Developer docs
API playgroundTry for free, no card

Search company profiles

Moore Threads

Full company profile

uuid00005wv

Namestring
Moore Threads
Legal namestring
Moore Threads Intelligent Technology (Beijing) Co., Ltd. (摩尔线程智能科技(北京)股份有限公司)
Websiteurl
mthreads.com
Company typeenum
Public
Founded yearint
2020
Descriptiontext

Moore Threads (摩尔线程) is a Chinese full-function GPU designer founded in October 2020 by former Nvidia China general manager Zhang Jianzhong. The company is publicly listed on the Shanghai Stock Exchange STAR Market (stock code 688795) since December 5, 2025, with a post-IPO market capitalization of approximately US$40 billion. Moore Threads designs and sells proprietary GPU accelerators, servers, intelligent computing clusters, and AI software under its MUSA (Meta-computing Unified System Architecture) platform, serving cloud hyperscalers, state-owned enterprises, government agencies, financial institutions, autonomous driving companies, embodied AI developers, broadcast/media firms, and consumer gamers across China.

The company's core technology is the proprietary MUSA full-stack architecture spanning chips (Pinghu PH100 powering MTT S5000, QuYuan, Chunxiao, and the upcoming Huashan), the MUSA SDK 5.1.0 with deep CUDA 12.8 compatibility (3,194 PyTorch operators, 100% math library alignment), MTLink high-speed interconnect, MT Mesh 2.0 GPU virtualization (up to 32-way vGPU splitting), and the KUAE integrated intelligent computing cluster solution. Flagship products include the MTT S5000 (1 PFLOPS FP8, 80GB HBM), MTT S4000, MTT SGX5000 8U server, the 10,000-card KUAE cluster (95% linear scaling efficiency, 60% MFU on dense LLM training), MTT AIBOOK AI laptop (9,999 CNY), MTT AICUBE smart home AI hub, MTT E300 edge AI module, and consumer gaming GPUs (S80, S70, S30). Self-developed AI capabilities include the MooER voice model (MOS 4.51), MusaCoder-27B code LLM (first open-source code LLM trained on domestic GPU stack), DigitalME digital humans, and MT Lambda embodied AI simulation platform.

Moore Threads generates revenue primarily through one-time hardware sales of GPU cards, SGX5000 servers, and KUAE cluster solutions to enterprise customers (China Mobile, ByteDance pipeline, Alibaba Cloud, Pony.ai, BOC, Ministry of Water Resources); commercial software licensing for KUAE Platform and Model Studio under the MUSA Software License Agreement; subscription and usage-based revenue from AI model services (MooER, MUSAChat); channel-driven distribution through a three-tier partner program (Registered, Gold Preferred, Platinum Elite); JD.com flagship store direct-to-consumer sales of gaming GPUs and AIBOOK laptops; and professional services including trial programs, training, and integration. Q1 2026 revenue reached 737.6M yuan (+155% YoY) with the company's first profitable quarter (29.4M yuan net profit), following 2025 full-year revenue growth of 247%.

Short descriptiontext

Moore Threads is a Chinese full-function GPU designer founded in 2020 by former Nvidia China GM Zhang Jianzhong. It builds proprietary MUSA-based chips, servers, and the 10,000-card KUAE cluster, serving cloud, government, finance, and autonomous driving customers in China.

Operating statusenum
Operating
Ownership categoryenum
Headcount rangeband
101–250
akta.pro rankint
HeadquartersBeijing, China
HQ citystring
Beijing
HQ countrystring
China
HQ regionstring
Asia
Markets served

Serves global market

Offices5 records

Each record includes

City, Country, Type, Description, Source

Keyword5 values
AI training chips, GPU accelerators, data center compute, graphics processing units, intelligent computing clusters
Industry7 codes
PRIMARYHDAAAAAA
AI Accelerators (GPUs/TPUs/NPUs/ASICs)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDAAAAAB
AI Server Systems & HGX/Accelerator Platforms
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDACABAG
GPU-Accelerated & AI Training/Inference Servers
SECONDARYHDAHAJAF
AI/Neural Processing Units (NPUs) & ML Accelerators
SECONDARYHDAAAAAI
AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDAAAAAC
AI Inference Hardware & Edge Accelerators
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDAAAAAG
AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
NAICS code5 codes
  • Electronic Computer Manufacturing334111
  • Computer and Peripheral Equipment Manufacturing33411
  • Computer Systems Design and Related Services5415
  • Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services51821
  • Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services5182
SIC code3 codes
  • Electronic Computers3571
  • Services-Computer Integrated Systems Design7373
  • Services-Computer Programming, Data Processing, Etc.7370
Product category
GPU Accelerators and AI Compute Hardware
GTM motion6 records

Each record includes

Type, Description, Source

Revenue model7 records
1Discrete GPU card sales
TypeHardware Sales
Description

Direct sales of consumer/professional GPU cards (MTT S80, S70, X300, S50, S30/S10) via JD.com flagship store and OEM/ODM partners. One-time hardware sales with bundled driver/software.

mthreads.com
2Enterprise AI training/inference card sales (MTT S5000, S4000)
TypeHardware Sales
Description

One-time hardware sales of high-end data center GPU accelerators (S5000 at 1 PFLOPS FP8, 80 GB; S4000 with 48 GB and Tensor cores) typically procured via quote-based enterprise sales for cloud, telco, and government data centers.

mthreads.com
3KUAE cluster solution & SGX5000 server sales
TypeHardware Sales
Description

Integrated intelligent computing center solutions (KUAE Platform, KUAE ModelStudio, server hardware) sold as multi-million-yuan deals to hyperscalers, telcos (China Mobile, China Telecom), state-owned data centers, and large enterprise customers. Includes 30-day cluster delivery.

mthreads.com
4KUAE Platform & Model Studio commercial software licensing
TypeLicensing Royalties
Description

Commercial license required for KUAE Platform and Model Studio (per Software License Agreement: no commercial use, redistribution, hosting, SaaS, or transfer without authorization). Recurring/enterprise licensing model for software stack.

mthreads.com
5AI model & application services (MooER voice, MUSAChat, 摩笔马良, 摩笔天书)
TypeSubscription Recurring
Description

Self-developed AI models offered as services/applications — voice AI (MooER) via voice.mthreads.com, image generation (摩笔马良), content creation (摩笔天书), and embedded digital human assistants — supporting commercial customization and B2B deployment.

6Channel partner distribution (Distributors, VARs, OEM/ODM)
TypeTransaction Fee
Description

Multi-tier distribution through authorized distributors and value-added resellers covering sales, OEM/ODM manufacturing, embedded integration (SI), ISVs, and service providers across China. Tiered benefits include MDF, sales leads, technical training.

mthreads.com
7Professional services: enterprise trial (夸娥试用), solution consulting, integration, training
TypeProfessional Services
Description

Service revenue from free evaluation (KUAE trial), paid consulting, system integration, channel partner enablement training (摩尔学院), and 7×12-hour service hotline (400-667-5666) backed by L2/L3 technical support and 100+ authorized service institutions.

Marketing channels10 records

Each record includes

Title, Type, Stage, Description, Source

Distribution channels6 records

Each record includes

Title, Type, Scope, Target buyer, Description, Source

Cost components6 values
Technology or R&D, Personnel, Supply Chain, Marketing or Sales, Operations, Infrastructure
Pricing details6 tiers
1MTT AIBOOK laptop consumer pricing
ModelOne time/ perpetual licenseBilling cadencePay-as-you-go
Notes

9,999 CNY (approximately US$1,447); available via JD.com and direct online purchase.

tomshardware.com
2MTT S5000 / SGX5000 enterprise AI training & inference cards
ModelUnit PricingBilling cadenceMulti-year contract
Notes

Quote-based; not publicly disclosed. Procurement typically via contact form or sales rep. Includes evaluation trials via KUAE trial program.

mthreads.com
3KUAE (夸娥) intelligent computing cluster solution
ModelHybridBilling cadenceMulti-year contract
Notes

Quote-based integrated delivery including hardware, software, and services with 30-day deployment. Multi-million-yuan deployments common for telco/state customers.

mthreads.com
4KUAE Platform & Model Studio commercial software license
ModelSubscriptionBilling cadenceAnnual
Notes

Commercial license required per EULA for any commercial use, hosting, SaaS, transfer, or redistribution. License terms not publicly disclosed.

mthreads.com
5MooER voice AI API / commercial customization
ModelUsage-basedBilling cadencePay-as-you-go
Notes

Free trial available via voice.mthreads.com; commercial customization and premium voice development offered via sales contact. Pricing quote-based.

mthreads.com
6Channel partner program (Registered / Gold Preferred / Platinum Elite)
ModelTransaction based/ take rateBilling cadenceMulti-year contract
Notes

Multi-tier channel program with revenue share / margin via distributors, VARs, OEM/ODM partners with sales incentives (MDF) at Gold/Platinum levels.

mthreads.com
GTM typeB2B
B2B
Offering typeHardware or Manufacturing
Hardware or Manufacturing
Brand1 of 6 records shown
1MTT
Description

Product brand prefix used across the company's GPU product line (e.g., MTT S5000, MTT S4000, MTT S80, MTT AIBOOK, MTT KUAE, MTT X300).

mthreads.com
+5 more records
Core offering1 text field

Moore Threads designs, manufactures, and sells full-function GPU accelerator hardware (S5000, S4000, S3000, S80, S70, X300, S50, S30/S10), AI training/inference integrated servers (SGX5000), and large-scale intelligent computing cluster solutions (KUAE) built on its proprietary MUSA architecture. The company also provides a complete software stack (MUSA SDK, training/inference suites, vGPU, MUSACODE), edge/endpoint devices (AIBOOK laptop, AICUBE home AI center, E300 module), and AI applications (MooER voice, Maliang drawing, Tianshu content, DigitalME digital human) primarily for Chinese enterprise, government, and prosumer markets.

