XMight is exhibiting at OCP Global Summit 2026 · San Jose · 12–15 October Meet us →
THE CONTROL LAYER BETWEEN GRID AND AI COMPUTE

Active power orchestration for grid-compliant AI racks — from ORv3 today to Mt Diablo ±400 VDC / 800 VDC.

The grid won't bend for AI. So we make the rack bend for the grid.

XMight's POM implementation coordinates BBU and PSU output at rack level to meet ERCOT / NERC ride-through and post-fault recovery requirements — on ORv3 / ORW racks today, and designed for Mt Diablo ±400 VDC / 800 VDC power racks.

Talk to us Read the architecture

THE PROBLEM

Grid compliance just became a continuous control problem.

On 10 July 2024, a single transmission fault in Northern Virginia caused roughly 1,500 MW of data-center load to disconnect from the grid within seconds. AI training and inference create power transients on millisecond time scales — the kind of behavior grid operators can no longer ignore.

Nine grid operators are now writing rules so that does not repeat. ERCOT NOGRR 282 has been in force since 1 August 2026: voltage ride-through and post-fault active power recovery are mandatory for large computational loads — including data centers above the size threshold.

Read the compliance landscape →

WHAT WE BUILD

Two coordinated layers. One closed loop.

Facility-level policy layer

XMight POM Manager

Sets the power envelope for each rack from grid conditions and site policy. Records the per-rack evidence grid operators increasingly ask for.

Interfaces: Redfish · PMBus · POM interface

Learn more →

Rack-mounted hardware module

XMight POM Module

Rack-level module in the ORv3 power shelf's PMI-class slot, with its own control plane. Coordinates BBU discharge and PSU output shaping in real time, and keeps operating autonomously if the facility layer is unreachable.

Form factor: ORv3 power shelf · PMI-class slot · POM implementation

View spec →

ARCHITECTURE

Set the envelope, coordinate the rack, prove it.

XMight's POM implementation sits between the facility power management layer and the ORv3 rack — setting the envelope at facility level and enforcing it autonomously at the shelf. Built on POM, the open rack-level coordination interface XMight has contributed to OCP Rack & Power.

DATA CENTER EMSFleet scheduler · workload orchestratorPOWER GRIDDemand · capacity · frequency schedulerXMIGHT POM MANAGER — FACILITY LAYERFacility-level envelope and policyVRT POLICYVoltage ride-through envelope enforcementPFAPR RECOVERY CONTROLPost-fault active power recovery slopeCONTROL PLANEXMIGHT POM MODULE — IN THE POWER SHELFRack-mounted power orchestration moduleVRT EDGE POLICYLocal VRT enforcementBBU/PSU ORCHESTRATORDischarge + output shapingLOCAL EVENT DETECTIONAutonomous rack-level responseTELEMETRY AGGREGATORFleet data collectionAC INPUTGrid powerORv3 RACKACPOWER SHELFAC → DC conversionPOM48V DC48V DC BUSBBU SHELFStandard backupCharge ⇄ dischargePOMBBU — VRT/FRT/PFAPRGrid compliance eventsCharge ⇄ dischargePOMAI SERVER ×6+GPU compute load
POM (Power Orchestration Module) — XMight's open contribution to OCP Rack & Power
See architecture detail →

WHERE WE STAND

Where we stand.

XMight is a member of the OCP Startup Program 2026, a named contributor to the OCP DCF LVDC Power Distribution white paper v1.1, and has POM in review with the OCP Rack & Power project. The core rack-level power-orchestration method is granted as US 12,510,873 B2.

FAQ

Common questions about XMight and POM.

What does XMight build?
Two products on POM for OCP ORv3 / ORW AI racks: the XMight POM Module, a rack-mounted module in the power shelf's PMI-class slot that coordinates BBU discharge and PSU output and keeps operating autonomously, and the XMight POM Manager, a facility-level layer that sets the power envelope for each rack. Together they meet ride-through and recovery requirements without throttling GPU compute.
What is NOGRR 282?
NOGRR 282 is the ERCOT Nodal Operating Guide revision that mandates voltage ride-through (VRT) and post-fault active power recovery (PFAPR) for large computational loads — including AI data centers above a size threshold. It has been in force since 1 August 2026; one provision, continued consumption in the 0.50–0.80 p.u. band, becomes mandatory on 1 January 2028. SPP has adopted its own ride-through rules (RR 696), and other North American grid operators are following.
Why can't ORv3 / ORW racks pass NOGRR 282 without help?
OCP ORv3 / ORW racks distribute battery backup (BBU) across each rack for energy efficiency but provide no orchestration layer. On grid restoration, PSU startup follows a randomized 0–5.5 s delay defined by the ORv3 PSU specification, on top of a cold-boot of several seconds. That is incompatible with the recovery windows grid codes now set — ERCOT requires at least 90% of pre-disturbance load back within 2 seconds of voltage recovery.
How is POM different from a traditional UPS?
A facility UPS or BESS supplies bulk energy. XMight's POM implementation coordinates how the energy already in the rack — PSU hold-up and BBU — is used during a disturbance, and sequences the recovery. The two are complementary: rack-level coordination lets a facility size its UPS or BESS for the services that need bulk energy. It operates at rack level: the XMight POM Module installs in the ORv3 power shelf as a POM-class module — hot-swap, minutes per rack, no facility-level electrical work. IT and OT are fully decoupled.
Does the XMight POM Module throttle GPU compute?
No. The XMight POM Module orchestrates the power layer (BBU + PSU) without touching the compute layer. AI workloads continue at full throughput during grid events — in contrast to software GPU throttling approaches, which interrupt synchronous training with straggler effects and NCCL communication failures.
Is POM an open specification or an XMight product?
POM (Power Orchestration Module) is an open specification, in review with the OCP Rack & Power project. Any vendor can implement it. The XMight POM Module and POM Manager are XMight's commercial implementations.
Is XMight a member of the Open Compute Project (OCP)?
Yes. XMight Smart Energy Corp. is a member of the OCP Startup Program 2026. We contribute to the OCP Rack & Power working group (advancing POM as an open standard) and to the OCP Data Center Facility LVDC Work Stream (named contributor to the LVDC Power Distribution white paper v1.1). POM is designed for Mt Diablo ±400 VDC and NVIDIA 800 VDC sidecar architectures, where one power rack feeds several IT racks.
Is NOGRR 282 already enforced?
Yes. NOGRR 282 took effect on 1 August 2026 and applies to large computational loads on the ERCOT grid now. A load that fails to ride through a disturbance must submit a root-cause investigation and a remediation plan, and ERCOT can require disconnection until compliance is shown. SPP has its own ride-through rules (RR 696); NERC is drafting continent-wide standards under FERC order RD26-7-000.

The grid code is already in force. Talk to the team.

Contact us Read the architecture