Cabling bottlenecks leave a third of data centers with wasted AI hardware
Onnec’s survey of 300 data center decision-makers in the UK, Ireland and the Nordics finds 78% cut corners on cabling to deploy AI faster, and 44% have paid for it in delays or remediation
- Sapio ran the fieldwork online for Onnec between 27 May and 12 June 2026, with 300 respondents: 150 in the UK, 50 in Ireland and 100 in the Nordics.
- Cabling is becoming a critical bottleneck to supporting AI workloads, say 77% of operators, and 41% say it always or frequently delays AI capacity projects.
- AI requirements change so fast that new data centers need upgrades soon after go-live, according to 83%.
- A health and safety incident or near miss during a rushed cabling installation was reported by 32%.
Businesses renting AI capacity in new halls should ask how the network linking the GPUs was designed and tested before signing. Poor cabling is likely to show up as slower training and inference, late handover or outages after go-live, and the cost of fixing it can reach tenants through delays. Contracts should tie service levels to delivered network performance, and buyers should ask whether the hall can scale without significant re-cabling.
“Everyone is focused on securing power and GPUs for AI, but the network connecting that hardware together is too often finalised only after those decisions have already been made.”Matt Salter, global head of data centers, Onnec
Onnec, a structured cabling and network integration specialist, published a survey on 10 September in which 31% of data center operators said the network around their AI hardware had failed to keep up, leaving the expensive kit it links unable to do its job. The research, which Onnec commissioned and Sapio ran between 27 May and 12 June, questioned 300 senior decision-makers who design, manage or operate facilities: 150 in the UK, 50 in Ireland and 100 across the Nordics. Onnec sells the cabling work it says operators are getting wrong, so its report, The AI Cabling Time Bomb, carries that caveat.
The demand side explains the rush. Two-thirds (65%) of respondents saw demand for AI capacity rise over the previous 12 months, and those reporting growth put the average increase at 44%. Under that pressure, 78% admitted compromising on cabling quality or specification to get AI infrastructure in faster. Of those, 41% put immediate needs ahead of longer-term bandwidth, 41% spent too little time on pathway and capacity design and 37% used less experienced contractors to move faster.
The bill arrives after go-live
The consequences show up once the facility is live. Some 44% of operators have paid for remediation work or suffered deployment delays because upfront cabling planning was thin, 33% could not scale an AI environment without significant extra cabling work and 27% have had unplanned downtime caused by cabling failures. TechRadar Pro, which reported the findings on 15 September, framed them as a brake on the UK’s data center build-out.
Sources
- Onnec, “New Onnec Report: AI Hardware Investment at Risk as Network Fails to Keep Pace”, press release, 10 September 2026. https://www.onnecgroup.com/2026/09/10/new-onnec-report-ai-hardware-investment-at-risk-as-network-fails-to-keep-pace/
- TechRadar Pro, “The UK’s £100 billion data center boom is suffering from a major bottleneck that puts capacity and performance at risk”, Benedict Collins, 15 September 2026. https://www.techradar.com/pro/the-uks-gbp100-billion-data-center-boom-is-suffering-from-a-major-bottleneck-that-puts-capacity-and-performance-at-risk
