Hyperscaler capex may still be real demand

$AMZN and $GOOGL are becoming the center of the AI capex debate.
The bear case is simple: hyperscalers may be overbuilding infrastructure before demand is proven.
But the bull case says the opposite.
Cloud growth has re-accelerated, AI workloads are absorbing new capacity, and capex may be following real demand rather than creating empty supply.
Some investors now see total hyperscaler capex moving toward $1T within the next one to two years.
The real risk is the signal from management.
If $MSFT, $GOOGL, or $AMZN says supply has finally caught up with demand, high-multiple AI hardware names could reprice fast.
Until then, the market may keep treating the buildout as infrastructure for a much larger AI cycle.