Differentiator
Functional benefit
Problem solved
Quantifiable outcome1 of 10 values shown
  • Q1 2026 net profit of 29.4M yuan (vs. 112.5M loss prior year)
+9 more records
Product overview1 text field

Moore Threads operates a platform-plus-modules architecture centered on its self-developed MUSA (Meta-computing Unified System Architecture) and full-function GPU technology. The product portfolio spans a comprehensive "cloud-edge-end" (云-边-端) stack: at the cloud level, the flagship MTT KUAE intelligent computing center with MTT S5000 AI training/inference GPU cards (powered by PH100 'Pinghu' architecture) and MTT SGX5000 servers form the foundation for large-scale AI training and inference, capable of 10,000-card cluster configurations; at the edge level, MTT E300 AI modules provide embedded AI compute; at the device/endpoint level, MTT AIBOOK AI laptops and the newly announced MTT AICUBE smart home AI center bring AI to personal devices. The hardware portfolio is complemented by a complete software stack including the MUSA SDK, KUAE Training Suite, KUAE Inference Suite, GPU Virtualization (vGPU), KUAE Cloud Native Suite, MUSACODE AI programming agent, and Moore Perf System profiling tools. Vertical solutions include DigitalME digital humans, AI Reality 3D reconstruction, MT Lambda embodied AI simulation, GPU cloud computer (VDI), digital twin/GIS, broadcast/professional video, and video conferencing solutions. Application-layer products include MooER voice AI, Maliang drawing assistant, and Tianshu content creation. All products share a common architectural foundation based on multiple generations of Moore Threads' proprietary GPU chips: Chunxiao (Spring Dawn), QuYuan (second/third-gen), the current PH100 Pinghu, and the upcoming Huashan AI chip.

Product and service21 records
1MTT S5000 AI Training/Inference GPU Card
CategoryCore product - Data Center GPU
Description

Flagship all-function GPU card powered by PH100 chip on 4th-gen Pinghu MUSA architecture. Delivers up to 1 PFLOPS dense AI compute (FP8), 80GB HBM memory, 1.6TB/s memory bandwidth, and 784GB/s MTLink inter-GPU bandwidth. Supports full FP8-to-FP64 precision for AI training and inference workloads, including trillion-parameter large language models. Targeted at cloud, hyperscale, government, and research data centers.

2MTT S4000 Large Model Computing Accelerator
CategoryCore product - Data Center GPU
Description

Large-model AI computing accelerator based on the third-generation QuYuan MUSA architecture with Tensor cores. Features 48GB memory, 768GB/s bandwidth, and MTLink inter-GPU connectivity at 240GB/s. Supports trillion-parameter LLM training, fine-tuning, and inference with 91%+ linear scaling at 1,000-card cluster configurations.

3MTT S3000 Server GPU
CategoryCore product - Server GPU
Description

Server GPU based on the Chunxiao (Spring Dawn) chip with 4096 MUSA cores at 1.9GHz, 15.5 TFLOPS FP32, 32GB GDDR6 memory, and 448GB/s bandwidth on PCIe Gen5 x16. Supports cloud gaming, cloud rendering, video cloud, digital twin, and meta-computing scenarios with up to 32-way vGPU partitioning per card.

4MTT S2000 Multi-Purpose GPU
CategoryCore product - Server GPU
Description

Multi-purpose GPU supporting AI computing, graphics rendering, physics simulation, scientific computing, and ultra-HD video encoding/decoding. Features 4096 cores, 10.4 TFLOPS FP32, 32GB memory, and supports x86/Arm/LoongArch CPU architectures. Targets cloud desktop, cloud rendering, and cloud gaming workloads.

5MTT S80 Gaming GPU
CategoryConsumer product - Gaming GPU
Description

Gaming GPU featuring the Chunxiao chip with 4096 MUSA cores at 1.8GHz, 14.7 TFLOPS FP32, 16GB GDDR6, and PCIe Gen5 x16 (the industry's first PCIe Gen5 consumer GPU). Supports 4K gaming, AV1/H.264/H.265/VP9 codec, and 8K display output via 3x DisplayPort 1.4a plus 1x HDMI 2.1. Sold via JD.com flagship store.

6MTT S70 Gaming GPU
CategoryConsumer product - Gaming GPU
Description

Mid-range gaming GPU with Chunxiao chip at 1.6GHz, 11.5 TFLOPS FP32, 7GB GDDR6, 392GB/s bandwidth. Supports DirectX 11, 1080P/2K/4K resolutions, and is compatible with major graphics APIs for gaming and content creation.

7MTT X300 Professional GPU
CategoryCore product - Professional GPU
Description

High-performance professional graphics card with Chunxiao architecture, 4096 cores, 14.7 TFLOPS FP32, 16GB GDDR6, PCIe Gen5 x16. Supports 4 displays, 8K output, and 36 channels of 1080P30 video codec. Designed for CAD/BIM/GIS and professional visualization workloads.

8MTT S50 Multi-function Smart GPU
CategoryCore product - Professional GPU
Description

Multi-function smart GPU with 2048 cores, 5.2 TFLOPS FP32, 8GB memory. The world's first GPU to support AV1 format hardware encoding acceleration. Supports DirectX, Vulkan, OpenGL, and OpenGL ES for gaming, GIS, and CAD/BIM applications.

9MTT S30 / S10 Digital Office GPU
CategoryCore product - Office GPU
Description

Compact, low-power GPUs for digital office environments. MTT S30 has 1024 cores, 2.6 TFLOPS FP32, with 2GB/4GB options; MTT S10 has 1024 cores, 2.0 TFLOPS FP32, 4GB. Supports 4K display, OpenGL/OpenGL ES/Vulkan, H.264/H.265 codec, and WebGL online 3D rendering.

10MTT SGX5000 AI Training/Inference Integrated Machine
CategorySystem - Server Hardware
Description

8U rack-mounted AI server optimized for large model training and inference. Features 8 OAM GPU modules, dual x86/C86 CPUs, up to 2 TB DDR5-5600 RDIMM, 8x 400 Gb InfiniBand HDR/Ethernet CX7 adapters. Supports 10,000-card cluster training with linear scaling efficiency up to 95%. Available in liquid-cooled and air-cooled configurations.

11MTT KUAE Intelligent Computing Center
CategoryCore platform - Intelligent Computing Center
Description

Full-stack intelligent computing center solution based on MTT S5000 GPUs and AI training/inference integrated machines. Includes KUAE Platform (Kubernetes-based cluster management), KUAE ModelStudio (LLM training/fine-tuning/inference), and 30-day cluster delivery. Supports 10,000-card cluster scale for trillion-parameter model training with 95% linear scaling efficiency and 60% MFU on Dense LLM training.

12MTT AIBOOK AI Computing Laptop
CategoryEdge/Device - AI Laptop
Description

AI computing laptop powered by Moore Threads' self-developed Yangtze intelligent SoC with 12-core Arm-based CPU, 50 TOPS AI compute, 32GB LPDDR5X 7500MT/s memory, 14-inch 2.8K OLED display. Runs Linux-based AIOS with Windows via virtualization. Priced at 9,999 CNY for developers, AI enthusiasts, and prosumers.

13MTT E300 AI Computing Module
CategoryEdge/Device - AI Module
Description

AI computing module for embedded edge scenarios powered by Yangtze intelligent SoC. Integrates GPU/CPU/NPU heterogeneous compute providing up to 50 TOPS INT8 dense compute. Supports INT8/FP16/FP32/FP64 precision and operates from -20°C to +65°C. Compatible with Ubuntu 22.04, ROS 2, and Android 14.

