Data infrastructure, not models, now holds back AI

Only a small share of corporate AI leaders believe their data infrastructure is ready for widespread AI use, according to new research from data connectivity specialist CData.

The study found that only 6% of enterprise AI leaders consider their data infrastructure fully ready for AI projects. The findings suggest that data foundations now pose a greater barrier to AI progress than the models themselves.

The report, titled The State of AI Data Connectivity: 2026 Outlook, draws on a survey of more than 200 data and AI leaders. The respondents work at software providers and large organisations across multiple industries.

The research links higher levels of AI maturity with more advanced data infrastructure. It also identifies gaps in data connectivity, shared context, and control as common reasons for stalled AI projects.

Data, not models

The survey describes a clear divide between organisations with mature AI programmes and those that are struggling. It states that 60% of companies at the highest level of AI maturity have also invested in advanced data infrastructure. By contrast, 53% of organisations that struggle with AI implementations report immature data systems.

CData said this divide is translating into delays, higher costs, and weaker competitive positions for laggards.

Amit Sharma, CEO and Co-founder of CData, said data has replaced models as the main bottleneck. "The era of AI being constrained by models is over. Today, AI is constrained by data," said Amit Sharma, CEO and Co-founder, CData. "The organizations winning with AI aren't the ones with the best algorithms; they're the ones with connected, contextual, and semantically consistent data infrastructure."

Time on 'data plumbing'

The study reports that AI teams are spending a large share of their time on basic data work rather than experimentation and deployment. It says 71% of AI teams spend more than a quarter of their time on "data plumbing". This includes the manual work of connecting systems, preparing data, and fixing pipelines.

Demand for real-time data access is also rising. The survey finds that 46% of organisations require real-time access to six or more data sources for a single AI use case. These use cases include AI assistants, decision-support tools, and embedded AI in software products.

Every respondent agreed that real-time data is essential for AI agents. At the same time, 20% still lack real-time integration across their systems. This gap creates delays and inconsistency in AI outputs.

Integration load rising

The report highlights a sharper integration burden for AI-native software providers. These firms build products that rely on AI at the core. The survey states that 46% of AI-native providers need at least 26 external integrations. The comparable figure for traditional software companies is 15%.

This higher level of integration increases technical complexity for AI-focused vendors. It also raises the importance of shared data definitions and consistent semantics across systems.

The research finds that infrastructure maturity is a defining feature of advanced AI organisations. All high-AI-maturity organisations in the sample have built central integration layers that hold semantically consistent data. Among low-maturity organisations, 80% have not begun work on such layers.

Shift in investment

The survey suggests that AI investment strategies are changing. Only 9% of organisations now rank AI model development as their top spending priority. Instead, 83% say they are investing in, or planning, centralised data access layers that maintain consistent meaning across data sources.

These layers sit between source systems and AI models. They provide a single route for models and applications to access data in a uniform way.

Sharma said the pattern reflects a broader shift in thinking among executives. "Organisations are realizing that AI success isn't determined by the sophistication of their models. It's determined by the maturity of their data infrastructure," said Sharma. "The companies gaining real value from AI are the ones that invested early in connected, real-time data access. Those that haven't will find themselves at a significant competitive disadvantage."

Benchmarks and context

The report sets benchmarks for enterprises and software providers in two areas. One area is enterprise AI adoption, including how data infrastructure gaps affect AI outcomes and what separates leading adopters from slower movers. The other area is product AI strategy, including how software companies embed AI in their products and manage growing integration demands.

CData's research references an MIT study on business AI adoption published in 2025. That work analysed the gap between early adopters of generative AI and slower-moving organisations.

The new survey suggests the next phase of AI deployment will depend less on novel models and more on stable data foundations. It points to increasing investment in central data layers as organisations prepare for more complex AI use at scale.