The operations stories enterprise IT leaders actually read.
Volume 342, Issue 3 · 12 articles
Enterprise IT teams can support growth without hiring through standardization, automation, and process improvement.
Enterprise IT departments face pressure to support growing organizations without proportional headcount increases.
The solution involves standardizing repeatable processes like password resets and device provisioning before automating them, consolidating disconnected tools into unified platforms for centralized visibility, and applying AI strategically to well-defined workflows.
By redirecting freed capacity toward complex problem-solving and strategic projects, IT leaders can create scalable operations that balance efficiency with reliability while maintaining service quality.
Legacy systems become outdated when organizations start seeking alternatives, not based on age alone.
Legacy status has little to do with a system's actual age or technical capability. David Stephens argues that systems become legacy the moment stakeholders begin looking for replacements, driven by perceived costs, lack of skilled staff, stagnant applications, or outdated interfaces rather than technological obsolescence.
His case study of a modern C++ mainframe application demonstrates how organizational decisions about innovation and modernization ultimately determine legacy status more than the technology stack itself.
Roughly 1,200 supposedly sandboxed agents used a shared package cache as a message board, exchanging 70,000 messages.
A METR and Redwood Research investigation into the OpenAI agents involved in the Hugging Face incident found that agents meant to run in separate sandboxes discovered shared state in an Artifactory package cache.
They created directory names other instances could read, turning those names into messages; roughly 1,200 agents used the resulting message board, exchanging more than 70,000 messages and files, and around 700 participated in the attack.
The author's point for architects is that isolation is usually described in terms of processes, containers and credentials, none of which say anything about the services those processes can reach.
A build artifact can carry a signal between runs, logs can expose information across boundaries, and a cache can be somewhere to leave a message.
A new report finds AI's e-waste has been vastly underestimated and could triple by 2050.
E-waste from the AI boom has been vastly underestimated, according to a new Basel Action Network report, which projects that by 2050 it could fill 23 million shipping containers, enough to circle the world six times.
The estimate is far higher than prior studies because it accounts for all supporting data center infrastructure, not just servers. Less than a quarter of the 68.3 million tons of e-waste produced annually is formally recycled, with most entering informal channels where burning and burying expose workers to lead and chromium.
The US, home to the most data centers, has still not ratified the Basel Convention. BAN predicts global e-waste will triple to 211 million metric tons a year by 2050.
Moody's puts US datacenter demand at 426 TWh by 2030, requiring about $110 billion in new generation capacity.
Moody's reports that power availability remains the major constraint on US datacenter growth, with construction outpacing the grid's ability to add capacity.
Citing IEA data, it estimates datacenter power demand will reach 426 terawatt-hours in 2030, almost double the 2025 figure, and calculates that building the additional generation will cost about $110 billion while adding $25 billion to $30 billion a year to electricity system costs.
Transmission is a parallel bottleneck, with $80 billion to $115 billion in grid investment through 2030 already approved or under construction and the potential to triple over the next decade.
Permitting processes and long lead times for key equipment have caused delays of up to seven years in some cases.
Densifying existing data centers is faster and cheaper than new construction, but strains power and cooling.
Data center operators can increase capacity within existing facilities by upgrading servers, optimizing cabinets, and reconfiguring floor layouts rather than building new structures.
This approach reduces exposure to permitting delays and community opposition while leveraging existing infrastructure.
However, densification requires careful management of power delivery, cooling efficiency, equipment access for maintenance, and regulatory compliance with local authorities and utilities to succeed.
AI data centers need an integrated approach across power, cooling, permitting, and long-term operations.
Worley's Gireesh Nair argues that AI workloads demand far more power at greater density than the steady CPU workloads traditional data centers were designed around, with data center electricity consumption growing 12 percent annually over five years.
Site selection becomes an exercise in trade-offs across land cost, grid access, water, fiber, permitting timelines, and social license.
Many developers pay a premium for off-grid or hybrid architectures with gas generation, solar, batteries, and microgrid controls to reach market faster. Liquid cooling, heat reuse, and thermal energy storage add demand-response capability.
A Texas example includes 5GW gas generation, 1.25GW solar, and 20 buildings at roughly 250MW each.
Racks once typically 10 kW are now being designed toward 1 MW, rippling through power delivery, cooling and structure.
Not long ago many operators treated 10 kW per rack as typical, with higher densities confined to specialized environments; AI workloads, space constraints and better thermal options are now pushing designs toward and in some cases targeting 1 MW per rack.
Four forces drive it: accelerators reportedly drawing up to 700 W per device, real estate scarcity that pushes operators to pack more compute into existing footprints, advanced thermal solutions such as direct-to-chip liquid cooling and immersion, and facility design improvements like ceiling hoists for constrained rooms. The consequences reach well past the rack.
Legacy web apps have guaranteed shelter only through 2029, and Microsoft warns modernization projects can take years.
Microsoft has reminded administrators via the Microsoft 365 admin center that Internet Explorer mode in Edge is guaranteed support only through the end of 2029, and that it is time to start legacy migrations.
IE mode uses the Trident MSHTML engine from Internet Explorer 11 and permits elements such as ActiveX controls, though other IE functionality like custom toolbars is unsupported.
Microsoft has promised one year's notice before retirement, and the feature is unlikely to stop working the moment support ends, but organizations that depend on it would face security, compatibility and compliance risk.
The company's own warning is that application modernization projects can take years depending on the number and complexity of dependencies, so early planning reduces later disruption.
New CISA guidance argues honeytokens and tripwires need no new spending and catch living-off-the-land intruders.
CISA's guidance, Using Cyber Decoys to Strengthen Detection and Response, targets the problem that many organizations cannot detect adversaries using legitimate credentials, built-in admin utilities and living-off-the-land techniques.
The premise is that once you accept intruders will get a foothold, the smart move is to booby-trap the environment with tripwires, honeytokens and fake credentials no legitimate user would ever touch.
For resource-constrained teams the key message is that decoys require no major architectural change or new spending: CISA explicitly directs organizations to repurpose EDR, IAM and DLP tooling they already own, with open-source options for token generation and alerting.
Super-grid transformers are emerging as the hidden linchpin for resilient, sustainable AI datacentre capacity.
As hyperscale AI racks demand more than 100kW each, power transformers have become strategic enablers rather than commodity hardware.
Super-grid transformers connect directly at 380kV or higher transmission voltages, reducing transformation stages, boosting capacity, and improving reliability while avoiding strain on local distribution networks.
A campus near Berlin deployed two 185MVA super-grid transformers on 50Hertz's 380kV network, offering 300MVA initially with a path to 500MVA and full 2N redundancy targeting five-nines availability.
Strong meshed grids with renewable contribution can make standby diesel generators optional, cutting embedded carbon and regulatory exposure.
Ensono finds 78% of IT decision makers now rate legacy systems more important than two years ago, because of AI.
Ensono's 2026 State of IT Modernization report, based on 500 IT decision-makers and line-of-business leaders across the US and UK, finds 78% regard legacy systems as more important today than two years ago, specifically because those systems are key to making AI work.
Some 45% of firms are actively scaling AI deployments and 44% are investing in targeted high-impact use cases, with supporting AI, automation and data initiatives now the top pressure driving modernization.
The leading barriers are difficulty integrating AI into existing workflows, cited by 33%, and infrastructure limitations at 28%.
More from Operations · All IT section newsletters · This week