Recommended Free Tools
The chip in the headline is Alibaba’s Hanguang 800, unveiled on September 25, 2019. Alibaba described it as a neural processing unit (NPU) designed to speed up machine-learning inference: running a trained model to produce an answer, such as a product recommendation or translation. The announcement was about Alibaba’s own commerce and cloud infrastructure—not a consumer chip launch.
What is Alibaba’s Hanguang 800 chip?
Alibaba said its chip-design subsidiary T-Head developed Hanguang 800 under the Alibaba DAMO Academy. The company characterized it as an NPU specialized for accelerating machine-learning tasks. Alibaba’s 2019 announcement said the chip was already being used internally in e-commerce operations.
Inference is the stage where a trained machine-learning model is put to work on new inputs. For example, a model can infer which products are relevant to a search, translate text, or generate a customer-service response. That differs from training, which is the process of building or updating a model from data.
What did Alibaba use it for?
Alibaba named product search, automatic translation, personalized recommendations, advertising, and intelligent customer service as internal uses. The announcement framed Hanguang 800 as infrastructure for the company’s operations and cloud business, rather than hardware intended for individual buyers.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors#1 Best Overall
- High-Performance AI Processing: The MX3 is designed to handle the most demanding AI computer vision workloads, delivering exceptional performance and efficiency.
- Flexible Integration: The MX3 can be easily integrated into your existing systems via its M.2 M-key form factor and support for Linux operating systems.
- Energy Efficient: The MX3 is designed to provide high performance while minimizing power consumption.
- Comprehensive Software Development Kit (SDK): The MX3 is supported by a comprehensive SDK that simplifies development and deployment.
- Hardware compatability: The MX3 is compatible with the PCI-SIG M.2 M-key 2280 Specification. It can be used with the Raspberry Pi 5 with a M-key 2280 HAT.
How fast did Alibaba say Hanguang 800 was?
Alibaba’s launch release reported a peak single-chip performance of 78,563 IPS and computation efficiency of 500 IPS/W on a ResNet-50 inference test. These are vendor-reported figures; the release does not establish independent laboratory verification or provide enough test detail to apply the results to other models or workloads. Alibaba’s release also used a Taobao image-categorization workload as an example:
- Alibaba said around one billion product images were uploaded to Taobao per day.
- It said Hanguang 800 reduced the time to categorize that daily volume and prepare it for search and recommendation use from one hour to five minutes.
Those time and throughput claims describe Alibaba’s stated workload and launch-era results. They should not be read as a general guarantee for customer systems.
Rank #2
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
Could customers buy or use the chip?
Alibaba’s 2019 release said the company planned to make access to the chip’s advanced computing available to clients through Alibaba Cloud. Jeff Zhang, Alibaba Group CTO and president of Alibaba Cloud Intelligence at the time, said the company planned to provide that access through its cloud business. That statement records a plan at launch; it does not confirm current Hanguang-specific availability. The announcement does not establish a retail purchase route or consumer sales.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How Hanguang 800 fits with Alibaba’s later AI chips
Hanguang 800 is a 2019 product, not Alibaba’s latest announced AI processor. Alibaba introduced later Zhenwu chips in 2026. The available announcements describe different generations and workloads, so their headline performance claims are not a direct benchmark comparison.
| Chip | Announcement and intended workloads | Company-reported specifications or claims | Availability context |
|---|---|---|---|
| Hanguang 800 | Announced September 25, 2019; described as an NPU for inference. | 78,563 IPS peak single-chip performance and 500 IPS/W on a ResNet-50 inference test, as reported by Alibaba in 2019. | Alibaba said it was in internal use and planned cloud access for clients; the release does not establish present Hanguang-specific access. |
| Zhenwu M890 | Announced May 20, 2026; described by Alibaba as supporting both training and inference. | Alibaba reported 144 GB on-chip memory, 800 GB/s inter-chip bandwidth, support from FP32 down to FP4, and three times the performance of its predecessor, Zhenwu 810E. | Alibaba also described its Panjiu AL128 system and cloud model services. The cited announcement does not establish a Hanguang comparison under common benchmark conditions. |
| Zhenwu V900 | Announced September 22, 2026; described as a training and inference processor. | Alibaba reported 216 GB of GPU memory, 1,200 GB/s inter-chip bandwidth, support for FP8 and FP4, and three times the performance of M890. | Alibaba scheduled mass production and commercial release for Q1 2027; this was a forward-looking schedule in the September 2026 announcement. |
The M890 and V900 specifications and performance claims are Alibaba’s statements, not independently established figures in the cited announcements. Their memory, bandwidth, precision support, and stated workloads are useful context for how Alibaba’s product announcements evolved, but the reported multiples do not show how the chips compare on a shared independent test.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
