Zhenwu 810E by Alibaba T-Head: status and timeline

Milestones

Next · Expansion to external cloud providers

Hanguang 800, T-Head's first AI chip, launchedSep 2019Complete.
Zhenwu 810E announced and launched29 Jan 2026Complete.
Highest shipment volume among domestic GPUs in 20252025Complete.
Zhenwu V900 successor unveiledSep 2026Complete.
Expansion to external cloud providersTarget 2027Not yet reached.

Most important updates

Alibaba T-Head unveils Zhenwu V900 with three times the performance of M89019d ago
Alibaba T-Head unveils Zhenwu 810E, a domestic AI chip supporting training and inference29 Jan 2026

Upcoming

2027Expansion to external cloud providers (next)
2027T-Head IPO

Current obstacles

Export control exposureTSMC manufacturing in Taiwan makes Zhenwu vulnerable to US trade policy changes affecting foundry capacity or advanced nodes.
Competing with Nvidia ecosystemEnterprises and AI researchers invest heavily in Nvidia's CUDA ecosystem. Adopting Zhenwu requires rewriting code or using compatibility layers.

Physics limits

Node availabilityUS export controls limit T-Head to older process nodes than Nvidia's latest chips, capping performance per watt.
Memory bandwidth physicsMemory bandwidth between chip and processors is limited by pin count and module bandwidth. No design trick gets around this fundamental constraint.

How it works

3 parts

Unified design

Training and inference on one chip

Zhenwu 810E efficiently handles both training (high bandwidth) and inference (low latency) in a single flexible compute architecture.

Integration

Alibaba Cloud software stack

T-Head provides software frameworks and drivers optimized for Zhenwu and Alibaba Cloud's deployment model.

Scale

Proven in production

Zhenwu runs in clusters of tens of thousands at Alibaba Cloud, proven at data center scale.

Spec sheet

Spec	Zhenwu 810E
Performance positioning	Between Nvidia A800 and H20^R (reported)
Supported workloads	LLM training, multimodal, autonomous driving^R (reported)
Deployments at launch	Tens of thousands in Alibaba Cloud^R (reported)

^R reported by the company

Update log

2 updates

Tue 22 Sep

Major: Alibaba T-Head unveils Zhenwu V900 with three times the performance of M890PressHardware

Thu 29 Jan

Major: Alibaba T-Head unveils Zhenwu 810E, a domestic AI chip supporting training and inferencePressHardware

About Alibaba T-Head

The team behind Zhenwu 810E

Alibaba T-Head

Zhenwu chips, AI processors, domestic GPU, spin-off IPO

T-Head is Alibaba's semiconductor division, established in 2018 to design AI and general-purpose processors. Its Zhenwu 810E (Jan 2026) achieved the highest shipping volume among domestic Chinese GPUs in 2025. Alibaba is spinning off T-Head as an independent company pursuing separate IPO.

Zhenwu 810E (January 2026) integrates training and inference, positioned between Nvidia A800 and H20, running at scale in Alibaba Cloud.
Zhenwu family achieved highest shipment volume among domestic Chinese GPUs in 2025; V900 unveiled September 2026.
Alibaba is spinning off T-Head as independent entity with separate IPO plans to accelerate growth beyond Alibaba Cloud.

t-head.cn (opens t-head.cn)Company profile →

Founded: 20188 yrs
Headquarters: China
Status: Private
Valuation: private
Works in: ComputingAI accelerators & CPUs
Coverage: 1 program · 3 updateslatest 19d agochecked 11 Oct
Customers: Alibaba Cloud (opens alibabacloud.com)
Partners: Alibaba Group (opens alibaba.com)
Suppliers: TSMC (opens tsmc.com)
People: Alibaba Cloud executives (opens alibabacloud.com)Management

Spec

Zhenwu 810E

Performance positioning

Between Nvidia A800 and H20^R (reported)

Supported workloads

LLM training, multimodal, autonomous driving^R (reported)

Deployments at launch

Tens of thousands in Alibaba Cloud^R (reported)