"Consistently achieving >99.95% availability of our GPU nodes" What do GPU node availability numbers look like for other companies? Do they share these numbers?