Define the failure case at the branch

Identify which power cords, PSUs, rack PDUs and upstream circuits remain after the reviewed loss. HPE's feed guidance requires the remaining arrangement to sustain the actual system load. That is different from proving that the two paths are independent. This question concerns capacity in the path that survives, including its permitted voltage and equipment conditions.

Avoid halving the required service by assumption

Hypothetically, a server taking 600 W across two feeds might take approximately 300 W per feed in one normal sharing state. That observation alone cannot justify two branches each capable of only 300 W if either must sustain the server after the other is lost. The figures are illustrative; actual remaining input can also change with PSU efficiency and control mode.

Check the input-voltage case used for the surviving feed. A redundancy statement for the server does not establish identical PSU capacity or branch current under every supported supply-voltage condition.

Check combined degraded demand

Include other equipment that transfers onto the same remaining source and any required cooling or communications loads. Ask the IT and electrical teams for supported load cases, with redundancy configuration and firmware identified. Do not demonstrate capacity by deliberately removing live supplies outside approved procedures. Provide normal and degraded input profiles to the wind integrator. Wind can reduce energy purchases while the retained electrical infrastructure still needs the complete accepted failure-case capacity; no MARSTON server-backup or feed-transfer capability follows from the average demand.