chokepoints.ai
SUBSCRIBE
10 layers580 nodes2,376 dependencies9 chokepoints112 bottlenecks6,500+ companiesnode size = companies identified

Data centre physical

BOTTLENECK

Years-long lead times for liquid-cooling and construction capacity let Nscale capture lease premiums and power mark-ups.

Physical infrastructure housing, powering, cooling and securing AI compute hardware, from utility service entrance to rack face. Excludes external grid connection and in-rack IT hardware. Operators earn lease premiums and power mark-ups; construction and liquid-cooling supply chains gain from constrained capacity and long lead times.

All physical infrastructure required to house, power, cool and secure AI compute hardware, from land acquisition and facility construction through to ongoing operations and decommissioning. Covers everything from the utility service entrance to the rack face; excludes the external grid connection (L2) and in-rack IT hardware (L4).

What the evidence shows

Transformer lead times from traditional OEMs run 26 weeks or longer.

gigaenergy.com

Equipment lead times range from eight to 24 months.

bain.com
RESCORED JUL 2026fragmentedscaling739 companies

Who controls it

Nscale+738 more tracked
$295B market · 202511.66% CAGRsource

Where it sits in the stack

Takes in: Entitled land, utility power at service entrance, construction labour and materials

Sends on: Commissioned, operational physical capacity (power + space + cooling + connectivity) ready for IT hardware

41 upstream · 5 downstream

view in atlas

Related nodes

No other tracked node in the atlas currently shares this parent.

Full scorecard, owner shares, supply edges and the full tracked roster are in the desk letter.

GET THE BRIEFING