Reference
·Supporting source
·PreviewHyperscale data centre capex per MW IT load (Thailand 2026)
~USD 10-13M
Triangulated capex range per MW of IT load for new Thai hyperscale builds, including shell, MEP, generators, and chillers. AI-dense GPU halls run 15-30 percent higher per MW than general-purpose cloud halls.
Figure in context
Triangulated capex range per MW of IT load for new Thai hyperscale builds, including shell, MEP, generators, and chillers. AI-dense GPU halls run 15-30 percent higher per MW than general-purpose cloud halls.
Triangulated capex range per MW of IT load for new Thai hyperscale builds, including shell, MEP, generators, and chillers. AI-dense GPU halls run 15-30 percent higher per MW than general-purpose cloud halls.
Time scope
2026 build cohort
Source basis
Supporting source
Interpretation notes
What this tells you
Triangulated capex range per MW of IT load for new Thai hyperscale builds, including shell, MEP, generators, and chillers. AI-dense GPU halls run 15-30 percent higher per MW than general-purpose cloud halls.
What not to do with it
Use the linked report for interpretation and keep basis differences explicit.
Related figures
Adjacent numbers that add context without drowning the value.
Thailand datacenter operating IT load (2024 baseline)
Structure Research, operator disclosures
Thailand datacenter operating IT load (2027 base case)
Structure Research, operator phasing, BOI Section 8 pipeline
Hyperscaler combined Thailand commitment
AWS, Google Cloud, Microsoft press releases
New build PUE target band (Bangkok hyperscale 2026-27)
Operator sustainability disclosures, MDES certifications
GPU-as-a-service share of Thai AI workload (2027 modelled)
Operator disclosures, NVIDIA AI Center Thailand, hyperscaler partner channel data
BoT virtual banking licence winners with GPU-cluster commitments
Bank of Thailand virtual bank licence press, winners' disclosures
Report context
Atlas actors in this figure's reports
Profiles covered in the report that cite this number.