two years, per the commissioned study.
- About two-thirds of respondents described their AI environments as too complex to manage.
- 97% of those surveyed said scaling AI for their organization will need to happen in the cloud.
- DDN commissioned the research in partnership with Google Cloud and Cognizant; DDN works with vendors including Nvidia, xAI and Google.
- DDN’s CEO pointed to underused GPUs and rising power costs as signs that operational foundations for AI are often insufficient.
- Other industry analyses cited: MIT’s Project NANDA found most organizations report no measurable return from generative AI; Gartner predicted many agentic AI projects will be canceled by 2027; Forrester forecasted delays to AI spending and limited EBITDA lifts for most firms.
What to watch next
- Whether systems integrators and consultancies (named examples include Accenture and Deloitte) accelerate enterprise readiness and turnkey AI deployments.
- If organizations shift from generic chatbot pilots to use cases that directly connect existing data assets with AI to produce measurable business value.
- Whether cloud migrations reduce the operational complexity DDN highlights, or whether unified data and orchestration gaps persist (not confirmed in the source).
Quick glossary
- AI infrastructure: The combination of hardware, software, networking and data systems required to develop, train and run AI models.
- GPU: A graphics processing unit, often used to accelerate training and inference of AI and machine learning models.
- Cloud orchestration: Automated coordination of cloud resources and services to deploy and manage applications at scale.
- Systems integrator: A consultancy or vendor that designs and implements end-to-end IT solutions, including complex AI deployments.
- Generative AI: A class of AI models that create new content—text, images, audio or code—based on learned patterns from training data.
Reader FAQ
How many AI projects were affected?
The commissioned survey found that more than half of AI projects at large U.S. firms were delayed or canceled over the past two years.
What were the main reasons for shelving projects?
Respondents cited complexity of AI infrastructure; executives also noted underused hardware and rising power and operational costs.
Will moving to the cloud fix the problem?
Most surveyed decision-makers said scaling requires the cloud, but DDN’s leadership warned cloud migration alone does not automatically resolve data unification and orchestration needs.
Who ran the research?
The study was commissioned by DDN in partnership with Google Cloud and Cognizant.
AI + ML Over half of AI projects are shelved due to complex infrastructure The answer seems to be educating the enterprise workforce, and creating smarter use cases O'Ryan Johnson…
Sources
Related posts
- Remails: Building a European Mail Transfer Agent for Transactional Email
- AI labs see accelerating churn as staff jump between leading companies
- Inside OpenAI’s Talent Grab at Thinking Machines Lab and What’s Next