GPU available on the VM but not on the Kubernetes node

On the node I see a Tesla GPU:

$ cat /proc/driver/nvidia/gpus/0000\:00\:07.0/information 
Model: 		 Tesla T4
IRQ:   		 10
GPU UUID: 	 GPU-01fd306e-ac2f-6c63-329a-c0f60196601e
Video BIOS: 	 90.04.38.00.03
Bus Type: 	 PCIe
DMA Size: 	 47 bits
DMA Mask: 	 0x7fffffffffff
Bus Location: 	 0000:00:07.0
Device Minor: 	 0
GPU Excluded:	 No

While the Kubernetes node reports 0 GPUs:

Capacity:
  cpu:                4
  ephemeral-storage:  66573292Ki
  hugepages-1Gi:      0
  hugepages-2Mi:      0
  memory:             15999492Ki
  nvidia.com/gpu:     0
  pods:               110

This happens for ml-prod-2022-07-01-t4-kzd3ozfhuqbz-node-2 and ml-prod-2022-07-01-t4-kzd3ozfhuqbz-node-3.

Selinux is disabled, nvidia-gpu-device-plugin running. No errors.