Nvidia and Firmus will build a 170,000-GPU data center in Indonesia, Yahoo Finance reported. The disclosed accelerator count places the planned facility among the largest single-site AI compute deployments publicly announced, should it reach the named scale.

The Yahoo Finance report carries no project timeline, capital expenditure, site selection, or power-procurement detail. The 170,000-GPU figure is the only quantitative specification in the available coverage. No executive statements or direct quotes appeared in the report.

Who is involved

The report names two counterparties:

• Nvidia, the GPU supplier
• Firmus, the infrastructure partner

Firmus's specific role — developer, owner, or operator — is not detailed in the report. The arrangement's equity, offtake, and tenancy structure also remain undisclosed.

How 170,000 GPUs sizes the project

A 170,000-GPU cluster sits in the upper tier of single-site AI builds. The accelerator count drives three downstream engineering requirements:

• Power delivery typically measured in the hundreds of megawatts
• Full-bisection networking between every GPU, generally served by InfiniBand or Spectrum-X fabrics
• Capital expenditure in the multi-billion-dollar range, by reference to comparable US projects

None of those downstream figures appear in the Yahoo Finance report. The 170,000-GPU count, however, sets the upper bound on what the facility will require, and is unusually large by single-site standards. Many publicly disclosed US clusters have been sized in the tens of thousands of accelerators, not hundreds of thousands.

Why Indonesia

Indonesia is the fourth-most-populous country globally. It sits on subsea cable systems linking Southeast Asia to India, the Middle East, and the US West Coast, factors relevant to data ingestion and inference latency for AI workloads.

A 170,000-GPU project, if built at the announced scale, would rank as one of the largest single-site AI compute deployments disclosed in Southeast Asia. Regional capacity to date has centered on Singapore, with secondary projects in Malaysia and Thailand. The Indonesia announcement, if executed, would shift the regional balance of disclosed capacity.

What remains unspecified

The Yahoo Finance report does not detail:

• Site selection or land agreement
• Power sourcing and grid commitments
• Phased GPU delivery schedule
• Tenant, offtake, or sovereign-AI arrangement
• Construction timeline
• Capital structure

Outlook

The 170,000-GPU figure functions as a public benchmark even in the absence of financial or schedule detail. Public single-site US clusters have typically been disclosed in the tens of thousands of accelerators; a public 170,000-GPU figure for Indonesia sets a new order of magnitude for non-US AI infrastructure announcements.

Project specifications beyond the Yahoo Finance report are not currently public. As site, schedule, power, and capex details emerge, the project will be a reference point for the cost and power profile of single-site AI builds outside the major US cloud regions.