14MTT AICUBE Smart Home AI Center
CategoryEdge/Device - Smart Home AI
Description

Smart home AI center integrating Intelligent Agent + AI PC + AI NAS capabilities, powered by the Yangtze SoC. Pre-installs 60+ XiaoMai agent skills and supports 36+ APP cross-application control. Features an all-flash AI NAS module for secure local storage and smart data management.

15ZhiYu MoFang (智娱摩方) Gaming PC
CategorySystem - Gaming PC
Description

All-in-one gaming PC platform featuring Moore Threads MTT S80/S70 gaming GPUs. Includes 27-inch 2K 180Hz monitor, mechanical keyboard/mouse, esports headset, and smart streaming camera. Combines gaming capability with AI development support.

16MUSA SDK
CategorySoftware - SDK
Description

Full-featured development platform providing compilers, runtime libraries, debugging tools, and API interfaces for Moore Threads GPUs. Latest version 5.1.0 includes 248 new APIs (761 total), 100% math library alignment with CUDA 12.8, and complete support for all 3,194 PyTorch operators. Distributed under the Moore Threads Software License Agreement.

17MooER (摩耳) Voice AI
CategoryAI Application - Voice Model
Description

Self-developed AI voice large model providing speech synthesis (MOS 4.51), speech recognition (0.6% Mandarin error rate), voice cloning from 30-60s samples with >90% similarity, voice conversion, and customized voice services. Supports streaming low-latency generation. Deployed in digital human, video conferencing, audiobooks, and smart customer service applications.

18Maliang (摩笔马良) AI Drawing Assistant
CategoryAI Application - Image Generation
Description

Intelligent drawing assistant based on self-developed GPU and AI technology. Supports high-quality image generation with Chinese/English bilingual capabilities, image editing, and creative assistance for digital content creators.

19Tianshu (摩笔天书) AI Content Creation
CategoryAI Application - Content Creation
Description

AI content creation system that automatically generates plots, illustrations, subtitles, and music from user-provided titles and outlines, producing high-quality illustrated books or videos.

20DigitalME Digital Human Solution
CategorySolution - Digital Human
Description

Full-stack digital human solution providing customized image/voice/LLM services. Features 30-second voice cloning (>90% similarity), >95% speech recognition accuracy, sub-2-second end-to-end interaction delay, and single-photo 2D digital human generation. Supports GPT/LLaMA/GLM/Qwen/Baichuan/Mixtral model customization for customer service, education, healthcare, and financial services.

21MT Lambda Full-Stack Embodied AI Simulation Platform
CategorySolution - Embodied AI Platform
Description

Full-stack embodied AI simulation platform enabling data synthesis, strategy training, and simulation verification workflows. Includes MT Lambda-Lab for embodied strategy development and training, and MT Lambda-Sim for high-fidelity physics simulation and rendering. Based on all-function GPU architecture with zero-copy data flow between rendering, physics, and AI compute.

Scale indicator12 records

Each record includes

Type, Value, Description, Source

Partnership21 partners
Strategic tierFlagshipTypeTechnology or IntegrationAnnounced on2026-06-17
Description

Day-0 adaptation of Zhipu GLM-5.2 model completed on Moore Threads MTT S5000 GPUs. Demonstrates ability to support flagship Chinese foundation models within hours of release.

Strategic tierFlagshipTypeTechnology or IntegrationAnnounced on2026-06-12
Description

Day-0 support and deep optimization for DeepSeek-V4 (Pro and Flash) and DeepSeek-R1-Distill series on MTT S5000. Native FP8 engine enables training performance improvements of >30% on DeepSeek architectures.

Strategic tierFlagshipTypeTechnology or IntegrationAnnounced on2026-06-12
Description

Day-0 support for MiniMax M3 — the first Chinese frontier-class open-source model with native multimodality, MSA architecture (1M context), and frontier Coding/Agentic capabilities. vLLM-MUSA and SGLang-MUSA both available for deployment.

4Wuxi (Huishan) National Computing Center / 'East Data West Computing' Yangtze River Delta Hub
Strategic tierFlagshipTypeStrategic or Co-development PartnerAnnounced on2026-06-11
Description

Moore Threads provides the core computing power for the Wuxi (Huishan) National Computing Center Phase 1, launched June 11, 2026. Part of the Yangtze River Delta hub of the national '东数西算' (East Data West Computing) project, serving local universities, research institutes, and embodied AI / large model enterprises.

mthreads.com
5Bank of China International Securities (中银证券) & BOC Financial Technology (中银金科)
Strategic tierFlagshipTypeStrategic or Co-development PartnerAnnounced on2026-05-27
Description

Strategic partnership to jointly develop 'AI + Finance' benchmark solution. Brings together domestic GPU compute with major Chinese financial institution, signaling push into regulated financial vertical.

mthreads.com
6Wuxi (Huishan) Industrial Embodied AI Innovation Center — 16 partner consortium
Strategic tierFlagshipTypeStrategic or Co-development PartnerAnnounced on2026-05-14
Description

Moore Threads launched the Wuxi Industrial Embodied AI Innovation Center with 16 partners including universities and tech companies to accelerate industrial embodied AI through cloud-edge-end collaboration. Leverages KUAE cluster and full-function GPUs for large model training, simulation, and edge deployment. Also signed strategic partnership with Guanglun Intelligence (光轮智能) to jointly develop high-confidence simulation data synthesis solutions.

autonews.gasgoo.com
Strategic tierFlagshipTypeTechnology or IntegrationAnnounced on2026-05-10
Description

Official merge of MUSA backend into SGLang mainline — developers can now use SGLang directly with Moore Threads GPUs without third-party adaptation layers. SGLang × MUSA Meetup held May 10, 2026 in Beijing with ~100 developers. Signals China's GPU open-source ecosystem has entered native support phase.

Strategic tierCoreTypeOEM/ Whitelabel/ Licensing PartnerAnnounced on2026-05-09
Description

Ecosystem partnership announced at CTONE's AI Computing Strategy Transformation event. CTONE positioned its Agent Computer as a strategic anchor leveraging Moore Threads GPU ecosystem. Targets 'AI to complete 80%-90% of users' daily tasks'.

Strategic tierFlagshipTypeStrategic or Co-development PartnerAnnounced on2026-02-06
Description

Strategic partnership announced February 6, 2026 to accelerate L4 autonomous driving with domestic AI computing. Pony.ai will adopt Moore Threads' MTT S5000 computing cards and KUAE intelligent computing cluster for world model and in-vehicle model training. First large-scale adoption of domestic AI compute for Pony.ai's critical workloads. Targets deployment of 3,000+ Robotaxi vehicles by end-2026.

1051Sim (Wuyi Vision / 五一视界) — 'Physical AI's first stock'
Strategic tierFlagshipTypeStrategic or Co-development PartnerAnnounced on2025-12-20
Description

Collaboration to build next-generation Physical AI simulation system based on domestic GPU computing power for autonomous driving and embodied intelligence applications. Wuyi Vision operates 51Aes, 51Sim, 51Earth platforms serving 1,000+ enterprises across 19 countries with 30% revenue CAGR.

moomoo.com
11AI-Chip Ecosystem Alliance (StepFun + Huawei Ascend + Cambricon + Biren + Moore Threads)
Strategic tierFlagshipTypeStrategic or Co-development PartnerAnnounced on2025-07-29
Description

Formed in July 2025 with domestic AI model startup StepFun and other major Chinese AI chipmakers (Huawei Ascend, Cambricon, Biren). Alliance optimizes StepFun's foundation model Step 3 for domestic chips to counter Nvidia's renewed presence in China. Signals coordinated push for self-reliance in AI infrastructure.

technode.com
12Beijing Academy of Artificial Intelligence (BAAI / 智源研究院)
Strategic tierFlagshipTypeStrategic or Co-development Partner
Description

Long-term close partnership. Joint completion of RoboBrain 2.5 end-to-end training (Jan 2026) and Qwen3-0.6B training (Mar 2026) on MTT S5000 / FlagOS stack; Day-0 adaptation of DeepSeek-V4 Pro/Flash models (Apr 2026); co-validation of full-stack domestic AI training infrastructure.

mthreads.com
13Alibaba Cloud (Qwen models)
Strategic tierFlagshipTypeTechnology or Integration
Description

Full adaptation of Qwen3.5 and ongoing Day-0 support for Qwen model series on Moore Threads GPUs. Flagship AI chip compatible with Alibaba Cloud model ecosystem.

technode.com
Strategic tierStrategicTypeStrategic or Co-development Partner
Description

In active talks for cloud infrastructure orders alongside Iluvatar CoreX, Biren, MetaX, and Enflame. Part of ByteDance's strategic shift to domestic AI chips following Nvidia export restrictions, having purchased tens of thousands of processors from Chinese AI chipmakers.

