Nvidia H800 80g
Overview
Product sourcing insights & recommendations
The NVIDIA H800 80GB is a specialized AI accelerator built on the Hopper architecture, specifically engineered to comply with U.S. export controls for restricted markets such as China PhilisunGetdeploying. It serves as a modified version of the flagship H100, maintaining high AI compute performance while intentionally throttling communication speeds Philisun.
1. Key Product Specifications
The H800 retains the core AI processing power of the H100 but features significant limitations in interconnect and high-precision scientific computing GetdeployingTechpowerup:
* AI Performance: Delivers up to 1,513 TFLOPS (FP16/BF16) and 3,026 TFLOPS (FP8) with sparsity, making it a powerful engine for Large Language Model (LLM) training and inference Philisun.
* Memory: Features 80GB of HBM2e (PCIe) or HBM3 (SXM) memory with a bandwidth of 2.0 TB/s GetdeployingTechpowerup.
* NVLink Bandwidth: Capped at 400 GB/s bidirectional, which is roughly 44% of the H100's speed. This restricts its efficiency in massive "super-clusters" of thousands of GPUs PhilisunGetdeploying.
* Scientific Computing: High-precision (FP64) performance is gutted to ~1 TFLOPS (vs. 34 TFLOPS on H100) to prevent its use in military-grade scientific modeling Philisun.
2. Market Trends & Pricing (2026)
The market for H800 is driven by geopolitical restrictions, making it a highly sought-after "legal" high-end option in restricted regions Getdeploying.
* Retail Pricing: Due to limited supply and high demand in secondary markets, individual units can command between $25,000 and $40,000 IntuitionlabsHashrateindex.
* Cloud Trends: While H100s are widely available on Western clouds for ~$3.34/hr, H800s are more common in specialized regional clouds, with pricing often slightly lower due to the reduced interconnect performance GetdeployingGetdeploying.
* Stability: Used H100 prices have stabilized around $18,000–$22,000, while the H800 remains volatile as newer models like the B200 and H200 shift enterprise focus elsewhere GpunexIntuitionlabs.
3. Sourcing Insights
B2B sourcing for the H800 80GB remains active, with suppliers primarily located in tech hubs like Shenzhen and Guangzhou.
* Available Stock: Listings range from brand new SXM5 modules to passive PCIe cards for data centers.
* Pricing Variations: Market prices on B2B platforms range from $15,000 to over $36,000 depending on the specific SKU (PCIe vs. SXM) and order quantity.
Here is the analysis of the NVIDIA H800 market. You can continue with:
H800 Ai Accelerator 80gb
Products · 2 lists

H800 Tensor Core GPU Center AI Data Analysis Training Deep Learning 80GB Acceleration High Performance Computing
Shenzhen Huatenmax Technology Co., Ltd.🇨🇳CN3 yrs
YF A 800 80GB HBM2 PCIe Data Center GPU Accelerator Card AI Training Inference Server Workstation High Performance Graphics Card
Shenzhen Vbest Trading Company Limited🇨🇳CN1 yr
TESLA A800 80G TESLA A100 80G TESLA H100 80G AI Computing Card 80GB HBM2e, Data Center GPU Accelerator with 2TB/s Bandwidth
Beijing Huli Tong Technology Company Limitedverified🇨🇳CN9 yrs
A100 80GB Data Center AI Accelerator, 2TB/s Memory Bandwidth Dedicated GPU for Deep Learning Training

A100 80GB HBM2e AI Accelerator Card 6912 CUDA Cores 300W PCIe 4.0 for Data Center Deep Learning
Beijing Zai Ming Technology Co., Ltd.🇨🇳CN2 yrs
New Factory Supply NVlDlA 800 HBM2 Tensor Core GPU AI Training Deep Learning Workstation Accelerator Fan 5120-bit PCIe 4.0
CONG TY CO PHAN TU VAN VA DAO TAO KHAI MINH🇻🇳VN1 yr
New AI GPU Accelerator Ampere Tensor Core Graphics Card Tesla 100 40GB HBM3 PCIe Passive Cooling Data Center Server Workstation

