B300 Reserved Capacity in California · From $6.49/GPU/hour
D-DAO

INFRASTRUCTURE TECHNOLOGY

ENGINEEREDAS ONE SYSTEM.

Power, cooling, compute, networking, and storage are engineered as a unified foundation that enables reliable, scalable AI infrastructure.

Integrated Architecture

AI Factory Architecture integrated infrastructure system

POWER DENSITY EVOLUTION

AI COMPUTE IS CHANGINGTHE RULES OF INFRASTRUCTURE.

Every new NVIDIA architecture generation fundamentally reshapes rack power, cooling, electrical distribution, and AI factory design.

Infrastructure requirements across enterprise AI architecture generations
Enterprise AIHGX H100 / B200GB300 NVL72Rubin NVL72
Rack Power10–20 kW30–80 kW120–150 kW250 kW+
Cooling ArchitectureAir CoolingAir / HybridDirect-to-Chip LiquidHigh-Capacity Direct-to-Chip
Power Architecture208 / 415V AC415 / 480V AC480V AC800V DC

Enterprise AI

Rack Power
10–20 kW
Cooling Architecture
Air Cooling
Power Architecture
208 / 415V AC

HGX H100 / B200

Rack Power
30–80 kW
Cooling Architecture
Air / Hybrid
Power Architecture
415 / 480V AC

GB300 NVL72

Rack Power
120–150 kW
Cooling Architecture
Direct-to-Chip Liquid
Power Architecture
480V AC

Rubin NVL72

Rack Power
250 kW+
Cooling Architecture
High-Capacity Direct-to-Chip
Power Architecture
800V DC

The infrastructure challenge is no longer adding more servers.

It's delivering dramatically more power to every AI rack.

ADAPTIVE POWER DELIVERY

DESIGNEDFOR CHANGE.

Power delivery architecture engineered to evolve across future AI generations without rebuilding the AI Factory.

  1. 1Utility Interface
  2. 2Medium Voltage Distribution
  3. 3Power Delivery Architecture
  4. 4Rack Power Distribution
  5. 5AI Rack Interface

Compatible with

Conventional Transformer

Hybrid Power Architecture

Solid-State Transformer (SST)

Designed to support

480V AC

800V DC

without redesigning the facility.

Designed for

GB300

Rubin

Beyond

Built for

Scalability

Flexibility

Long-Term Evolution

  • Adaptive Power Delivery

    Architected to evolve with AI generations.

  • Modular Distribution

    Modular building blocks for every scale.

  • Rack-Ready Infrastructure

    Consistent power from facility to rack.

  • Future Voltage Architecture

    Designed for emerging voltage standards.

Design infrastructure for change—not for replacement.

AI FACTORY COMPOSITION

ONE FACTORY.SEVEN ENGINEERING DOMAINS.

Power, cooling, compute, fabric, storage, platform, and operations work together as one integrated AI Factory.

AI Factory

Power Infrastructure

Utility-to-rack power delivery for increasing AI density.

Cooling Infrastructure

Facility water and direct-to-chip cooling as one thermal system.

AI Compute

Rack-scale accelerated computing for dense AI workloads.

Compute Fabric

High-bandwidth scale-up and scale-out AI networking.

AI Storage

High-throughput data infrastructure for models, datasets, and checkpoints.

AI Platform

Provisioning, scheduling, orchestration, and workload lifecycle management.

Factory Operations

Monitoring, automation, capacity, reliability, and daily operations.

Every AI Factory is engineered as one system.