cd /news/ai-infrastructure/ai-workloads-hit-by-power-quality-is… · home › topics › ai-infrastructure › article
[ARTICLE · art-172465] src=datacenterknowledge.com ↗ pub= topic=ai-infrastructure verified=true sentiment=↓ negative

AI Workloads Hit by Power-Quality Issues at Data Centers

Nearly a third of 500 US data center operators and professionals surveyed by Fluke said power-quality issues have reduced their ability to deploy or operate high-density AI and GPU workloads, and only 36% said their electrical infrastructure is fully prepared for higher-density AI workloads. Fluke senior manager and technical fellow Michael Stuart said "the biggest risk is whether a data center's internal infrastructure can reliably distribute and support that power at scale," with 79% reporting power-quality problems contributed to delays, disruptions, unplanned downtime or near-misses in the past year and 78% lacking comprehensive real-time visibility across critical electrical systems.

recommendation

Measure power quality as close as possible to GPU loads along the distribution chain and fix electrical or cooling shortfalls before deploying high-density racks, since 79% of surveyed operators reported power-quality-driven delays or downtime.

more in playbook: Running AI in production
by read3 min views1 publishedNov 19, 2026
AI Workloads Hit by Power-Quality Issues at Data Centers
Image: Datacenterknowledge (auto-discovered)

Power quality, not just grid access, is limiting AI growth at US data centers, exposing weaknesses in internal infrastructure and monitoring.

Electrical problems are constraining AI deployments at US data centers, with nearly a third of respondents to a Fluke survey saying power-quality issues have reduced their ability to deploy or operate high-density AI and GPU workloads.

The survey, which polled 500 data center operators and professionals across the United States, suggests facilities are facing internal infrastructure constraints alongside the broader struggle to secure electricity from utilities. Just 36% of respondents said their electrical infrastructure is fully prepared for higher-density AI workloads, while 74% said insufficient power and grid capacity are moderately or severely limiting expansion.

“The biggest risk is whether a data center’s internal infrastructure can reliably distribute and support that power at scale,” Michael Stuart, senior manager and technical fellow at Fluke, told Data Center Knowledge.

Power, Cooling, and Cabling Failures Create Risks #

About three-quarters of respondents (76%) reported experiencing electrical or power-system issues. Separately, 79% said power-quality problems had contributed over the past year to delays, service disruptions, unplanned downtime, or serious near-misses.

The figure groups operational failures together with incidents that did not result in an outage. It does not mean nearly four in five respondents experienced downtime.

Stuart said weaknesses often emerge where power, cooling, and cabling systems intersect. Problems can include harmonics, voltage sags and swells, thermal hotspots, deteriorating cooling systems, and infrastructure connections that have not been adequately documented or verified.

“Since each data center is so complex and diverse, determining weaknesses within facilities should be done on a case-by-case basis,” he said.

Those conditions can limit the computing capacity a facility can support even after an operator has secured sufficient electricity.

“If electrical, mechanical, and network systems aren’t performing to specification, operators may be unable to fully utilize the capacity they’ve already secured from the grid,” he added.

Testing and Monitoring Before GPU Rollouts #

For operators preparing to install high-density GPU racks, Stuart recommends measuring power quality as close as possible to the GPU loads along the electrical distribution chain. The appropriate measurement points will vary by facility and may change during commissioning as actual load conditions become clearer. Electrical distribution and cooling also need to be evaluated together to establish whether they can sustain the intended workloads.

“If measurements reveal that either system cannot reliably support the intended GPU workloads, operators should address those issues before proceeding with deployment,” Stuart said.

The survey also identified gaps in operators’ ability to detect developing problems. Some 78% of respondents said they lacked comprehensive, continuous, real-time visibility across critical electrical systems. Another 15% said electrical problems were identified only after performance degradation, equipment failure, or downtime had already occurred.

Stuart said operators should assess existing monitoring coverage to identify where additional visibility would most improve reliability. Potential areas include power distribution, backup power, batteries, and switchgear, although priorities will depend on each facility’s infrastructure and current monitoring capabilities.

Labor Shortages Add Pressure to Maintenance #

Recruiting qualified electrical technicians has become harder for 57% of respondents, adding pressure to the teams responsible for installing, testing, and maintaining infrastructure.

That hiring difficulty has not uniformly translated into slower expansion. Some 65% of respondents said skills shortages had not delayed infrastructure upgrades or new capacity.

Still, Stuart said workforce constraints can contribute to localized delays, inconsistent installation practices, and additional work to correct problems or verify quality.

“Not every problem is visible during an initial inspection; some only surface once electrical systems are operational,” he said.

Stuart said qualified electricians should perform and validate the work, supported by testing and diagnostic tools to identify faults that may emerge after installation and help ensure long-term reliability.

── more in #ai-infrastructure 4 stories · sorted by recency
── more on @fluke 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ai-workloads-hit-by-…] indexed:0 read:3min 2026-11-19 · —