ArusGridProject
Directory › Thailand › Bangkok metro › Datasection Bangkok AI Data Center

Datasection Bangkok AI Data Center

Bangkok, Thailand · Bangkok metro market · location approximate to city level

Datasection Bangkok AI Data Center is a data center project in Bangkok, Thailand, operated by Datasection Inc. It is being commissioned. It is planned for 10 MW of IT load.

Record last checked against its sources on 19 October 2026.

Open on the mapSuggest a correction
Stage
Commissioning
Planned IT load
10 MW
Running now
0 MW
Operators rarely publish this
Operator
Datasection Inc.
Local developer
AI workloads
Confirmed
Evidence grade
Corroborated
Will it get built?Built7 of 7 gates passed
✓ Land secured✓ Power arranged✓ Approvals granted✓ Money committed✓ Tenant or own use✓ Builder appointed✓ Construction started

This project is built or running, so every gate counts as passed.

Build odds is ArusGrid's reading of the public record, not a forecast, a rating or advice. It counts the gates this project has passed in public sources. Private progress does not show, a passed gate can be undone, and projects that pass every gate still stop. The method has not yet been tested against outcomes. How it works.

How Thailand approves, powers and taxes data centers is on Data center rules in Thailand.

Thailand, 2026-09-24. Thailand asked developers on 4 September 2026 to pause 49 data centres under construction and froze 117 pending approvals while it writes new rules, due by mid October. The government has said it cannot legally enforce the pause. No list of affected projects has been published, so no Thai project has changed stage on this basis. A separate data centre electricity tariff has been created with no rate set.

Where it sits

Bangkok metro holds 23 active projects with 893 MW of planned IT load between them, and Datasection Bangkok AI Data Center ranks 16th by planned capacity. Of those, 6 are operational, 3 are operational and expanding, 5 are under construction, 2 are being commissioned, 3 are at the land and permitting stage and 4 are announced only. Published colocation vacancy in Bangkok metro was 22.1% as of June 2026, on 113 MW of operational colocation capacity as counted by Cushman & Wakefield. Vacancy excludes hyperscaler self-build.

It is the only Datasection project ArusGrid tracks. No tenant has been disclosed, which is normal for colocation and says nothing about occupancy.

Grid power for a large industrial user in Thailand costs about US$0.10 per kWh as of July 2026, the country figure ArusGrid applies where a site has no published rate. The US export control position for advanced AI chips in Thailand, as of 27 September 2026, is recorded as “No US destination-based licence requirement as enforced”. ArusGrid holds 2 public sources for this project, the latest dated 8 July 2026, and last checked the record on 19 October 2026.

How well evidenced is it

Construction or financing is evidenced, or several independent signals agree. The rules are fixed and published on the Methodology page.

Reaching this stage requires a dated construction source, which corroborates the project on its own.

Named funderNo
Disclosed client (confirmed)Yes
Committed or approved capitalNo
Dated, sized phaseYes
Three or more sourcesNo
Dated source in the last 24 monthsYes

AI workloads

StatusConfirmed

Confirmed GPU deployment: 4,696 Nvidia B200 GPUs (587 servers) purchased from ASRock Rack for ~US$257.4m, to be installed in this Bangkok facility with phased operations from Jul-Aug 2026 (Datasection disclosure, 4 Jun 2026; DCD 5 Jun 2026).

Access to AI chips

GradeNo US licence required

This country carries no US destination-based licence requirement for advanced AI chips as the rules are enforced today. Access is a commercial matter between the project and its vendors. Criteria are published on the Methodology page.

US licence required for this destination as enforcedNo
A party headquartered in China or Macau on the projectNo
Country-level export frameworkNo
A party named in a government authorisationNo
US approved operator or confirmed client on the projectNo
Named chip-vendor partnershipNo

No US destination-based licence requirement as enforced · No US destination-based licence requirement as enforced (2026-09-27) source

Phases

Phase 1 · Commissioning10 MW, target 2026

Who is behind it

OperatorDatasection Inc. · Local developer

Who uses the compute

Runs compute here · enterpriseUndisclosed US technology company · confirmed

10 MW · 4,696 Nvidia B200 GPUs (587 servers), US$257.4m contract, phased start July and August 2026 · as of 2026-06-05

DCD reports Datasection's TSE disclosure of a US$257.4m contract to supply 4,696 B200 GPUs from a 10MW Bangkok data center to an unnamed US-based technology company, funded partly by advance customer payments. The operator of the building is not named in the report. ArusGrid has this as confirmed, which is right if it rests on the TSE filing itself; this press source is reported.

Source · Japan's Datasection to deploy 4,696 B200 GPUs in Thailand (DCD, 2026-06-05)

Platform deals that include it

Datasection GPU purchase for Bangkok AI data center
Other, 2026-06
US$257m

Platform capital spans several sites. It is never added to this project's own figure.

Colocation price per kW of IT load per month

No published price band exists for Thailand, and ArusGrid does not estimate from the regional benchmark alone. What was searched is listed on the Colo prices page.

Build, land, power and water
Nothing found in public sources yet. Most contractors and utilities never name their data centre clients. Know who built it or where its power comes from? Tell us.

Sources checked 2026-10-19

  1. datacenterdynamics.com
  2. บีโอไอ อนุมัติลงทุน 9 โครงการ 6.6 หมื่นล้าน ตั้งทีมรองรับดาต้าเซ็นเตอร์ (The Bangkok Insight, 2026-07-08)

More in Bangkok metro

Planned MW is IT load, the power available to servers, unless marked otherwise. Running capacity is rarely published, so a missing figure means undisclosed. Definitions, sourcing standards and evidence rules are on the Methodology page. Data is CC BY-NC 4.0. Corrections go to corrections@arusgrid.com. Dataset updated 2026-10-18.