Strategic tierFlagshipTypeStrategic or Co-development Partner
Description

Joint cloud computer (云电脑) deployment leveraging MTT S3000 vGPU technology to bring revolutionary performance upgrade to China Mobile's nationwide cloud desktop service. Partnership combines telecom operator scale with domestic GPU infrastructure.

16Open-source ecosystem partners — vLLM, PyTorch, Megatron-LM, Hugging Face
Strategic tierCoreTypeTechnology or Integration
Description

Native support for PyTorch, Megatron-LM, vLLM and SGLang as core MUSA development frameworks. vLLM-MUSA open-sourced. TileLang-MUSA merged upstream. MUSIFY enables automated CUDA-to-MUSA porting. Enables 'zero-cost' code migration.

mthreads.com
17Domestic industrial software vendors — ZWROFT CAD/3D, Haochen CAD, Tiangong CAD, Siver (斯维尔)
Strategic tierCoreTypeTechnology or Integration
Description

Native support and performance optimization for domestic industrial design and manufacturing software on Moore Threads professional GPUs. Partners include ZWROFT CAD/3D, Haochen 2D/3D CAD, Tiangong CAD, and Siver BIM software.

mthreads.com
18Dameisheng (达美盛) — eZWalker platform & Feidu Technology (飞渡科技) — Explore engine
Strategic tierCoreTypeTechnology or Integration
Description

Digital twin partners: Dameisheng's eZWalker platform enables 3D reverse modeling, design optimization, and construction planning. Feidu Technology's Explore engine powers CIM-based urban renewal full lifecycle information modeling.

mthreads.com
Strategic tierCoreTypeTechnology or Integration
Description

Broadcasting and video production partners. Newauto's Sky Eagle non-linear editing system uses MTT X300; Haidong Vision's Daiyan GD2000 multimedia server uses MTT S50; Meishe's smart video editing system on ARM architecture uses MTT S50 with domestic OS.

20Suirui Group (随锐科技)
Strategic tierCoreTypeTechnology or Integration
Description

Suirui's 'Huijian' (会见) video communication cloud split terminal integrates MTT S50 for government/enterprise video conferencing localization, ensuring security for critical information infrastructure industries.

mthreads.com
Strategic tierCoreTypeChannel Partner/ Reseller/ Distributor
Description

Flagship e-commerce distribution partnership. Official JD flagship store at moorethreads.jd.com for direct-to-consumer sales of gaming GPUs (MTT S80, S70, S30, S10, S50) and MTT AIBOOK laptop.

Recent move9 records

Each record includes

Date, Type, Title, Description, Source

Expansion highlight8 records

Each record includes

Type, Description

Peers10 records
TypeBroad incumbent
Description

Dominant global GPU and AI accelerator vendor whose data-center GPUs (H100/H200/Blackwell) define the performance benchmark Moore Threads targets. Moore Threads' CUDA-compatible MUSA architecture and S5000 product line are explicitly positioned as a domestic alternative to Nvidia in China.

TypeDirect peer
Description

Huawei's Ascend series (910B/910C) is the leading domestic AI accelerator in China, sharing the 41% combined domestic AI server share with Moore Threads and Cambricon. Both are competing for the same state-mandated procurement, hyperscaler orders (China Mobile, China Telecom), and inclusion in certified data centers under the Anke framework.

TypeDirect peer
Description

Cambricon is a leading Chinese AI chip designer (one of the AI-Chip Ecosystem Alliance members alongside Moore Threads) specializing in AI training/inference accelerators. It competes head-to-head with Moore Threads for the same Chinese foundation-model and government compute customers, including the A-share market for AI semiconductor investments.

4Biren Technology
TypeDirect peer
Description

Shanghai-based GPU designer developing AI training/inference accelerators competing in the same domestic Chinese AI chip pool as Moore Threads, Huawei, and Cambricon. Listed by ByteDance as a candidate alongside Moore Threads for cloud infrastructure orders as part of the domestic pivot away from Nvidia.

TypeDirect peer
Description

Chinese AI GPU startup competing for the same domestic hyperscaler and government compute demand as Moore Threads. Identified alongside Moore Threads, Biren, MetaX, and Enflame as candidates receiving ByteDance's procurement interest as it shifts away from Nvidia.

TypeDirect peer
Description

Chinese AI accelerator startup integrated into the same domestic AI chip competitive set as Moore Threads. Listed among the Chinese vendors (Moore Threads, Biren, Iluvatar CoreX, Enflame) that ByteDance and other hyperscalers are evaluating to replace Nvidia after US export controls.

7Enflame Technology
TypeDirect peer
Description

Shanghai-based AI training chip designer competing for the same Chinese cloud and data center GPU demand as Moore Threads, with overlapping customer interest from ByteDance and other hyperscalers pivoting to domestic AI compute.

TypeBroad incumbent
Description

Global GPU competitor with Instinct data-center accelerators (MI300/MI325) and Radeon gaming GPUs, serving the same broad customer base as Moore Threads' enterprise (S5000) and consumer (S80) product lines. Represents the international performance benchmark Moore Threads positions itself against in customer pitches.

9Hygon Information
TypeOthers
Description

Domestic Chinese x86-compatible CPU vendor that partners with Moore Threads in the same 'East Data West Computing' and domestic data center deployments (e.g., Wuxi Huishan National Computing Center). Part of the same Legend Capital portfolio and broader domestic compute ecosystem.

10Jingjia Micro
TypeEmerging player
Description

Chinese domestic GPU designer listed on the A-share market, developing GPUs for both AI and graphics workloads. Represents another emerging Chinese player in the same domestic GPU space, though smaller and more graphics-focused than Moore Threads.

Market position
Strengths5 records

Each record includes

Headline, Details, Source

Weaknesses5 records

Each record includes

Headline, Details, Source

Competitive moat6 records

Each record includes

Type, Details

Key risks7 records

Each record includes

Headline, Details, Source

Key highlights7 records

Each record includes

Headline, Details, Source

Customer concentration

Classification, Details

Named customers18 records

Each record includes

Name, Industry, Type, Use case, Source, UUID

Segment8 records

Each record includes

Title, Type, Primary, Description, Pain point addressed, Use case, Source

Ideal customer profile8 records

Each record includes

Profile, Firmographic size, Sales motion, Sales cycle length, Buying structure, Purchase trigger, Buyer persona, Geography, Industry vertical, Primary use case, Description, Pain points, Evidence proof points, Target buyer

Technology focused
Yes
API detail
Has APIbool
Yes

Docs URL, Description

Integration17 records

Each record includes

Title, Type, Description, Source

AI capability14 records

Each record includes

Type, Description, Source

AI maturity
App detail

Has app

Feature10 records

Each record includes

Title, Differentiator, Description, Source

Core technology
Revenue estimate
Valuation estimate
Number of profiles
Profiles5 records

Each record includes

Name, Designation, Designation category, Overview, Profile commentary, Source

No data
Compliance2 records

Each record includes

Name, Class, Description

Funding overview

Funding stage, Last funding date, Total funding USD

Funding rounds7 records

Each record includes

Round, Amount USD, Date, Pre money valuation, Total investors, Investors, News

Investors34 records

Each record includes

Name, Type, Date of entry, Rounds participated, Website

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Each record includes

Name, Acquisition type, Announced date, Completed date, Status, Website, News

Investment

Each record includes

Name, Round, Announced date, Lead investor, Website, News

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Moore Threads

GPU Accelerators and AI Compute Hardwaremthreads.com

Moore Threads is a Chinese full-function GPU designer founded in 2020 by former Nvidia China GM Zhang Jianzhong. It builds proprietary MUSA-based chips, servers, and the 10,000-card KUAE cluster, serving cloud, government, finance, and autonomous driving customers in China.

What Moore Threads does

Moore Threads (摩尔线程) is a Chinese full-function GPU designer founded in October 2020 by former Nvidia China general manager Zhang Jianzhong. The company is publicly listed on the Shanghai Stock Exchange STAR Market (stock code 688795) since December 5, 2025, with a post-IPO market capitalization of approximately US$40 billion. Moore Threads designs and sells proprietary GPU accelerators, servers, intelligent computing clusters, and AI software under its MUSA (Meta-computing Unified System Architecture) platform, serving cloud hyperscalers, state-owned enterprises, government agencies, financial institutions, autonomous driving companies, embodied AI developers, broadcast/media firms, and consumer gamers across China.

