Atlasresearch

Workspace

WorkspacePokedexHPC MapPortfolioCalendarTaiwan RevenueMemory
Calculators
  • 01AI Factory
  • 02CPO Stack
  • 03HBM Bridge
  • 04GPU Cloud TCO
  • 05Inference Tokens
  • 06Power Bottleneck
  • 07Custom ASIC
  • 08OCS TAM
  • 09N3 Allocation
  • 10Optical TAM
Agent APISubscribe
Sign in
Calculator
Atlasresearch

Workspace

WorkspacePokedexHPC MapPortfolioCalendarTaiwan RevenueMemory
Calculators
  • 01AI Factory
  • 02CPO Stack
  • 03HBM Bridge
  • 04GPU Cloud TCO
  • 05Inference Tokens
  • 06Power Bottleneck
  • 07Custom ASIC
  • 08OCS TAM
  • 09N3 Allocation
  • 10Optical TAM
Agent APISubscribe
Sign in
Calculators/AI Factory

Jason's Chips

Subscribe or sign in with your Substack email for live inputs, outputs, and saved scenarios.

Subscribe to Jason's ChipsSign in

Inputs

Last calculated: Not yet

Calculator 01AI infrastructure math

AI Factory Calculator

Convert datacenter MW/GW announcements into GPUs, racks, capex, semiconductor revenue, and power draw. Defaults are editable so the model can follow new platform disclosures.

Sensitivity

GPUs

600K

1.67 kW / GPU

NVL / AI racks

8.3K

72 GPUs / rack

Critical IT power

1.00 GW

Input basis

Total facility power

1.20 GW

PUE 1.20

Total capex

$73.71B

$73.7M / MW IT

Semiconductor revenue

$24.78B

GPU x 1.18 content

Capex breakdown

Total project capex

$73.71B

38%
25%
18%
9%
9%
Category% of totalDollars$/MW IT
IT equipment: GPUs, servers, networking, storage38%$28.35B$28.4M
Power and cooling infrastructure25%$18.14B$18.1M
Building and site infrastructure18%$13.61B$13.6M
Support systems and controls9%$6.80B$6.8M
Land, permits, contingency9%$6.80B$6.8M

Capex density

Total $ / MW IT$73.7M
IT equipment / MW$28.4M
Non-IT / MW$45.4M
Total $ / GPU$122.9K
Total $ / rack$8.8M

Sensitivity analysis

GPU ASP x non-IT build intensity

GPU ASP \ Non-IT / IT1.20x1.40x1.60x base1.80x2.00x
$28.0K$49.90B$54.43B$58.97B$63.50B$68.04B
$31.5K$56.13B$61.24B$66.34B$71.44B$76.55B
$35.0K base$62.37B$68.04B$73.71B$79.38B$85.05B
$38.5K$68.61B$74.84B$81.08B$87.32B$93.56B
$42.0K$74.84B$81.65B$88.45B$95.26B$102.1B

Sensitivities hold rack power, PUE, utilization, and platform mix constant. This isolates the two assumptions that move most equipment revenue: GPU ASP and non-IT build intensity.

Deployment rate

41.7 MW / mo.

Required average critical IT energization pace across the stated build timeline.

Useful GPU-hours

4.84B

Annual GPU-hour capacity at the selected steady-state utilization.

Rack density

1.67 kW / GPU

Implied accelerator power density from rack power and GPUs per rack.

Assumptions

GB200 NVL72

PlatformGB200 NVL72
NetworkQuantum-X / Spectrum-X
GPUs / rack72
Rack power120 kW
PUE1.20
GPU ASP$35.0K
IT equipment x GPU1.35x
Non-IT / IT1.60x

Saved scenarios

1GW GB200 base case

1.0 GW IT / 600K GPUs

$73.71B

300MW GPU building

300 MW IT / 180K GPUs

$22.11B

4.5GW AI campus

4.5 GW IT / 2.70M GPUs

$331.7B