The operations stories enterprise IT leaders actually read.
Volume 343, Issue 2 · 10 articles
US data center land purchases reached about $6B in 1H 2026, up 79%, as power access drives land value.
Avison Young reports that US data center land purchases hit roughly $6 billion in the first half of 2026, up 79% year over year, as hyperscalers, neoclouds and investors compete for AI campus sites.
In Northern Virginia, entitled land without power sells for about $4 million an acre, versus at least $6 million for powered land.
An Altus Group analysis found that committed utility load was the strongest predictor of price, with 230-kV or 500-kV sites trading at about 2.4 times the price of lower-voltage sites.
A truly power-ready site generally needs electrical infrastructure deliverable within about 18 months.
Smarter sensing, motor control, efficient power devices and edge AI let liquid cooling respond in real time to AI rack loads.
Texas Instruments' Patrick Zeng argues that rising AI densities require cooling systems that sense and adapt rather than run at fixed settings.
Precise flow sensing, such as ultrasonic sensors, can guide linear valves at each tray, while better leak detection is replacing resistive ropes.
GaN power devices can raise pump-system efficiency by up to 2%, fans are moving from 12V to 48V, and pumps will need to support 800V HVDC distribution.
Edge AI "sensor fusion" across temperature, humidity, pressure and flow data enables faster leak detection and predictive maintenance across any cooling architecture.
Policymakers may start asking whether the services data centers run justify their power, water and grid costs.
Regulation and public criticism of data centers still focus on physical footprint: power, water, land and grid impact.
Discussions at Informa's Data Center Europe event in Vienna suggested that policymakers may eventually also weigh the societal value of the workloads inside.
Video streaming, social media and related applications account for about 65% of internet volume, and Goldman Sachs projects that AI's share of the data center market will double to 30% within two years.
The article notes that this is still an emerging policy question, but social media rules show governments are increasingly willing to intervene in digital activity.
AI-era data centers need electricians, HVAC and construction skills, so IT and facilities must jointly close workforce gaps.
AI and other energy-intensive workloads mean data center managers now need electrical engineers, HVAC technicians, physical security specialists and construction workers, not just IT staff.
As sustainability, heavier workloads and edge sites blur the line between IT and facilities, the article says both groups should plan, budget and support the whole data center together.
Its recommendations include proactive service agreements with providers, preventive maintenance and basic facility health checks such as UPS and switchgear inspections.
It also recommends remote monitoring to stretch scarce skilled labor.
October's Windows update now blocks NTLMv1-derived credentials by default, which can break Wi-Fi, Ethernet and VPN logons.
Before the October 2026 Windows security update, NTLMv1-derived credentials for single sign-on were allowed with warning events.
Now, if the BlockNtlmSSO registry value is unset and Credential Guard is not enabled, Windows uses Enforce mode and blocks them, potentially breaking Wi-Fi, Ethernet and VPN authentication.
Remnants often surface with MS-CHAPv2-based connections such as PEAP-MSCHAPv2 in domain-joined environments.
Admins can set Audit mode as a temporary exception, but Microsoft says the change is by design as it finalizes NTLMv1 disablement, so migration should not wait.
JLL says 3% to 6% annual construction inflation is the new baseline, squeezing projects without power or anchor tenants.
JLL's Global Construction Costs Index 2026 says annual construction cost escalation of 3% to 6% is now the baseline, up from 2% to 3% before 2020, with a further 3.1% to 4.2% rise in tender prices forecast for 2027.
Strong AI demand does not protect projects lacking power certainty, anchor tenants and credible delivery timelines.
JLL's Matthew Giddings recommends reassessing projects when costs erode returns or utility schedules slip, and using phased construction to match spending with confirmed power and demand.
Electricians, high-voltage specialists, pipefitters, cooling technicians and commissioning teams are the tightest labor constraints.
NERC stakeholders approved reliability standards for computational loads of 50 MW or more on 100 kV or higher lines.
NERC stakeholders approved new reliability standards for AI data centers, crypto mining sites and other large computational loads, which now go to the NERC board and then to FERC.
As written, they apply to loads of 50 MW or more connected at 100 kV or higher, and they follow a May Level 3 alert issued after data centers unexpectedly dropped load or oscillated demand, threatening regional grid stability.
Requirements cover interconnection studies and modeling, operational data and communications, and load protection coordination and disturbance monitoring.
NERC forecasts that summer peak demand will grow 224 GW over 10 years.
Recovering idle data center capacity starts with understanding why it is unused: overprovisioning, misallocation or constraints.
Two facilities can show similar utilization for very different reasons, the article explains, so identifying unused capacity is only the first step.
Monitoring tools for networks, AI workloads, cloud and applications are separate from physical capacity tracking, which hides problems such as empty rack space in a row that has hit its power limit.
Operators should map workload requirements across compute, memory, storage, networking, power and cooling, then classify low utilization as overprovisioned, poorly allocated or limited by the application.
They must also distinguish truly idle capacity from reserves kept for failover, maintenance or peaks.
Layered site and supply chain controls protect data centers from theft, sabotage, counterfeit gear and planted devices during the build.
The article argues that a data center's period of greatest physical exposure is often construction, when sites are porous and contractor populations are large and fluid.
Threats include theft of copper, cabling and generators, vandalism tied to community opposition, and rogue access points or hidden sensors planted for later backdoor access.
Supply chain tampering with counterfeit components and drone reconnaissance are also risks.
Recommended controls include strong perimeters with staffed gates, professional surveillance and lighting, contractor and delivery vetting, and supply chain verification, all designed to reduce risk without slowing construction.
Retired laptops, servers and drives remain a breach risk until data-bearing parts are sanitized under a documented chain of custody.
Enterprises invest heavily in protecting systems while they run but often neglect hardware retirement.
Decommissioned laptops, servers, switches and storage arrays sit in unsecured rooms or loading docks, and formatted or seemingly dead drives can still yield data to forensic tools.
The article explains how an enterprise ITAD program closes these gaps with protocols that begin when hardware leaves the network and an unbroken chain of custody across multiple sites using asset tags, GPS-tracked vehicles, tamper-proof seals and inventory reconciliation.
The program also requires verified sanitization or destruction and full documentation for auditors.
More from Operations · All IT section newsletters · This week