The company's core technology is the proprietary MUSA full-stack architecture spanning chips (Pinghu PH100 powering MTT S5000, QuYuan, Chunxiao, and the upcoming Huashan), the MUSA SDK 5.1.0 with deep CUDA 12.8 compatibility (3,194 PyTorch operators, 100% math library alignment), MTLink high-speed interconnect, MT Mesh 2.0 GPU virtualization (up to 32-way vGPU splitting), and the KUAE integrated intelligent computing cluster solution. Flagship products include the MTT S5000 (1 PFLOPS FP8, 80GB HBM), MTT S4000, MTT SGX5000 8U server, the 10,000-card KUAE cluster (95% linear scaling efficiency, 60% MFU on dense LLM training), MTT AIBOOK AI laptop (9,999 CNY), MTT AICUBE smart home AI hub, MTT E300 edge AI module, and consumer gaming GPUs (S80, S70, S30). Self-developed AI capabilities include the MooER voice model (MOS 4.51), MusaCoder-27B code LLM (first open-source code LLM trained on domestic GPU stack), DigitalME digital humans, and MT Lambda embodied AI simulation platform.

Moore Threads generates revenue primarily through one-time hardware sales of GPU cards, SGX5000 servers, and KUAE cluster solutions to enterprise customers (China Mobile, ByteDance pipeline, Alibaba Cloud, Pony.ai, BOC, Ministry of Water Resources); commercial software licensing for KUAE Platform and Model Studio under the MUSA Software License Agreement; subscription and usage-based revenue from AI model services (MooER, MUSAChat); channel-driven distribution through a three-tier partner program (Registered, Gold Preferred, Platinum Elite); JD.com flagship store direct-to-consumer sales of gaming GPUs and AIBOOK laptops; and professional services including trial programs, training, and integration. Q1 2026 revenue reached 737.6M yuan (+155% YoY) with the company's first profitable quarter (29.4M yuan net profit), following 2025 full-year revenue growth of 247%.

Moore Threads firmographics

Firmographics
Name
Moore Threads
Legal name
Moore Threads Intelligent Technology (Beijing) Co., Ltd. (摩尔线程智能科技(北京)股份有限公司)
Website
https://mthreads.com
Company type
Public
Founded year
2020
Operating status
Operating
Headcount range
101–250 employees
Short description
Moore Threads is a Chinese full-function GPU designer founded in 2020 by former Nvidia China GM Zhang Jianzhong. It builds proprietary MUSA-based chips, servers, and the 10,000-card KUAE cluster, serving cloud, government, finance, and autonomous driving customers in China.
Ownership category
akta.pro rank

Moore Threads industry classification

Industry
Product category
GPU Accelerators and AI Compute Hardware
NAICS
Electronic Computer Manufacturing (334111), Computer and Peripheral Equipment Manufacturing (33411), Computer Systems Design and Related Services (5415), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (51821), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)
SIC
Electronic Computers (3571), Services-Computer Integrated Systems Design (7373), Services-Computer Programming, Data Processing, Etc. (7370)
akta.pro industry
PRIMARYHDAAAAAA
AI Accelerators (GPUs/TPUs/NPUs/ASICs)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDAAAAAB
AI Server Systems & HGX/Accelerator Platforms
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDACABAG
GPU-Accelerated & AI Training/Inference Servers
SECONDARYHDAHAJAF
AI/Neural Processing Units (NPUs) & ML Accelerators
SECONDARYHDAAAAAI
AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDAAAAAC
AI Inference Hardware & Edge Accelerators
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)
SECONDARYHDAAAAAG
AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)

Keywords

  • AI training chips
  • GPU accelerators
  • Data center compute
  • Graphics processing units
  • Intelligent computing clusters

Where Moore Threads is headquartered

Location

Headquarters

HQ city
Beijing
HQ country
China
HQ region
Asia

Offices5 records

Markets served

Moore Threads business model

Business model
GTM type
B2B
Offering type
Hardware or Manufacturing
Cost components
Technology or R&D, Personnel, Supply Chain, Marketing or Sales, Operations, Infrastructure

Revenue model

  1. Discrete GPU card sales: Direct sales of consumer/professional GPU cards (MTT S80, S70, X300, S50, S30/S10) via JD.com flagship store and OEM/ODM partners. One-time hardware sales with bundled driver/software.
  2. Enterprise AI training/inference card sales (MTT S5000, S4000): One-time hardware sales of high-end data center GPU accelerators (S5000 at 1 PFLOPS FP8, 80 GB; S4000 with 48 GB and Tensor cores) typically procured via quote-based enterprise sales for cloud, telco, and government data centers.
  3. KUAE cluster solution & SGX5000 server sales: Integrated intelligent computing center solutions (KUAE Platform, KUAE ModelStudio, server hardware) sold as multi-million-yuan deals to hyperscalers, telcos (China Mobile, China Telecom), state-owned data centers, and large enterprise customers. Includes 30-day cluster delivery.
  4. KUAE Platform & Model Studio commercial software licensing: Commercial license required for KUAE Platform and Model Studio (per Software License Agreement: no commercial use, redistribution, hosting, SaaS, or transfer without authorization). Recurring/enterprise licensing model for software stack.
  5. AI model & application services (MooER voice, MUSAChat, 摩笔马良, 摩笔天书): Self-developed AI models offered as services/applications — voice AI (MooER) via voice.mthreads.com, image generation (摩笔马良), content creation (摩笔天书), and embedded digital human assistants — supporting commercial customization and B2B deployment.
  6. Channel partner distribution (Distributors, VARs, OEM/ODM): Multi-tier distribution through authorized distributors and value-added resellers covering sales, OEM/ODM manufacturing, embedded integration (SI), ISVs, and service providers across China. Tiered benefits include MDF, sales leads, technical training.
  7. Professional services: enterprise trial (夸娥试用), solution consulting, integration, training: Service revenue from free evaluation (KUAE trial), paid consulting, system integration, channel partner enablement training (摩尔学院), and 7×12-hour service hotline (400-667-5666) backed by L2/L3 technical support and 100+ authorized service institutions.

Pricing tiers

ModelBillingPrice
One time/ perpetual licensePay-as-you-goMTT AIBOOK laptop consumer pricing
Unit PricingMulti-year contractMTT S5000 / SGX5000 enterprise AI training & inference cards
HybridMulti-year contractKUAE (夸娥) intelligent computing cluster solution
SubscriptionAnnualKUAE Platform & Model Studio commercial software license
Usage-basedPay-as-you-goMooER voice AI API / commercial customization
Transaction based/ take rateMulti-year contractChannel partner program (Registered / Gold Preferred / Platinum Elite)

Go-to-market motion6 records

Distribution channels6 records

Marketing channels10 records

Moore Threads product offering

Product offering

Core offering

Moore Threads designs, manufactures, and sells full-function GPU accelerator hardware (S5000, S4000, S3000, S80, S70, X300, S50, S30/S10), AI training/inference integrated servers (SGX5000), and large-scale intelligent computing cluster solutions (KUAE) built on its proprietary MUSA architecture. The company also provides a complete software stack (MUSA SDK, training/inference suites, vGPU, MUSACODE), edge/endpoint devices (AIBOOK laptop, AICUBE home AI center, E300 module), and AI applications (MooER voice, Maliang drawing, Tianshu content, DigitalME digital human) primarily for Chinese enterprise, government, and prosumer markets.

Product overview

Moore Threads operates a platform-plus-modules architecture centered on its self-developed MUSA (Meta-computing Unified System Architecture) and full-function GPU technology. The product portfolio spans a comprehensive "cloud-edge-end" (云-边-端) stack: at the cloud level, the flagship MTT KUAE intelligent computing center with MTT S5000 AI training/inference GPU cards (powered by PH100 'Pinghu' architecture) and MTT SGX5000 servers form the foundation for large-scale AI training and inference, capable of 10,000-card cluster configurations; at the edge level, MTT E300 AI modules provide embedded AI compute; at the device/endpoint level, MTT AIBOOK AI laptops and the newly announced MTT AICUBE smart home AI center bring AI to personal devices. The hardware portfolio is complemented by a complete software stack including the MUSA SDK, KUAE Training Suite, KUAE Inference Suite, GPU Virtualization (vGPU), KUAE Cloud Native Suite, MUSACODE AI programming agent, and Moore Perf System profiling tools. Vertical solutions include DigitalME digital humans, AI Reality 3D reconstruction, MT Lambda embodied AI simulation, GPU cloud computer (VDI), digital twin/GIS, broadcast/professional video, and video conferencing solutions. Application-layer products include MooER voice AI, Maliang drawing assistant, and Tianshu content creation. All products share a common architectural foundation based on multiple generations of Moore Threads' proprietary GPU chips: Chunxiao (Spring Dawn), QuYuan (second/third-gen), the current PH100 Pinghu, and the upcoming Huashan AI chip.

