Skip to content

HUMAIN–NVIDIA Saudi AI Factories

Kingdom of Saudi Arabia; sites unresolved, Saudi ArabiaStatus under reviewProgram
Evidence
Headline capacity
See constituent sites
Updated
Oct 28, 2026

HUMAIN and NVIDIA announced a five-year Saudi AI-factory program with projected capacity up to 500 MW and a first deployment of 18,000 NVIDIA GB300 Grace Blackwell systems. The initial deployment has not been assigned to a named site.

Capacity

GPU fleet ambition

Forecast

Several hundred thousand NVIDIA GPUs

Five-year AI-factory buildout · Phase

Compute quantity or benchmark; not power capacity

Source

Operating IT capacity: Not established in published records

A missing figure does not mean zero. Planned, utility and generation figures are not operating IT capacity.

Development

First NVIDIA deploymentPlanned

Compute deployment phase; no named site, MW allocation or achieved delivery date is directly disclosed.

Five-year AI-factory buildoutPlanned

Program ceiling, not delivered facility IT load.

Milestones (3)
Saudi AI-factory program announced

May 13, 2025 · Recorded actual event

Completed · Program designation

Source not linked
Riyadh and Dammam works reported without NVIDIA allocation

Aug 25, 2025 · Recorded actual event

Completed · Status changed

Source not linked
Up-to-500-MW program horizon

2030 · Forecast / planned record

Forecast · Forecast completion

Source not linked

Participants

HUMAIN

Investor

Sources by role
Public Investment Fund

Investor

Sources by role
Other participants (1)

Technical specifications

Networking

Forecast

NVIDIA InfiniBand

First NVIDIA deployment · Phase

Source

Additional project facts

Partnership

Actual

HUMAIN and NVIDIA strategic partnership

Source

Deployment horizon

Forecast

5 years

Five-year AI-factory buildout · Phase

Source

First deployment

Forecast

18,000 NVIDIA GB300 systems

First NVIDIA deployment · Phase

Source

Omniverse deployment

Forecast

Country's first NVIDIA Omniverse Cloud as a multi tenant system

Source

Parent context

Actual

HUMAIN launched under PIF in May 2025

Source

Infrastructure & permits

Technical details (5)
Electric grid

Planned or represented

500 MW up to

Projected national AI-factory capacity over five years; not delivered IT load and not allocated to a site.

Compute and platform system

Planned or represented

Program ambition; no exact count or site allocation.

Compute and platform system

Planned or represented

18,000 NVIDIA GB300 systems

First deployment described as an 18,000 NVIDIA GB300 Grace Blackwell AI supercomputer; delivery date and site undisclosed.

Compute and platform system

Planned or represented

First phase includes NVIDIA InfiniBand networking.

Compute and platform system

Planned or represented

HUMAIN intends to deploy NVIDIA Omniverse as a multi-tenant system for physical AI and digital twins.

Public-source research. Plans may change. See Evidence for sources and qualifications.