ENGLISH OVERVIEW
AI Servers
Accelerators, networking, power and cooling in one system
An AI server integrates accelerators, CPUs, memory, storage, internal links, network interfaces, power and cooling. Clusters also require switches and optical or electrical links between servers; a GPU card alone is not the system.
Board-level and system-level connections
GPU links inside a system, CPU and storage links, and links between servers have different requirements. NVIDIA DGX B200 is a documented example combining GPUs, CPUs, NVSwitch, storage and network devices.
- NVSwitch serves internal GPU connections and is not the same role as a cluster network switch.
Follow the workload to the infrastructure
Bandwidth, latency and topology depend on workload and cluster scale. Compute growth can increase network demand without a fixed linear ratio.
- Track server models, interface speeds, rack power, cooling and architecture before inferring changes to PCB materials or optical links.
Separate PSU and board-level power
PSUs or shelves provide DC output, followed by server-side DC/DC and core regulation. Classify switches, inductors and capacitors by their actual location.
Companies by industry role
Connected industry roles
Links describe supply, classification, technology or services. They do not establish transactions between individual companies.
Printed Circuit Boards → AI Servers
System-board components
Multiple PCB types support server components and interconnects; board counts and materials depend on the system.
NVIDIA | DGX B200 hardware componentsHigh-Layer-Count PCBs → AI Servers
Board option for complex systems
High-layer-count boards serve complex data-center equipment; not every board in every server needs them.
TTM | High-layer-count and processing capabilities for data-center PCBsGPUs and AI Accelerators → AI Servers
Computing component
GPUs and accelerators combine with CPUs, memory, networking, power and cooling in AI systems.
NVIDIA | DGX B200 hardware componentsAI Servers ↔ Network Switches
Networking across nodes
Servers connect through network interfaces and switches; this differs from dedicated GPU interconnects within a system.
NVIDIA | DGX B200 hardware componentsAI Servers → AI Data Centers
Computing systems
AI servers are data-center computing components that work with networking, storage and facilities.
NVIDIA | AI data-center system architectureCooling supply chain ↔ AI Servers
Thermal-load-driven design
Power and density affect server cooling; not all AI servers use the same liquid-cooling architecture.
Delta | Data-center air and liquid coolingData Center Power → AI Servers
System power
Rack and system supplies power servers; conversion and distribution depend on design.
Delta | Data-center power and cooling systemsGlass Substrates for AI, HPC and Interconnects ↔ AI Servers
Large AI-server requirements
A development-application connection, not a purchase claim for a named server or GPU vendor.
AUO | August 31, 2026 GCS, RDL and reliability development with CorningServer and rack cooling integration → AI Servers
Server and rack integration
product-role link, not an inferred company transaction.
QCT | 2026 liquid-cooled AI rack solutionsDRAM modules → AI Servers
Compatible server modules
a role relationship, not a contract among all listed companies.
Nanya | DRAM design, manufacturing and productsSSDs and embedded storage → AI Servers
Workload-specific SSDs
a role relationship, not a contract among all listed companies.
Silicon Motion | NAND controllers, firmware and supported flashAI memory extension and cache ↔ AI Servers
Support compatible AI workloads
a role relationship, not a contract among all listed companies.
Phison | aiDAPTIV cache memory: role and limitationsAI Servers → Compute trays and server modules
Rack compute-unit form factor
Functional or process relationship; not a contract among every company listed.
NVIDIA | MGX architecture and system partnersServer PSU modules → AI Servers
PSUs integrated for individual-server architectures
Material, component or system roles; not proof of supply contracts between all representative companies.
TI | 12 V and 48/54 V server PSU designBoard DC/DC and core regulation → AI Servers
Local regulation for processors
Material, component or system roles; not proof of supply contracts between all representative companies.
Infineon | Power conversion from 48 V to the processor coreActive optical cables (AOC) → AI Servers
Server optical-cable links
Server links select AOC by port, protocol, rate and reach.
Foxlink | AOC productsAI Servers → IT asset retirement and reuse
Retired IT hardware
Servers may enter ITAD on retirement; no named manufacturer contract is implied.
ERI | IT asset disposition and trackingCompany examples & product roles
Examples illustrate roles, not rankings or customer orders. Locations refer to headquarters, principal operations or the stated registration jurisdiction, not every production site.
Quanta ↗
Server design and manufacturing
Design, manufacturing and integration of enterprise servers, storage and data-center systems.
Official product / role sourceWiwynn ↗
Server and rack integration
Servers and rack solutions for cloud, AI and edge computing.
Official product / role sourceFoxconn ↗
AI server manufacturing and integration
Foxconn Group AI servers, computing trays, racks and system integration services.
Official product / role sourceNVIDIA ↗
AI server platforms
DGX systems integrate GPUs, networking, power, cooling and other subsystems.
Official product / role sourceSources
- NVIDIA | DGX B200 system hardware
- NVIDIA | AI-cluster compute and management network topologies
- NVIDIA | DGX B200 system hardware
- NVIDIA | AI-cluster compute and management network topologies
- NVIDIA | NVSwitch internal GPU connections
- NVIDIA | AI Factory power and cooling
- Infineon | Power conversion from 48 V to the processor core
- TI | 12 V and 48/54 V server PSU design