Differentiator

Problem solved

Functional benefit

Brands

  • MTT: Product brand prefix used across the company's GPU product line (e.g., MTT S5000, MTT S4000, MTT S80, MTT AIBOOK, MTT KUAE, MTT X300).
  • KUAE (夸娥)
  • DigitalME
  • MooER (摩耳)
  • MT Lambda
  • MusaCoder

Products and services

  • MTT S5000 AI Training/Inference GPU Card Flagship all-function GPU card powered by PH100 chip on 4th-gen Pinghu MUSA architecture. Delivers up to 1 PFLOPS dense AI compute (FP8), 80GB HBM memory, 1.6TB/s memory bandwidth, and 784GB/s MTLink inter-GPU bandwidth. Supports full FP8-to-FP64 precision for AI training and inference workloads, including trillion-parameter large language models. Targeted at cloud, hyperscale, government, and research data centers.
  • MTT S4000 Large Model Computing Accelerator Large-model AI computing accelerator based on the third-generation QuYuan MUSA architecture with Tensor cores. Features 48GB memory, 768GB/s bandwidth, and MTLink inter-GPU connectivity at 240GB/s. Supports trillion-parameter LLM training, fine-tuning, and inference with 91%+ linear scaling at 1,000-card cluster configurations.
  • MTT S3000 Server GPU Server GPU based on the Chunxiao (Spring Dawn) chip with 4096 MUSA cores at 1.9GHz, 15.5 TFLOPS FP32, 32GB GDDR6 memory, and 448GB/s bandwidth on PCIe Gen5 x16. Supports cloud gaming, cloud rendering, video cloud, digital twin, and meta-computing scenarios with up to 32-way vGPU partitioning per card.
  • MTT S2000 Multi-Purpose GPU Multi-purpose GPU supporting AI computing, graphics rendering, physics simulation, scientific computing, and ultra-HD video encoding/decoding. Features 4096 cores, 10.4 TFLOPS FP32, 32GB memory, and supports x86/Arm/LoongArch CPU architectures. Targets cloud desktop, cloud rendering, and cloud gaming workloads.
  • MTT S80 Gaming GPU Gaming GPU featuring the Chunxiao chip with 4096 MUSA cores at 1.8GHz, 14.7 TFLOPS FP32, 16GB GDDR6, and PCIe Gen5 x16 (the industry's first PCIe Gen5 consumer GPU). Supports 4K gaming, AV1/H.264/H.265/VP9 codec, and 8K display output via 3x DisplayPort 1.4a plus 1x HDMI 2.1. Sold via JD.com flagship store.
  • MTT S70 Gaming GPU Mid-range gaming GPU with Chunxiao chip at 1.6GHz, 11.5 TFLOPS FP32, 7GB GDDR6, 392GB/s bandwidth. Supports DirectX 11, 1080P/2K/4K resolutions, and is compatible with major graphics APIs for gaming and content creation.
  • MTT X300 Professional GPU High-performance professional graphics card with Chunxiao architecture, 4096 cores, 14.7 TFLOPS FP32, 16GB GDDR6, PCIe Gen5 x16. Supports 4 displays, 8K output, and 36 channels of 1080P30 video codec. Designed for CAD/BIM/GIS and professional visualization workloads.
  • MTT S50 Multi-function Smart GPU Multi-function smart GPU with 2048 cores, 5.2 TFLOPS FP32, 8GB memory. The world's first GPU to support AV1 format hardware encoding acceleration. Supports DirectX, Vulkan, OpenGL, and OpenGL ES for gaming, GIS, and CAD/BIM applications.
  • MTT S30 / S10 Digital Office GPU Compact, low-power GPUs for digital office environments. MTT S30 has 1024 cores, 2.6 TFLOPS FP32, with 2GB/4GB options; MTT S10 has 1024 cores, 2.0 TFLOPS FP32, 4GB. Supports 4K display, OpenGL/OpenGL ES/Vulkan, H.264/H.265 codec, and WebGL online 3D rendering.
  • MTT SGX5000 AI Training/Inference Integrated Machine 8U rack-mounted AI server optimized for large model training and inference. Features 8 OAM GPU modules, dual x86/C86 CPUs, up to 2 TB DDR5-5600 RDIMM, 8x 400 Gb InfiniBand HDR/Ethernet CX7 adapters. Supports 10,000-card cluster training with linear scaling efficiency up to 95%. Available in liquid-cooled and air-cooled configurations.
  • MTT KUAE Intelligent Computing Center Full-stack intelligent computing center solution based on MTT S5000 GPUs and AI training/inference integrated machines. Includes KUAE Platform (Kubernetes-based cluster management), KUAE ModelStudio (LLM training/fine-tuning/inference), and 30-day cluster delivery. Supports 10,000-card cluster scale for trillion-parameter model training with 95% linear scaling efficiency and 60% MFU on Dense LLM training.
  • MTT AIBOOK AI Computing Laptop AI computing laptop powered by Moore Threads' self-developed Yangtze intelligent SoC with 12-core Arm-based CPU, 50 TOPS AI compute, 32GB LPDDR5X 7500MT/s memory, 14-inch 2.8K OLED display. Runs Linux-based AIOS with Windows via virtualization. Priced at 9,999 CNY for developers, AI enthusiasts, and prosumers.
  • MTT E300 AI Computing Module AI computing module for embedded edge scenarios powered by Yangtze intelligent SoC. Integrates GPU/CPU/NPU heterogeneous compute providing up to 50 TOPS INT8 dense compute. Supports INT8/FP16/FP32/FP64 precision and operates from -20°C to +65°C. Compatible with Ubuntu 22.04, ROS 2, and Android 14.
  • MTT AICUBE Smart Home AI Center Smart home AI center integrating Intelligent Agent + AI PC + AI NAS capabilities, powered by the Yangtze SoC. Pre-installs 60+ XiaoMai agent skills and supports 36+ APP cross-application control. Features an all-flash AI NAS module for secure local storage and smart data management.
  • ZhiYu MoFang (智娱摩方) Gaming PC All-in-one gaming PC platform featuring Moore Threads MTT S80/S70 gaming GPUs. Includes 27-inch 2K 180Hz monitor, mechanical keyboard/mouse, esports headset, and smart streaming camera. Combines gaming capability with AI development support.
  • MUSA SDK Full-featured development platform providing compilers, runtime libraries, debugging tools, and API interfaces for Moore Threads GPUs. Latest version 5.1.0 includes 248 new APIs (761 total), 100% math library alignment with CUDA 12.8, and complete support for all 3,194 PyTorch operators. Distributed under the Moore Threads Software License Agreement.
  • MooER (摩耳) Voice AI Self-developed AI voice large model providing speech synthesis (MOS 4.51), speech recognition (0.6% Mandarin error rate), voice cloning from 30-60s samples with >90% similarity, voice conversion, and customized voice services. Supports streaming low-latency generation. Deployed in digital human, video conferencing, audiobooks, and smart customer service applications.
  • Maliang (摩笔马良) AI Drawing Assistant Intelligent drawing assistant based on self-developed GPU and AI technology. Supports high-quality image generation with Chinese/English bilingual capabilities, image editing, and creative assistance for digital content creators.
  • Tianshu (摩笔天书) AI Content Creation AI content creation system that automatically generates plots, illustrations, subtitles, and music from user-provided titles and outlines, producing high-quality illustrated books or videos.
  • DigitalME Digital Human Solution Full-stack digital human solution providing customized image/voice/LLM services. Features 30-second voice cloning (>90% similarity), >95% speech recognition accuracy, sub-2-second end-to-end interaction delay, and single-photo 2D digital human generation. Supports GPT/LLaMA/GLM/Qwen/Baichuan/Mixtral model customization for customer service, education, healthcare, and financial services.
  • MT Lambda Full-Stack Embodied AI Simulation Platform Full-stack embodied AI simulation platform enabling data synthesis, strategy training, and simulation verification workflows. Includes MT Lambda-Lab for embodied strategy development and training, and MT Lambda-Sim for high-fidelity physics simulation and rendering. Based on all-function GPU architecture with zero-copy data flow between rendering, physics, and AI compute.

Quantifiable outcome

  • Q1 2026 net profit of 29.4M yuan (vs. 112.5M loss prior year)
  • +9 more outcomes

Companies that use Moore Threads

Customer profile

Named customers18 records

Segments8 records

Ideal customer profiles8 records

Moore Threads technology and API

Technology

Technology focussed Yes

API detail

Has API
Yes
API docs
API detail

Core technology

AI maturity

App detail

Integration17 records

AI capability14 records

Feature10 records

Moore Threads partnerships and signals

Strategic signal

Partnerships

21 partnerships are on record, tiered flagship, core and strategic.

