News & Events

New Power Gear Is Years Away. Maintain What You Have Like It Has to Last.

September 17, 2026
By William Newby

September 17, 2026

Operators waiting on new switchgear and transformers have to keep their existing electrical equipment reliable in the meantime. The 2026 edition of NFPA 70B spells out how, including a new cybersecurity requirement.

By William “Bill” Newby, Co-founder & COO, BNS, and CEO, BNS Power

Executive summary: Data center and critical facility operators cannot count on new electrical equipment to arrive soon. Switchgear takes about a year and substation transformers take more than three. Until replacements land, the UPS systems, transfer switches, breakers, and switchgear already installed have to carry the load, and failures in exactly that equipment remain the leading cause of serious data center outages. The point of this article is simple: service, test, and where needed refurbish or retrofit the equipment you already own, on a documented program, so it lasts until the new gear shows up. NFPA 70B, a mandatory standard since its 2023 edition, requires that program. The 2026 edition adds rules for transfer switches and requires a cybersecurity risk assessment of the networked relays, meters, and controls inside the electrical system. Below: what changed, why it matters for AI facilities, and how to sequence maintain, refurbish, retrofit, and replace.

A substation transformer ordered this month shows up sometime in 2029. Wood Mackenzie puts lead times for that class of transformer above 160 weeks this year, up from roughly 140 weeks in 2023. Switchgear is closer to a year. Prices moved with the wait: circuit breakers are up 47% since 2021 and medium-voltage switchgear is up 50%.

Run those numbers against a purchase order signed today. New switchgear lands in late 2027. The transformer lands in 2029. Every megawatt a facility serves between now and then flows through equipment that is already bolted to the floor, and most of that equipment was bought for a lighter load, a gentler duty cycle, and a maintenance budget somebody trimmed years ago.

If that equipment fails before its replacement arrives, there is no quick fix to order. Its condition is now a capacity question as much as a safety one.

An airplane older than most of its crew

I spent twelve years as a crewmember and instructor on U.S. and NATO E-3 AWACS aircraft. The E-3 is a Boeing 707 airframe. The Air Force took delivery of its first one in March 1977, production ended in 1992, and NATO is running a lifetime extension program to keep its E-3 fleet flying toward 2035.

Airplanes do not stay in service for half a century by luck. They stay because the maintenance program is written down, scheduled, inspected, and treated as part of the mission rather than a line item to cut when the budget gets tight. Nobody on a flight line waits for a part to fail to find out what condition it was in.

Commercial electrical infrastructure has mostly run the other way: test it when there is an outage window, fix it when it fails, and buy new when the budget allows. That approach held up when replacement gear was a routine order. It does not hold up when the replacement sits on a multi-year order book.

Power is still where facilities go down

Uptime Institute’s 2026 Annual Outage Analysis says power remains the leading cause of impactful data center outages, and that failures involving UPS systems, transfer switches, and generators dominate. 57% of operators surveyed said their most recent major outage cost more than $100,000, and for the second year in a row one in five put it above $1 million. Uptime also found that failure to follow established procedures is the leading driver of outages caused by human error.

None of those failure points is exotic. A transfer switch that has not been exercised under load. A UPS battery string nobody has trended. A breaker that has not been trip-tested since commissioning. Each one is a maintenance finding, and each one costs far less to find on a work order than in the middle of an event.

AI workloads shrink the margin. GPU halls run dense and at high utilization for long stretches, which leaves less headroom for a connection that is loosening or a protective device that has drifted out of calibration.

NFPA 70B stopped being optional

For most of its history, NFPA 70B was a recommended practice. The 2023 edition made it a standard, and “should” became “shall.” That handed authorities having jurisdiction, insurers, and owners a clear yardstick for judging whether a maintenance program is real.

The standard requires a documented electrical maintenance program (EMP) with a person accountable for it, an equipment survey, written procedures, inspection and testing plans, records, corrective action, and periodic review. Each piece of equipment gets a condition rating of 1, 2, or 3 based on its physical condition, its criticality, and its operating environment, and the worst of the three sets the interval. Infrared thermography, for example, runs every 12 months for Condition 1 and 2 equipment and every 6 months for Condition 3.

In a facility where an outage costs seven figures, criticality alone can push a lot of equipment into the shorter intervals.

The 2026 edition puts cybersecurity inside the maintenance program

The 2026 edition of NFPA 70B goes further in three places that matter for data centers:

  • Transfer switches: a dedicated chapter for the same equipment Uptime keeps flagging in its outage reports.
  • Continuous thermal monitoring: permanently installed temperature sensors are now recognized as a way to perform thermal inspection, alongside or in place of the periodic infrared scan.
  • OT cybersecurity: an operational technology cybersecurity risk assessment is now a required element of the EMP where control systems reach beyond isolated local interfaces.

That last one is where my other career comes in. I spent thirteen years as an Air Force communications and cyber operations officer, including time as a senior network defense officer. What I learned there applies directly to a switchgear room. Every microprocessor relay, networked power meter, PLC-based transfer control, and remote monitoring gateway is a device on a network. Most of them were installed by electrical crews, commissioned by electrical engineers, and never reviewed by anyone responsible for network defense.

The electrical side and the security side rarely sit in the same meeting. Relay settings and firmware can end up managed from a laptop that also handles email. Monitoring gateways get opened to a vendor’s cloud because it is convenient. The continuous thermal sensors the 2026 edition now recognizes add more connected devices to the same network. The standard now asks the maintenance program to account for all of it, and that puts both teams in the same plan.

Maintain, refurbish, retrofit, then replace

When new gear sits on a multi-year order book, the order of operations changes.

Maintain. Build the EMP the standard requires, rate the equipment honestly, and test to the intervals. Know what condition the gear is in before the facility needs it at full load.

Refurbish. Plenty of existing switchgear has a sound structure and bus. Reconditioning breakers, replacing worn components, and retesting can return it to reliable service on a schedule set by an outage window rather than a factory queue.

Retrofit. Replacing legacy trip units and protective relays with modern devices, or adding monitoring, extends service life and improves protection. Each of those new devices also belongs in the OT risk assessment.

Replace. Order the new transformers and switchgear now, because the lead times are real, and plan on the existing equipment carrying the load until they arrive.

Where BNS Power fits

BNS Power brings together two companies that have done this work for decades. Cadick Corporation, founded in 1986, has provided electrical maintenance, engineering studies, and safety training for forty years, and John Cadick co-authored The Electrical Safety Handbook. Our Cadick team has maintained, repaired, and refurbished the electrical systems aboard USCGC HEALY, the largest icebreaker in the U.S. fleet, for thirty years. Carolina Power Systems has built switchgear for the fleet in Sumter, South Carolina since 1994, and our Sumter plant manufactures and refurbishes switchboards, switchgear, motor controllers, and control and relay panels.

I have a direct interest here. BNS Power sells maintenance, engineering studies, training, refurbishment, and new equipment, so more of any of it is good for us. Weigh what I have written with that in mind. It is also why I can tell you that the difference between maintained and neglected equipment shows up long before it shows up in an outage report.

On an E-3, the jet flew because the maintenance got done. The megawatts a facility can count on in 2027 are running through equipment it owns today, and whether that equipment is ready depends on the program behind it.

Sources