AI capex just got revised up $225B in 6 months and earnings are actually following it, this is not how bubbles work and it's messing with my thesis.

So hyperscaler capex started 2026 at $515B and it's now $740B mid-year, with 2027 already pencilled in at $889B that's 2.7% of US GDP going into AI infrastructure, in 2019 this number was 0.3%. The part that's really breaking my bubble thesis is that EPS is actually moving with it. tech earnings estimates for 2027 are up nearly 20% YTD and in 1999 profits lagged the spending by a mile that's not what's happening here, at least not yet, also nobody seems to be talking about the equity returns in