  • Zhipu AI (智谱)flagshipTechnology or Integration · 17 June 2026Day-0 adaptation of Zhipu GLM-5.2 model completed on Moore Threads MTT S5000 GPUs. Demonstrates ability to support flagship Chinese foundation models within hours of release.
  • DeepSeekflagshipTechnology or Integration · 12 June 2026Day-0 support and deep optimization for DeepSeek-V4 (Pro and Flash) and DeepSeek-R1-Distill series on MTT S5000. Native FP8 engine enables training performance improvements of >30% on DeepSeek architectures.
  • MiniMax (M3 model)flagshipTechnology or Integration · 12 June 2026Day-0 support for MiniMax M3 — the first Chinese frontier-class open-source model with native multimodality, MSA architecture (1M context), and frontier Coding/Agentic capabilities. vLLM-MUSA and SGLang-MUSA both available for deployment.
  • Wuxi (Huishan) National Computing Center / 'East Data West Computing' Yangtze River Delta HubflagshipStrategic or Co-development Partner · 11 June 2026Moore Threads provides the core computing power for the Wuxi (Huishan) National Computing Center Phase 1, launched June 11, 2026. Part of the Yangtze River Delta hub of the national '东数西算' (East Data West Computing) project, serving local universities, research institutes, and embodied AI / large model enterprises.
  • Bank of China International Securities (中银证券) & BOC Financial Technology (中银金科)flagshipStrategic or Co-development Partner · 27 May 2026Strategic partnership to jointly develop 'AI + Finance' benchmark solution. Brings together domestic GPU compute with major Chinese financial institution, signaling push into regulated financial vertical.
  • Wuxi (Huishan) Industrial Embodied AI Innovation Center — 16 partner consortiumflagshipStrategic or Co-development Partner · 14 May 2026Moore Threads launched the Wuxi Industrial Embodied AI Innovation Center with 16 partners including universities and tech companies to accelerate industrial embodied AI through cloud-edge-end collaboration. Leverages KUAE cluster and full-function GPUs for large model training, simulation, and edge deployment. Also signed strategic partnership with Guanglun Intelligence (光轮智能) to jointly develop high-confidence simulation data synthesis solutions.
  • SGLang communityflagshipTechnology or Integration · 10 May 2026Official merge of MUSA backend into SGLang mainline — developers can now use SGLang directly with Moore Threads GPUs without third-party adaptation layers. SGLang × MUSA Meetup held May 10, 2026 in Beijing with ~100 developers. Signals China's GPU open-source ecosystem has entered native support phase.
  • CTONE GroupcoreOEM/ Whitelabel/ Licensing Partner · 9 May 2026Ecosystem partnership announced at CTONE's AI Computing Strategy Transformation event. CTONE positioned its Agent Computer as a strategic anchor leveraging Moore Threads GPU ecosystem. Targets 'AI to complete 80%-90% of users' daily tasks'.
  • Pony.aiflagshipStrategic or Co-development Partner · 6 February 2026Strategic partnership announced February 6, 2026 to accelerate L4 autonomous driving with domestic AI computing. Pony.ai will adopt Moore Threads' MTT S5000 computing cards and KUAE intelligent computing cluster for world model and in-vehicle model training. First large-scale adoption of domestic AI compute for Pony.ai's critical workloads. Targets deployment of 3,000+ Robotaxi vehicles by end-2026.
  • 51Sim (Wuyi Vision / 五一视界) — 'Physical AI's first stock'flagshipStrategic or Co-development Partner · 20 December 2025Collaboration to build next-generation Physical AI simulation system based on domestic GPU computing power for autonomous driving and embodied intelligence applications. Wuyi Vision operates 51Aes, 51Sim, 51Earth platforms serving 1,000+ enterprises across 19 countries with 30% revenue CAGR.
  • AI-Chip Ecosystem Alliance (StepFun + Huawei Ascend + Cambricon + Biren + Moore Threads)flagshipStrategic or Co-development Partner · 29 July 2025Formed in July 2025 with domestic AI model startup StepFun and other major Chinese AI chipmakers (Huawei Ascend, Cambricon, Biren). Alliance optimizes StepFun's foundation model Step 3 for domestic chips to counter Nvidia's renewed presence in China. Signals coordinated push for self-reliance in AI infrastructure.
  • Beijing Academy of Artificial Intelligence (BAAI / 智源研究院)flagshipStrategic or Co-development PartnerLong-term close partnership. Joint completion of RoboBrain 2.5 end-to-end training (Jan 2026) and Qwen3-0.6B training (Mar 2026) on MTT S5000 / FlagOS stack; Day-0 adaptation of DeepSeek-V4 Pro/Flash models (Apr 2026); co-validation of full-stack domestic AI training infrastructure.
  • Alibaba Cloud (Qwen models)flagshipTechnology or IntegrationFull adaptation of Qwen3.5 and ongoing Day-0 support for Qwen model series on Moore Threads GPUs. Flagship AI chip compatible with Alibaba Cloud model ecosystem.
  • ByteDancestrategicStrategic or Co-development PartnerIn active talks for cloud infrastructure orders alongside Iluvatar CoreX, Biren, MetaX, and Enflame. Part of ByteDance's strategic shift to domestic AI chips following Nvidia export restrictions, having purchased tens of thousands of processors from Chinese AI chipmakers.
  • China Mobile (Smart Family Operation Center)flagshipStrategic or Co-development PartnerJoint cloud computer (云电脑) deployment leveraging MTT S3000 vGPU technology to bring revolutionary performance upgrade to China Mobile's nationwide cloud desktop service. Partnership combines telecom operator scale with domestic GPU infrastructure.
  • Open-source ecosystem partners — vLLM, PyTorch, Megatron-LM, Hugging FacecoreTechnology or IntegrationNative support for PyTorch, Megatron-LM, vLLM and SGLang as core MUSA development frameworks. vLLM-MUSA open-sourced. TileLang-MUSA merged upstream. MUSIFY enables automated CUDA-to-MUSA porting. Enables 'zero-cost' code migration.
  • Domestic industrial software vendors — ZWROFT CAD/3D, Haochen CAD, Tiangong CAD, Siver (斯维尔)coreTechnology or IntegrationNative support and performance optimization for domestic industrial design and manufacturing software on Moore Threads professional GPUs. Partners include ZWROFT CAD/3D, Haochen 2D/3D CAD, Tiangong CAD, and Siver BIM software.
  • Dameisheng (达美盛) — eZWalker platform & Feidu Technology (飞渡科技) — Explore enginecoreTechnology or IntegrationDigital twin partners: Dameisheng's eZWalker platform enables 3D reverse modeling, design optimization, and construction planning. Feidu Technology's Explore engine powers CIM-based urban renewal full lifecycle information modeling.
  • Newauto (新奥特), Haidong Vision (嗨动视觉), Meishe (美摄)coreTechnology or IntegrationBroadcasting and video production partners. Newauto's Sky Eagle non-linear editing system uses MTT X300; Haidong Vision's Daiyan GD2000 multimedia server uses MTT S50; Meishe's smart video editing system on ARM architecture uses MTT S50 with domestic OS.
  • Suirui Group (随锐科技)coreTechnology or IntegrationSuirui's 'Huijian' (会见) video communication cloud split terminal integrates MTT S50 for government/enterprise video conferencing localization, ensuring security for critical information infrastructure industries.
  • JD.com (京东)coreChannel Partner/ Reseller/ DistributorFlagship e-commerce distribution partnership. Official JD flagship store at moorethreads.jd.com for direct-to-consumer sales of gaming GPUs (MTT S80, S70, S30, S10, S50) and MTT AIBOOK laptop.

Scale indicators12 records

Recent moves9 records

Expansion highlights8 records

Moore Threads competitors and assessment

Company assessment

Broad incumbents

  • Nvidia: Dominant global GPU and AI accelerator vendor whose data-center GPUs (H100/H200/Blackwell) define the performance benchmark Moore Threads targets. Moore Threads' CUDA-compatible MUSA architecture and S5000 product line are explicitly positioned as a domestic alternative to Nvidia in China.
  • AMD: Global GPU competitor with Instinct data-center accelerators (MI300/MI325) and Radeon gaming GPUs, serving the same broad customer base as Moore Threads' enterprise (S5000) and consumer (S80) product lines. Represents the international performance benchmark Moore Threads positions itself against in customer pitches.

Direct peers

  • Huawei (Ascend AI chips): Huawei's Ascend series (910B/910C) is the leading domestic AI accelerator in China, sharing the 41% combined domestic AI server share with Moore Threads and Cambricon. Both are competing for the same state-mandated procurement, hyperscaler orders (China Mobile, China Telecom), and inclusion in certified data centers under the Anke framework.
  • Cambricon Technologies: Cambricon is a leading Chinese AI chip designer (one of the AI-Chip Ecosystem Alliance members alongside Moore Threads) specializing in AI training/inference accelerators. It competes head-to-head with Moore Threads for the same Chinese foundation-model and government compute customers, including the A-share market for AI semiconductor investments.
  • Biren Technology: Shanghai-based GPU designer developing AI training/inference accelerators competing in the same domestic Chinese AI chip pool as Moore Threads, Huawei, and Cambricon. Listed by ByteDance as a candidate alongside Moore Threads for cloud infrastructure orders as part of the domestic pivot away from Nvidia.
  • Iluvatar CoreX: Chinese AI GPU startup competing for the same domestic hyperscaler and government compute demand as Moore Threads. Identified alongside Moore Threads, Biren, MetaX, and Enflame as candidates receiving ByteDance's procurement interest as it shifts away from Nvidia.
  • MetaX: Chinese AI accelerator startup integrated into the same domestic AI chip competitive set as Moore Threads. Listed among the Chinese vendors (Moore Threads, Biren, Iluvatar CoreX, Enflame) that ByteDance and other hyperscalers are evaluating to replace Nvidia after US export controls.
  • Enflame Technology: Shanghai-based AI training chip designer competing for the same Chinese cloud and data center GPU demand as Moore Threads, with overlapping customer interest from ByteDance and other hyperscalers pivoting to domestic AI compute.

