Colt: digital complexity costs enterprises €1m in stalled AI growth

Colt Technology Services has published research showing that complex digital infrastructure is costing large enterprises nearly €1 million a year in combined delays and stalled innovation, with AI adoption emerging as the most visible casualty.

The study, conducted by Coleman Parkes in early 2026, surveyed 600 senior leaders, including CEOs, CIOs and IT directors, at organisations averaging more than 26,000 employees across the UK, France, Germany, the Netherlands and Japan. It found that respondents estimated an average of €401,400 in value at stake annually from delays and rework, with a further €508,000 tied up in innovation initiatives that failed to progress, at least partly because of infrastructure constraints.

AI ambitions blocked at the network layer

The AI implications are stark. Six in ten respondents (60%) said digital infrastructure complexity is a direct barrier to capturing the full benefits of AI at scale, while two thirds (66%) believe their businesses have already missed meaningful AI opportunities as a result. Critically, 91% reported delays in embracing emerging technologies such as agentic AI, a figure that rose to 100% among Japanese respondents.

The sources of that complexity are familiar: 57% of respondents cited managing multiple vendors, 48% pointed to legacy systems, and 36% flagged security and compliance requirements. The result is a layered, fragile architecture that was not designed to carry the demands of modern AI workloads. As Laura Farina, EVP of Enterprise Sales at Colt, put it: "Many complex enterprise networks have been built over time like LEGO bricks from different generations, compatible in theory, but not designed to form a clean, stable structure together."

Beyond AI, the operational drag is broad. The research found the equivalent of seven weeks of delays across the past 12 months. Some 93% of respondents said complexity had slowed M&A integration, 84% cited delays to market expansion, and 83% pointed to product launches held back by infrastructure friction.

The convergence read-across: infrastructure as a strategic bottleneck

For the cross-sector strategist, this research carries implications well beyond IT procurement. Enterprise digital infrastructure has quietly become the rate-limiting factor for the AI economy, sitting at the intersection of network investment, cloud strategy and competitive advantage. As agentic AI workflows, which require low-latency, highly reliable connectivity across distributed systems, move from pilot to production, the gap between firms with clean network architectures and those carrying legacy debt will widen materially.

The geographic spread of the data is also instructive. UK businesses reported the highest value at stake from delays (€450,400) and the largest pool of stalled innovation value (€743,600), ahead of France (€753,000 in stalled innovation) and Germany. Japan, despite the 100% rate of agentic AI delays, cited lower absolute financial exposure, likely reflecting differing enterprise scale and digital investment cycles across the sample.

For capital allocators and digital infrastructure investors, the research points to a structural investment thesis: the demand signal for network simplification, managed services and next-generation B2B connectivity is not softening. Colt, which operates Europe's largest B2B network spanning 40-plus countries and 32,000 enterprise buildings, has an evident commercial interest in amplifying this finding. That context is worth holding. Nonetheless, the underlying dynamic, that AI deployment ambitions are consistently outrunning the legacy infrastructure on which they are being placed, is corroborated by enterprise technology research from multiple independent sources.

The broader implication for boards and macro investors is that AI ROI timelines cannot be modelled in isolation from network modernisation spend. Firms that treat infrastructure simplification as a background IT task rather than a strategic enabler risk compounding the very delays this research quantifies.