num_undelivered_messages shows up malformed when running 'kubectl get hpa'

I followed this tutorial:
https://cloud.google.com/kubernetes-engine/docs/tutorials/autoscaling-metrics#pubsub

I noticed that these are the values that show up when I run 'kubectl get hpa'
% kubectl get hpa pubsub --watch 12:38:09 !10161

NAME REFERENCE TARGETS MINPODS MAXPODS REPLICAS AGE
pubsub Deployment/pubsub 26400m/2 (avg) 1 5 5 10m
pubsub Deployment/pubsub 24200m/2 (avg) 1 5 5 10m
pubsub Deployment/pubsub 23800m/2 (avg) 1 5 5 11m
pubsub Deployment/pubsub 23800m/2 (avg) 1 5 5 16m
pubsub Deployment/pubsub 0/2 (avg) 1 5 5 16m
pubsub Deployment/pubsub 0/2 (avg) 1 5 5 21m
pubsub Deployment/pubsub 0/2 (avg) 1 5 1 21m
pubsub Deployment/pubsub 2/2 (avg) 1 5 1 22m
pubsub Deployment/pubsub 0/2 (avg) 1 5 1 23m
pubsub Deployment/pubsub 34/2 (avg) 1 5 1 24m
pubsub Deployment/pubsub 8500m/2 (avg) 1 5 4 25m
pubsub Deployment/pubsub 6800m/2 (avg) 1 5 5 25m
pubsub Deployment/pubsub 15400m/2 (avg) 1 5 5 25m
pubsub Deployment/pubsub 19400m/2 (avg) 1 5 5 26m
pubsub Deployment/pubsub 18/2 (avg) 1 5 5 27m
pubsub Deployment/pubsub 16/2 (avg) 1 5 5 28m
pubsub Deployment/pubsub 13/2 (avg) 1 5 5 29m
pubsub Deployment/pubsub 9/2 (avg) 1 5 5 30m
pubsub Deployment/pubsub 3/2 (avg) 1 5 5 31m

I looked on Metrics Explorer, and the highest value of num_undelivered_messages was 163, so I'm not sure what the value of 26400m comes from. Does anyone know why these values are showing up?