Others

  • Hygon Information: Domestic Chinese x86-compatible CPU vendor that partners with Moore Threads in the same 'East Data West Computing' and domestic data center deployments (e.g., Wuxi Huishan National Computing Center). Part of the same Legend Capital portfolio and broader domestic compute ecosystem.

Emerging players

  • Jingjia Micro: Chinese domestic GPU designer listed on the A-share market, developing GPUs for both AI and graphics workloads. Represents another emerging Chinese player in the same domestic GPU space, though smaller and more graphics-focused than Moore Threads.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks7 records

Key highlights7 records

Customer concentration

Moore Threads social profiles

Digital presence

Moore Threads compliance and trust

Trust signal

Compliance2 records

Moore Threads financial estimates

Financial estimate

Revenue estimate

Valuation estimate

Moore Threads leadership team

Management profile

Number of profiles

Profiles5 records

Moore Threads funding detail

Funding detail

Funding overview

Funding rounds7 records

Investors34 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

Moore Threads M&A and investment

M&A and investment

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Moore Threads

What does Moore Threads do?

Moore Threads designs, manufactures, and sells full-function GPU accelerator hardware (S5000, S4000, S3000, S80, S70, X300, S50, S30/S10), AI training/inference integrated servers (SGX5000), and large-scale intelligent computing cluster solutions (KUAE) built on its proprietary MUSA architecture. The company also provides a complete software stack (MUSA SDK, training/inference suites, vGPU, MUSACODE), edge/endpoint devices (AIBOOK laptop, AICUBE home AI center, E300 module), and AI applications (MooER voice, Maliang drawing, Tianshu content, DigitalME digital human) primarily for Chinese enterprise, government, and prosumer markets.

Is Moore Threads a public or private company?

Moore Threads is a public company. It is classified as public and is currently operating.

When was Moore Threads founded?

Moore Threads was founded in 2020. It employs 101 to 250 people.

Where is Moore Threads based?

Moore Threads is headquartered in Beijing, China, in the Asia region.

How does Moore Threads make money?

Seven revenue lines are on record. Discrete GPU card sales are the primary driver. The others are enterprise AI training/inference card sales (MTT S5000, S4000), KUAE cluster solution & SGX5000 server sales, KUAE Platform & Model Studio commercial software licensing, AI model & application services (MooER voice, MUSAChat, 摩笔马良, 摩笔天书), channel partner distribution (Distributors, VARs, OEM/ODM) and professional services: enterprise trial (夸娥试用), solution consulting, integration, training.

Who are Moore Threads's main competitors?

Broad incumbents on record are Nvidia and AMD. Direct peers are Huawei (Ascend AI chips), Cambricon Technologies, Biren Technology, Iluvatar CoreX, MetaX and Enflame Technology. Hygon Information is listed as an others. Jingjia Micro is listed as an emerging player.

Does Moore Threads have an API?

Yes. Moore Threads offers the MUSA SDK as a full-featured development platform for its GPUs, providing compilers, runtime libraries, debugging tools, and API interfaces that allow developers to build and deploy optimized parallel computing applications. The MUSA SDK includes deep learning acceleration libraries, mathematical computation libraries, and is designed to be compatible with CUDA via the MUSIFY migration tool. Additionally, MUSACODE provides an AI programming agent with natural language interaction for code generation, project analysis, debugging/refactoring, and high-performance Kernel/TileLang operator development, supporting multiple programming languages including Python, JavaScript, TypeScript, Go, Rust, Java, and C++. The MUSA Deploy tool supports automated installation, uninstallation, updates, and environment detection for the MUSA software stack. Developer documentation is at developer.mthreads.com.

What industry is Moore Threads in?

Moore Threads's product category is GPU Accelerators and AI Compute Hardware. Its primary akta.pro industry code is HDAAAAAA, AI Accelerators (GPUs/TPUs/NPUs/ASICs), with a secondary code of HDAAAAAB, AI Server Systems & HGX/Accelerator Platforms. Its NAICS code is 334111 and its SIC code is 3571.

Unlock the full company data

50 free credits on sign-up, no credit card required.

Contact sales
Live signals
BBCThe Aftermath of Huang Renxun's Resignation: The 'Spy' Meme in China's Public Opinion, Election Speculations, and Stock Price 'Black Swan'After Huang Renxun's resignation, domestic chip stocks surged, with Cambrian Technology rising 8.11% and Moore Threads and MetaX gaining over 8%. The Shanghai Composite and CSI 300 remained flat before lunch.36krSharp Surge: Domestic Chips Are Undergoing Collective Explosive GrowthFour leading domestic GPU firms rallied on November 12 after disclosing Q3 results, with Moore Threads and Muxi posting profits. Moore Threads revenue rose 319% year-on-year to 3.289 billion yuan, while Enflame Technology remains loss-making. NVIDIA's board turmoil and rising computing power prices also drove the surge.YahooChina chip stocks rally as U.S. weighs curbs on remote AI accessChinese chip stocks rallied on Thursday, with Cambricon up over 11% and Moore Threads up about 7%, as investors weighed potential U.S. curbs on remote AI chip access. The Commerce Department is preparing a rule to block Chinese AI firms from using overseas data centers for advanced chips, which could extend to seven countries. The proposal remains under interagency review.PandailyMoore Threads Sets MDC 2026 for December 29-30, With S6000 Card and Lushan GPUs ExpectedMoore Threads announced its MDC 2026 conference in Beijing on December 29-30, promising over 10 new products. The S6000 AI card and Lushan graphics GPUs on the Huagang architecture are expected by year-end, with Lushan claiming 16x geometry, 64x AI, and 50x ray-tracing performance. Pricing and shipping dates remain unannounced.Simply Wall St3 China AI Infrastructure Stocks Investors Are Watching As Hardware Exports SurgeChina's AI hardware exports have generated trade surpluses above $1 trillion, with shipments of chips, servers, and networking gear surging. The article highlights Moore Threads Technology, Eoptolink Technology, and Accelink Technologies, noting Moore Threads' revenue of CN¥3.29 billion and Eoptolink's P/E of 17.6x versus an industry average of 27x.PandailyMoore Threads Updates AICUBE With Smart-Home Control and Local AI NAS ManagementMoore Threads updated its MTT AICUBE home AI hub with smart-home control, local file management, and a mobile client. The update adds Matter and ZigBee support, voice commands for appliances, and Olares file operations. A companion app is available on iOS and Android.PandailyMoore Threads Adds Day-Zero Kimi K3.1 Pro Support on MTT S5000 GPUsMoore Threads completed day-zero adaptation of Moonshot AI's Kimi K3.1 Pro model on its MTT S5000 AI cards, enabling model loading, service startup, and basic text and image inference. The company used SGLang-MUSA and MUSA components, but did not disclose performance metrics or production configuration details.MacaubusinessBeijing seeks to bolster AI shares | Macau BusinessBeijing is boosting AI shares after recent drops, reactivating liquidity tools and seeing ETF inflows reach 120 billion yuan ($18 billion) in nine funds. Tech stocks like Moore Threads and MetaX have fallen around 50% over six months, while the government aims to stabilize the market.South China Morning PostChina’s AI chipmakers report first profits, but will the trend continue?China's pure-play AI chipmakers turned profitable in the third quarter, with domestic GPU developers like Moore Threads, MetaX, and Iluvatar CoreX filling the void left by Nvidia. However, soaring memory costs and supply-chain difficulties are expected to limit long-term growth, prompting investment banks to cut margin forecasts.MarketScreenerAsian Chip Stocks Diverge Amid Concerns of Tighter Data-Center RegulationsAsian chip stocks were mixed Thursday as investors worried about tighter data-center regulations under Democratic control of Congress. Korean memory-chip giants Samsung and SK Hynix fell 4.8% and 7.0%, while Chinese chip stocks rallied, with Moore Threads gaining 5.65% and Shanghai Biren 11%. Analysts see limited risks to the AI investment boom.