Planned vs Unplanned Outages

Planned maintenance can be scheduled and backed up; unplanned trips cannot — separating the two turns a large raw downtime budget into a small disruption to compute.

Not all downtime hurts equally

A scheduled magnet or vacuum-system service, announced weeks ahead and run in a low-demand window while the grid carries the campus, barely disturbs a data center. An unannounced plasma trip at peak utilization is a different event: milliseconds matter, and the ride-through layer has to catch the load before anything notices.

For a fusion plant, the larger share of the downtime budget is expected to be planned: magnets, cryogenics, vacuum, and first-wall inspection are periodic, predictable activities. Because they are predictable, they are coverable — the grid tie imports during the window and the compute load never sees a gap. The felt availability can therefore be much better than the raw availability suggests.

Unplanned outages are the ones that test the design. A plasma trip must hand off to the battery layer within its ride-through window, and the transfer to grid import must complete before the battery depletes. This is standard data-center practice, applied to a new base source.

The honest point stands: the burner's raw 0.86-0.995 is real, but the disruptive fraction — sudden, uncovered outage — is what the architecture drives toward zero by scheduling the planned part and bridging the unplanned part.

• Planned outages are scheduled and grid-backed
• Unplanned trips are caught by the battery layer
• Most fusion downtime is planned and coverable
• Felt availability can beat raw availability with good scheduling