Most industries treat energized electrical work as the exception. Data centers are built around it. Concurrent maintainability — the ability to service any component without dropping the critical load — is a design goal, a marketing claim, and an operational religion. The practical consequence is that data center electricians and critical facility technicians spend more of their working lives near energized equipment than almost anyone else in the trade.
That is exactly why the 2027 edition of NFPA 70E, in effect since May 6, 2026, lands harder in mission-critical facilities than almost anywhere else. Several of its most significant changes — a mandatory additional person for permitted energized work, emergency response planning that has to exist before the work starts, tighter expectations for job briefings, and a sharper focus on managing change — read as if they were written with a data hall in mind. Here is the deep dive on what actually changes, and why mature data center operators may be closer to compliance than they think.
The Second Person Is Now a Staffing Requirement
The headline change: new section 130.2(A)(2) requires that whenever an energized electrical work permit is required under 130.3(A), at least one additional person must be present — in the vicinity of the work, but positioned outside the limited approach boundary or the arc flash boundary, whichever is greater. That person must meet the emergency response training requirements of 110.4(C)(1). Their job is not to hold the flashlight. It is to release a worker from contact, start CPR, and get help moving if the worst happens.
For operators who already run two-person rules on live work — and many large colocation and hyperscale platforms do — the adjustment is real but manageable: reposition the role, train it to the emergency response requirements, and document it on the permit. For everyone else, this is a new line item. Every permitted energized task now costs two qualified people instead of one, which changes the economics of doing marginal tasks energized at all. That pressure is intentional.
An Emergency Response Plan That Exists Before the Work Does
The second-person requirement quietly forces a bigger question: respond how, exactly? Pointing at the wall-mounted AED is not an emergency response plan. The plan for an energized task needs to be settled before anyone opens a panel: who on the crew is trained in contact release, where the rescue hook and AED physically are, who calls for help, and what the inbound path for emergency responders actually looks like.
That last item deserves special attention in a data center, because your physical security program works against your emergency response time. Badge readers, mantraps, security escorts, and a NOC that has to authorize entry are all fine — until paramedics are standing at the gate while a worker is in cardiac arrest on Row 14. An emergency response plan worth the name answers that scenario in advance: who meets EMS at the door, who overrides the access sequence, which path the gurney takes through the white space. Write it down, brief it before the task, and rehearse it the way you rehearse a generator failure.
Pre-Job Briefings: The MOP Read-Through Is Not the Whole Job
Data centers already brief. Before a critical switching sequence, someone walks the method of procedure step by step with everyone involved. NFPA 70E's job briefing requirement overlaps with that discipline but does not duplicate it. The MOP briefing exists to protect the load. The 70E job briefing exists to protect the person.
A compliant briefing covers the hazards of the specific task: the incident energy at the point of work, the PPE that number demands, where the approach and arc flash boundaries fall, how the energy sources are controlled, and the emergency plan — including where the additional person will stand and what they will do. The good news: bolting those items onto an existing MOP briefing is a template change and a habit, not a new meeting.
Management of Change: The Study Your Last Upgrade Invalidated
An arc flash label is a snapshot of the electrical system on the day the study was run. Data centers change faster than almost any facility type: UPS modules added, breaker settings adjusted during commissioning, new PDUs energized, busway taps landed for higher-density rows, generator paralleling schemes revised. Any of those can change available fault current or clearing times — which means the incident energy on the label may simply no longer be true.
NFPA 70E expects the incident energy analysis to be reviewed when the electrical system changes, not just on a five-year calendar. The mechanism data centers already have is change management: every one of those modifications went through a change ticket. The fix is wiring one question into that workflow: does this change affect available fault current, protective device settings, or clearing time? If yes, the change does not close until the study owner has reviewed it and the labels reflect reality.
Dry Runs: The Habit That Was Ahead of the Standard
The dry walk-through is one of the data center industry's best inventions. Before a critical MOP executes, crews rehearse it — walking the route, touching (without operating) each device in sequence, confirming field labels match the procedure, catching the step that says "open breaker 2B" when the nameplate reads "2B-ALT." Mission-critical operations programs treat the dry run as mandatory for the same reason 70E treats job safety planning as mandatory: most incidents are planning failures that surfaced at the worst possible moment.
The 2027-era upgrade is to let the safety choreography ride along on the rehearsal you already do. On the dry run, also confirm: where PPE gets donned, where the boundary tape goes, where the additional person stands, where absence of voltage will be verified and with which instrument, and which way a worker in trouble gets extracted. Five extra minutes on a walk-through, and the safety plan has been rehearsed instead of merely written.
Where Data Centers Are Ahead — and Where 2027 Still Bites
Line the two worlds up and the overlap is striking:
- MOP dry runs already do most of what job safety planning asks. Extend them to rehearse the safety steps, not just the switching steps.
- EOP libraries already rehearse failure response. Add an electrical-injury EOP: contact release, CPR/AED, and the EMS access path through security.
- Change management already captures every modification. Add the fault-current and clearing-time trigger question to close the MOC gap.
- Shift and MOP briefings already happen. Append the 70E items: incident energy, PPE, boundaries, emergency plan.
- Two-person rules already exist on paper. Reposition the second person outside the boundaries and train them to the emergency response requirements.
In most industries, NFPA 70E 2027 demands a culture change. In a good data center, it mostly demands that you formalize the culture you already have.
Where the new edition bites regardless of operational maturity: verification. Data center switchgear is full of current transformer metering, and the 2027 edition now explicitly recognizes that some circuits — CT secondaries among them — can read near zero volts and still carry hazardous current, so testing for absence of voltage alone is not always sufficient verification. Above 1,000 volts, the standard also cautions that noncontact capacitive test instruments are not suitable for proving voltage is absent. And battery energy storage systems — increasingly common on the critical bus — get their own risk assessment attention in the 2027 edition. None of these map neatly onto an existing data center habit. They have to be trained in.
The Bottom Line for Mission-Critical Operators
The 2027 edition raises the cost and the ceremony of energized work everywhere. Data centers, unusually, get to meet much of it by formalizing disciplines they invented for uptime reasons — the MOP, the dry run, the EOP, the change board. The gap that remains is specificity: a second person positioned and trained correctly, an emergency response plan that solves the locked-door problem before it happens, a briefing that covers incident energy and not just switching order, and verification methods matched to the circuit in front of you. Close those gaps, and a standard that reads like a burden becomes documentation of what your operation already believed.