Working Notes on Agent Systems/Brad Zhang

@teach_fireworks / X longform

By 2030, global data center computing power (measured in gigawatts of power consumption) is expected to

By 2030, global data center computing power (measured in gigawatts of power consumption) is expected to increase from 55 GW in 2023 to 171 ~ 298 GW, which is more than...

September 27, 2025 · 1 min read

By 2030, global data center computing power (measured in gigawatts of power consumption) is expected to increase from 55 GW in 2023 to 171 ~ 298 GW, which is more than a threefold increase.

Energy demand is surging: By 2030, data center power consumption could approach that of a mid-sized country.

AI is the main driver: the rise of large models and inference services has significantly increased the demand for computing power.

Sustainability challenges: Without improving energy efficiency or expanding the use of clean energy, the pressure on carbon emissions will be significant.

Investment opportunities: Chips, data center construction, power infrastructure, and renewable energy are all likely to experience accelerated development.

Driving factors: Rapid popularization of AI (training & inference requires a lot of computing power) Growth in shipments of different types of chips (GPU, ASIC, etc.) Corresponding energy consumption (power consumption is an intuitive proxy for data center scale) Increased demand for computing, storage and network

Visual summary

Article argument map

Generated from the post's content graph

FORMATTOPICCAPABILITYMARKETcoverscoversFORMATlongform noteTOPICinferenceTOPICretrieval
Mermaid outline
flowchart LR
  format-long_post["longform note"]
  topic-inference["inference"]
  topic-retrieval["retrieval"]
  format-long_post -->|covers| topic-inference
  format-long_post -->|covers| topic-retrieval

Visual structure

Essay structure map

Built from summary and key paragraph positions

By 2030, global data center computing power (measured in gigawatts of power consumpti...THESISBy 2030, global datacenter computing power(measured in gigawattsof power consumption)SIGNALBy 2030, global datacenter computing power(measured in gigawattsof power consumption)OPERATORSustainabilitychallenges: Withoutimproving energyefficiency orIMPLICATIONDriving factors: Rapidpopularization of AI(training & inferencerequires a lot of
Mermaid outline
flowchart LR
  thesis["By 2030, global data center computing power (measured in gigawatts of power consumption) is expected to inc..."]
  signal["By 2030, global data center computing power (measured in gigawatts of power consumption) is expected to inc..."]
  operator["Sustainability challenges: Without improving energy efficiency or expanding the use of clean energy, the pr..."]
  implication["Driving factors: Rapid popularization of AI (training & inference requires a lot of computing power) Growth..."]
  thesis -->|frames| signal
  signal -->|develops| operator
  operator -->|lands in| implication

Source: View the original post