ODM system designersVendor profile

Pegatron

Verified2026-07-15
Systems listed5

Pegatron designs HGX and NVL systems and has become a favored supplier to neoclouds. It sells complete AI racks with its own thermal engineering, and its willingness to move fast on new NVIDIA platforms has won it visible slots in recent GPU cloud buildouts.

What they design in-house01
  • In-house-designed RA4401-72N1 GB200 NVL72 rack, listed on the Open Compute Project
  • RA4803-72N3: third-generation NVIDIA MGX rack design for Vera Rubin NVL72, built for drop-in deployment into existing Blackwell datacenters
  • AS401-2T0-8H1 (HGX B200, liquid-cooled with in-row CDU) and AS800-2T0-8H1 (air-cooled) deployed at a Maryland datacenter with partner 5C
  • Dedicated Pegatron SVR server business unit distinct from its core Apple-assembly business
  • Co-engineering platform security silicon with Axiado for GB200 NVL72 racks
NVIDIA systems02
ModelPlatformForm factorCooling
RA4401-72N1
OCP-listed design: 36 Grace CPUs, 72 Blackwell GPUs
GB200 NVL7272-GPU rackDLC
RA4803-72N3
Designed for drop-in into existing Blackwell datacenters
Vera Rubin NVL7272-GPU rack (MGX gen 3)DLC
AS401-2T0-8H1
In-row CDU; deployed with 5C in Maryland
HGX B200rackDLC
AS800-2T0-8H1
Air-cooled counterpart
HGX B200rackair
GB300 NVL72 rack
Tri-party build with Together AI and 5C for US datacenters
GB300 NVL7272-GPU rackDLC
NVIDIA relationship03
  • NVIDIA MGX rack-scale partner (GB200 NVL72, GB300 NVL72, Vera Rubin NVL72)
  • NVIDIA HGX B200 partner, air and liquid-cooled variants; NVIDIA-Certified Systems
  • Vera Rubin integration partner announced at COMPUTEX 2026
Manufacturing
  • Taoyuan, Taiwan (home plant)
  • Mexico: new AI-server plant, mass production from Q3 2025, guided to ~5x Taoyuan capacity long-term
  • Texas, US: plant and land bought 2025, operational target 2026
  • Malaysia and India capacity ramps
Field notes
  • Disclosed AI customers: Lambda (GB200 NVL72) and a Together AI plus 5C tri-party build for GB300 NVL72 and HGX B200 racks
  • Best known as Apple's #2 iPhone assembler; the server unit is a newer, fast-growing division on that manufacturing base
  • Heavy US/Mexico build-out explicitly aimed at US AI-infrastructure customers under tariff pressure
Questions we get on this vendor

Is Pegatron an OEM, ODM or integrator for NVIDIA AI servers?

Pegatron sits in the "ODM system designers" tier of our index. In-house-designed RA4401-72N1 GB200 NVL72 rack, listed on the Open Compute Project

Does Pegatron design its NVIDIA servers in-house?

Pegatron designs HGX and NVL systems and has become a favored supplier to neoclouds. It sells complete AI racks with its own thermal engineering, and its willingness to move fast on new NVIDIA platforms has won it visible slots in recent GPU cloud buildouts.

Which NVIDIA platforms does Pegatron ship?

Pegatron ships systems on GB200 NVL72, Vera Rubin NVL72, HGX B200, GB300 NVL72. The systems table on this page lists the specific models.

Where does Pegatron manufacture its AI servers?

Taoyuan, Taiwan (home plant). Mexico: new AI-server plant, mass production from Q3 2025, guided to ~5x Taoyuan capacity long-term. Texas, US: plant and land bought 2025, operational target 2026. Malaysia and India capacity ramps