All the capex being built out in data centers is being lumped under cloud computing for their respective companies. However, these data centers are being populated with GPUs which are only good for parallel computing and vector math, not spinning up virtual systems, storage, etc.
Seems like data centers whose only use is AI inference and training need to be considered as a separate category. If there is a burst in an AI bubble and enterprise customers pull back from paying for tokens that entire ecosystem dies, or seriously retreats, but other proven cloud services will remain.
What am I missing here?