AN QIPTE. LTD. · SINGAPORE
Contact us
Concept rendering of a flat integrated AI inference GPU package with six memory regions
GPU PACKAGE · CONCEPT RENDERING

SINGAPORE ROOTS. GLOBAL AMBITION.

Engineering the next wave of
AI inference.

From AI inference GPUs to accelerator cards and servers, connecting core compute with enterprise and cloud deployment.

Explore T200M
DESIGN · INTEGRATE · ENABLE 01 / AN QI
OUR FOCUS

AI inference

High-bandwidth memory

2.5D advanced packaging

GPU-to-system solutions

AN QI / THE COMPANY

Focused on inference.
Connected to the world.

AN QI PTE. LTD. is headquartered in the Republic of Singapore, focused on three product lines: AI inference GPUs, accelerator cards and AI inference servers.

Leveraging world-class advanced logic design and 2.5D advanced packaging ecosystems, we bring compute architecture, memory integration, thermal-aware packaging and software enablement into a connected product vision.

COMPANYAN QI PTE. LTD.
UEN202028052C
HEADQUARTERSRepublic of Singapore

01 / THREE CORE PRODUCT LINES

GPU. Accelerator card.
AI server.

Three distinct product lines, from core silicon to board-level acceleration and complete inference systems.

01 / GPU · FLAGSHIP

T200M

Concept rendering of a flat mirrored GPU package with six memory regions
Product concept rendering

AI Inference GPU

An energy-efficient GPU for enterprise and cloud data centers, designed for low-latency, high-concurrency large-model inference.

  • HBM3E high-bandwidth memory architecture
  • 2.5D lidless packaging optimized for thermal integration
  • Precision formats: FP8 / BF16 / FP16 / INT8
HBM3E2.5D · LidlessFP8 / BF16 / FP16 / INT8
02 / ACCELERATOR CARD

AI Accelerator

Concept rendering of a passive-cooled PCIe inference accelerator card
Product concept rendering

Board-level inference acceleration

A separate accelerator card product line that brings GPU compute into standardized board-level solutions for system integration.

  • Board-level integration and reference design
  • Standard PCIe / OAM integration paths
  • Designed for scalable inference deployment
GPU IntegrationPCIe / OAM
03 / AI SERVER

AI Server

Concept rendering of a neatly aligned eight-card air-cooled server with modular fan wall and organized cabling
Eight-card air-cooled server · Concept rendering

System-level AI inference

An AI inference server product line that integrates accelerator cards, system architecture and software for enterprise and data center workloads.

  • Multi-card inference system integration
  • Private large-model deployment
  • Enterprise and cloud data center applications
Multi-cardPrivate AI

NEXT-GENERATION ROADMAP / T300

The next step in AI inference.

T300 is our next-generation high-throughput AI inference acceleration platform roadmap, designed for ultra-large parameter models.

02 / FULL-STACK INFERENCE SOFTWARE STACK

From runtime.
To model deployment.

A layered software architecture connecting core compute with modern inference workloads.

01

Enterprise Linux drivers & Runtime

The foundation for device management, execution and enterprise deployment.

02

Tensor kernels & memory optimization

High-performance FP8 / BF16 tensor kernels and KV-Cache optimization for efficient inference.

03

Inference framework integration

An integration path for vLLM, PyTorch and ONNX Runtime within familiar development workflows.

04

Open-model deployment

Out-of-the-box deployment as the software stack’s goal, from model onboarding to production inference.

03 / GPU-TO-SYSTEM SOLUTIONS

From GPU to system.
More possibilities.

Working with board and system manufacturers to turn core compute into deployable inference solutions.

TIER-1 ADD-IN CARD MANUFACTURERS

For board manufacturers

T200M GPUs and board-level reference designs enable card manufacturers to develop standardized PCIe / OAM accelerator cards.

T200MReference DesignPCIe / OAM
ENTERPRISE DATA CENTERS & AI SERVER OEMs

For systems and data centers

Enable enterprise data centers and AI server OEMs to integrate multi-card inference clusters and deploy private large-model environments.

Multi-card inference Private deployment

04 / CONTACT US

The next generation
starts with a conversation.

Connect with AN QI for GPU, accelerator card and AI server product enquiries and commercial collaboration.

GENERAL ENQUIRIES info@anqisemi.com SALES & BUSINESS sales@anqisemi.com
AN QI PTE. LTD.

Republic of Singapore
UEN: 202028052C

Opens a draft in your email app for you to review and send.