You must have a plan to ramp down to a steady state of consumption that keeps things safe. Even if they need many megawatts, the 50 million in damages figure quoted elsewhere buys a hell of a lot of generators.
I wonder if it's even possible to ramp down this single process without a similar hit to productivity as an uncontrolled power loss, never mind the other stages in the fab process. If not, fabs can consume tens of megawatts, so what does it cost to build and maintain the necessary backup generation capacity to weather a 30-minute outage?
One interesting example: USD $1.5 million for 10MW of flywheel-based backup capacity (http://ir.p10industries.com/news-releases/news-release-detai...) - but these will last ~15 seconds under full load.
Tesla's install at the Hornsdale Power Reserve [1] is 129 MW of storage capable of discharging at 100MW. This could supply power for far more than 30 minutes if an outage occurs. Estimated cost is $50 million, which appears to be roughly the same cost as this manufacturing failure, so there's a middle ground to be found.
[1] https://en.wikipedia.org/wiki/Hornsdale_Wind_Farm#Hornsdale_...
Flywheels are an alternative for battery backups. They are used to give enough time for generators to come online and stabilize.