3–5 MW
Initial GPU requirements
Search criteria include high-density power, cooling readiness, carrier access and a defined expansion path.
AI & GPU Infrastructure
Multi-megawatt capacity for dense compute
Brightstar sources and qualifies data-center environments for GPU cloud, AI training, inference, sovereign AI and HPC—including liquid-cooled and GB300/NVL72-class requirements.
5–15+ MW
AI factory requirements
Requirement-led sourcing for contiguous wholesale capacity supporting training, inference and customer-backed cloud deployments.
Phased
Large campus requirements
Qualification-led sourcing for phased campuses and longer-term infrastructure programs without implying ready-now inventory.
Technical Qualification
The accelerator does not define the facility by itself
Power, cooling, CDU topology, rack density, network fabric, commissioning, logistics and delivery schedule must be validated as one system.
Liquid cooling
Direct-to-chip design, loop temperatures, flow, water quality, heat rejection and CDU strategy.
Electrical architecture
Critical load, redundancy, substation capacity, phasing and energization evidence.
Network fabric
Carrier entry, cloud connectivity, cross-connects, InfiniBand/Ethernet responsibilities and latency.