Disclose Utility Outage History and SAIDI/SAIFI for GPU Offtake Sites

Consider a site owner offering unused grid capacity for GPU offtake. The opening packet includes a utility reliability report with a low average interruption duration. During diligence, the buyer asks about a lengthy feeder outage noted in the facility’s maintenance log. That event was excluded from the utility’s headline metric as a major-event day.

Neither record is necessarily wrong. The problem is an unexplained difference in scope.

For Pacific Intelligent Technologies, Inc., a useful site-owner disclosure separates utility statistics from observed site interruptions, identifies exclusions, and gives buyers evidence they can reconcile. Here is the playbook.

Provide five years where available—and disclose gaps

As a practical diligence target, provide five completed calendar years plus current year-to-date history. This is a recommended disclosure window, not a universal utility reporting requirement. If only three years—or less—are available, provide them and explain why the record stops there.

Include annual indices and an event-level interruption log. Annual averages can conceal a single long restoration; an event log shows duration, timing, and recurrence.

State:

• The reporting period and year-to-date cutoff.
• When site monitoring began and any missing intervals.
• Whether ownership, utility service configuration, or monitoring systems changed.
• Which records were requested but remain unavailable.

Do not turn “no retained records” into “no outages.” For an undeveloped pad or previously unenergized service, label nearby-facility or feeder history as contextual evidence, not measured performance at the proposed GPU site.

Define the metric and its electrical boundary

SAIDI is total customer interruption duration divided by customers served, commonly reported as minutes per customer for the reporting period. SAIFI is total customer interruptions divided by customers served, commonly reported as interruptions per customer.

These are customer-weighted system measures—not direct predictions of GPU workload availability.

For every figure, identify the reporting scope: utility territory, district, substation, feeder/circuit, or site service point. Utilities may use “circuit” and “feeder” interchangeably; reproduce their terminology and identify the actual asset rather than assuming two distinct levels.

Record the source’s sustained-interruption threshold, methodology, units, and customer population. If the utility publishes equivalent reliability indices, retain their definitions instead of relabeling them SAIDI or SAIFI.

Distinguish feeder statistics from site observations. A branch fault may affect the site without interrupting every feeder customer. Conversely, an outage elsewhere on the feeder may not affect the site.

Identify where site interruptions were observed: incoming utility service, downstream switchgear, or a customer meter. Reference the metering and settlement boundary disclosure for physical context without making settlement rules the subject here.

Show major events and planned outages separately

Where available, provide annual reliability results both including and excluding major-event days. Name the utility’s classification method, such as its stated IEEE 1366 methodology, and cite the report edition. Do not independently label every storm a qualifying major event.

Keep major-event interruptions visible in the event log even when the published index excludes them. Buyers need to understand severe-weather exposure and restoration duration, not just normalized utility performance.

Separate:

• Unplanned utility interruptions: faults, equipment failures, and other unscheduled supply losses.
• Planned utility interruptions: scheduled maintenance or construction, including notice received and actual duration.
• Customer-side interruptions: facility equipment or operational events, clearly distinguished from utility supply failures.
• Momentary interruptions and voltage disturbances: reported separately where records exist, because sustained-outage indices may omit them.

State whether each published index includes planned interruptions. Never combine incompatible annual series without explaining the methodological difference.

Attach an evidence annex buyers can reconcile

Build a compact annex with an annual summary, an event log, and a source register.

For each event, capture start time, restoration time, time zone, duration, affected service point, planned/unplanned status, reported cause, major-event classification, and supporting reference. Mark unknown causes as unknown. If restoration was staged, explain which timestamp represents restored service to the site.

Useful evidence includes utility reliability reports, regulator filings, utility outage notices, SCADA exports, customer-meter interval or event data, and maintenance logs.

Explain each source’s limits. A public utility report supports territory-level statistics; it does not establish site-specific history. A meter communications gap is not automatically an outage. SCADA records need signal descriptions and clock checks.

Preserve originals, document reconciliation differences, and share sensitive infrastructure records through an appropriately controlled diligence process.

Explain how buyers use the record—not what it guarantees

Buyers compare outage frequency, duration, seasonality, and restoration patterns with their operating assumptions and backup-power plans. For a proposed Supermicro HGX B300 deployment, describe the intended pad load and uptime sensitivity as project inputs—not as conclusions derived from SAIDI.

Utility interruption history and backup coverage answer different questions. Cross-reference the UPS and generator disclosure, but do not claim that installed backup erases historical exposure.

Owners can use the bridge-capacity overview and capacity overview to frame the opportunity. Pacific Intelligent Technologies, Inc. provides the broader company context.

Have utility capacity and an incomplete outage record? Schedule a site-owner discussion with the records available and a list of gaps. Ask counsel and electrical engineering to review proposed disclosure language before circulation, especially exclusions, attribution, and statements that could be read as reliability guarantees.

FAQ

Can I disclose utility-wide SAIDI/SAIFI if feeder data is unavailable?

Yes—label the geographic scope and limitation prominently. Include available site logs and identify feeder records as unavailable or pending in your site-owner offtake brief.

Does low SAIDI establish a site uptime commitment?

No. Historical customer averages do not establish future site or workload availability. Keep operating-mode claims in the separate emergency shutdown and islanding disclosure.

Should planned outages remain visible if the utility excludes them?

Yes. List them separately with notice and duration. Buyers need an operational history, not merely a reproduction of the utility’s headline index.