N.XP Ara240 40TOPS Edge AI Computing Machine Vision Robotics Object Detection PCIe Gen4 16GB Low Power M.2 AI Accelerator Module
Shenzhen Geniatech Inc., Ltd.verified🇨🇳CN3 yrs
ASUS AI compute accelerator card Google tpu AI TensorFlow embedded deep learning
Shenzhen Kaisheng Century Electronics Co., Ltd.verified🇨🇳CN3 yrs
PCIe Gen3 AI Accelerator 26 TOPS for Embedded Systems Made in China & Guatemala
LCFC (Hefei) Electronics Technology Co., Ltd.🇨🇳CN2 yrs
4GB Axera AX8850 SoC AI acceleration card M5Stack LLM-8850 Kit
Shenzhen Foing Industrial Co., Ltd.🇨🇳CN7 yrs
HFSecurity XCore ARM GPU AI Computing Station for AI Model Migration and Smart Park AI Solutions
Chongqing Huifan Technology Co., Ltd.verified🇨🇳CN18 yrs
ASUS AI Artificial Intelligence Server Acceleration Card Embedded Neural Computing Stick Deep Learning Linux AI Accelerator
Shenzhen TCT Electronics Co., Ltd.🇨🇳CN5 yrs
Raspberry Pi Hailo-8L Compute Module AI HAT + Neural Network Accelerator Made in China
Shenzhen Qinda Electronics Co., Ltd.🇨🇳CN7 yrs
GGDX Chipboard M5Stack AI-8850 LLM Accelerator M.2 Kit Offers an to Raspberry Pi AI HAT+2 Octa-core A55 Processor
Shenzhen Gegu Dingxin Technology Co., Ltd.🇨🇳CN1 yr
ARA240 16GB M.2 AI Computing Card Accelerator up to 40 eTOPS Applicable for Industrial Edge AI
Baoding Forlinx Embedded Technology Co., Ltd.🇨🇳CN10 yrs
Chipboard EMA EA300V Ai Accelerator Card With Ascend Chip, 8-core, 24gb Lpddr4x, Pcie 4.0, 100tops for Gpu Ai Computing
Shenzhen Anshengchuang Technology Co., Ltd.🇨🇳CN2 yrs
New & Original Atlas 300T A2 64G AI Inference Card AI Acceleration Card
Shenzhen Xulianjiesen Technology Co., Ltd.🇨🇳CN9 yrs
Hailo-8 M.2 AI Accelerator Module, 26TOPS Hailo-8 AI Processor Optional For PCIe To M.2 Adapter Board, For Raspberry Pi 5
Shenzhen Lonten Technology Co., Limited🇨🇳CN13 yrs
Chipboard EMA EA300V Ai Accelerator Card With Ascend Chip, 8-core, 24gb Lpddr4x, Pcie 4.0, 100tops for Gpu Ai Computing
Shanghai Xuanxin Technology Co., Ltd.verified🇨🇳CN2 yrs
Asus CRL-G18U-P3DF PCIe Gen3 AI accelerator card
Haoxin Micro-Control (Beijing) Technology Co., Ltd.🇨🇳CN1 yr
Chipboard Ema Ea300v Ai Accelerator Card With Ascend Chip, 8-core, 24gb Lpddr4x, Pcie 4.0, 100tops for Gpu Ai Computing
Shanghai Xuanxin Technology Co., Ltd.🇨🇳CN7 yrs
BL808C AIoT Chipset Video And Audio Encoding AI Hardware Accelerator With Wi-Fi/BT/BLE/Zigbee
Shenzhen YTF Technology Co., Limited🇨🇳CN12 yrs
Brand New Original Tesla L4 24GB PCIe Graphics Accelerator GPU HH HL 75W 900-2G193-0000-000 in Stock
Beijing Jinyingtai System Technology Co., Ltd.🇨🇳CN4 yrs
Techvision RK1828 AI Accelerator Card 20 TOPS M.2 2280 NPU Module Universal AI Co-Processor for SMARC SOMs & All Carrier Boards

Axera AX8850 SoC AI acceleration card M5Stack LLM-8850 Card
Shenzhen Pinkuo Trading Co., Ltd.🇨🇳CN2 yrs
DX-M1 AI Accelerator M.2 Module
Shenzhen Shengdingtong Technology Co., Ltd.verified🇨🇳CN1 yr
DX-M1 AI Accelerator M.2 Module
Shenzhen Dongjiang Chuangfu Technology Co., Ltd.🇨🇳CN1 yr
DX-M1 AI Accelerator M.2 Module
Shenzhen Sunhokey Electronics Co., Ltd.🇨🇳CN13 yrs
Edge AI Inference System Embedded Computer Core I3 10110U Processor Hailo-8 AI Accelerator 16GB RAM 1TB SSD POE RS232 RJ45
Yanling Industrial Computer Technology (shenzhen) Co., Ltd.verified🇨🇳CN11 yrsRelated Searches
Explore more sourcing topics & related categories
Sources & References
23 sources cited · Verified industry data & reports
NVIDIA H800 PCIe 80 GB Specs - GPU Database
TechPowerUp
The GPU is operating at a frequency of 1095 MHz, which can be boosted up to 1755 MHz, memory is running at 1593 MHz (3.2 Gbps effective). Being a dual-slot card ...
NVIDIA H800 GPU buying guide for AI workloads
Alibaba.com
Nvidia H800-80G Specification ; Tensor Cores, 320 ; Memory Size, 80 GB ; Memory Type, HBM2e ; Memory Bandwidth, 2 TB/s ; Peak FP16 Performance, 134.4 TFLOPS.
H100 GPU
NVIDIA
Product Specifications ; FP64 Tensor Core, 67 teraFLOPS, 60 teraFLOPs ; FP32, 67 teraFLOPS, 60 teraFLOPs ; TF32 Tensor Core · 989 teraFLOPS, 835 teraFLOPs ; BFLOAT16 ...
Exploring the Performance and Applications of the GPU ...
ResearchGate
This paper investigates the performance, capabilities, and use cases of the GPU H800 in the context of DeepSeek AI applications.
What Are the Benefits and Features of the NVIDIA H800 GPU?
szwecent.com
The NVIDIA H800 GPU is equipped with 80GB of high-bandwidth HBM2e memory, ensuring seamless handling of large-scale AI tasks, including training ...
Release Notes - NVIDIA Docs
NVIDIA Docs
Support for the following GPUs: NVIDIA H800 PCIe 80GB; NVIDIA L4; NVIDIA L40; NVIDIA RTX 6000 Ada. Changes to Virtualization Software in this Release.
