The AI buildout just hit puberty — investors are done buying promises and now demand proof of progress.
The Summary
- Aswath Damodaran calls this AI's "bar mitzvah moment" — investors shifting from faith-based valuations to evidence-based returns following Nvidia's latest earnings
- DataBank operates 70+ data centers across 25+ markets and faces infrastructure bottlenecks, particularly power availability
- The gap between AI compute capacity and real-world deployment constraints is the defining tension of 2026
The Signal
Damodaran's "bar mitzvah" framing cuts through the noise. We've moved past the childhood phase where every AI company could trade on narrative alone. Now the market wants receipts. After Nvidia's earnings, the question isn't whether AI will transform everything. The question is whether the companies building the picks and shovels can actually deliver infrastructure at the scale their valuations assume.
Kevin Ooley, incoming CEO at DataBank, gave a ground-level view of what that delivery problem looks like. His company runs over 70 data centers, and the binding constraint isn't demand. It's power. The AI buildout is infrastructure-limited, not imagination-limited.
"The market wants receipts now, not roadmaps."
This matters because we're watching two parallel realities unfold:
- Nvidia and its peers are selling compute capacity at unprecedented scale
- Data center operators are struggling to turn that capacity into live deployments
- Power grids, cooling systems, and physical real estate are the actual bottlenecks
The "bar mitzvah" metaphor works because it captures a phase transition. You can't fake adulthood. You can't smooth-talk your way through a power shortage. The companies that survive the next 18 months will be the ones who solved the boring problems: kilowatts, cooling loops, fiber routes, and permitting timelines.
The Implication
If you're tracking the agent economy, watch infrastructure operators more closely than chip designers. The companies building Web4 need compute, but compute needs power, and power needs real-world construction timelines. The AI winter won't come from a lack of ideas. It'll come from a lack of electricity in the right places.
For builders: your agent roadmap should account for infrastructure constraints. If you're planning to scale AI-heavy features in 2027, talk to the people running data centers today. The gap between what's technically possible and what's physically available is widening.