TBC Home →
Company profiles›Alibaba T-Head

Alibaba T-Head平头哥半导体

BABAFounded 2018HangzhouReviewed 2026-08-30

T-Head is a fabless semiconductor design unit of Alibaba Group that develops processor IP and custom chips, deploying silicon primarily inside Alibaba Cloud rather than through broad merchant chip sales. Its RISC-V processor IP is open-sourced to grow an ecosystem, while its AI accelerators and server CPUs serve Alibaba's own cloud infrastructure and AI services.

Scale

—no public financials

4 tracked chips with sourced pricing/specs

Gap vs frontier

Close behindsector position

Sector-level position; no priceable accelerator silicon to compare chip-by-chip.

The chip stackEach layer depends on the layers below it
1 · ChipsWhat the stack produces
CPU · Close behind
2 · Design toolsSoftware and IP to design them
IP cores · Substitute
3 · ManufacturingPlants that turn designs into silicon
4 · EquipmentMachines inside the plants
5 · MaterialsWhat the plants consume

Company at a glance

Scale

Listed · CPU

China competition

Competes with Loongson, Phytium, Hygon, Huawei Kunpeng in CPU — peers listed at sector level.

Close behind

Global comparables

Global reference set: Intel, AMD, Arm.

Global reference companies: Arm, Imagination Technologies, Google (TPU).

Technical gap

Sector rated close behind vs the frontier.

Obstacles

Sector bottleneck applies

Performance per watt, software ecosystem, server qualification, and dependence on advanced foundry processes.
All sourced facts (3)
  • Founded: 2018
  • Headquarters: Hangzhou, China
  • Listing status: NYSE (parent Alibaba Group) · BABA · Alibaba Group ADS

Tracked chips: T-Head Hanguang 800 (含光800) · T-Head PPU AI inference accelerator · T-Head Zhenwu M890 · T-Head Yitian 710

Why it matters

Alibaba T-Head is China's most strategically embedded open-silicon arm, pairing open-source RISC-V processor IP (XuanTie) with cloud AI accelerators and server CPUs to cut Alibaba Cloud's reliance on foreign-designed chips. The test that matters is whether its in-house NPUs and server processors reach volume deployment inside Alibaba data centers and win external adopters, not just IP and prototype launches. The thesis confirms if deployment scale or external design wins surface; it breaks if foundry access is cut off or internal adoption stays marginal.

Where it sits in the stack

Compute

CPU

Close behind

Performance per watt, software ecosystem, server qualification, and dependence on advanced foundry processes.

AI trainingAI inferenceRoboticsEV / auto
Design

IP cores

Substitute

High-performance CPU IP, interface IP, verification, ecosystem trust, and global licensing constraints.

AI trainingAI inferenceRoboticsEV / auto

Key semiconductor products

  • XuanTie RISC-V processor IP
  • Hanguang AI inference NPU
  • Yitian ARM server CPU

Chip pricing

Ecosystem

Investors / affiliates

  • Alibaba T-Head

Customers / end markets

  • Alibaba Cloud
  • Hyperscale cloud data centers
  • Cloud AI inference services
  • Edge, IoT, and embedded devices

Global peer set

  • Global reference companies: Arm, Imagination Technologies, Google (TPU)

Affiliates

  • Alibaba Group
  • Alibaba Cloud

Competitive position

  • China position: Near frontier.
  • RISC-V IP and cloud AI/server chips are credible but trail frontier merchant accelerators; differentiation comes from captive Alibaba Cloud deployment and open-source ecosystem rather than raw performance leadership.
  • Strategic value is gated by internal deployment scale, external RISC-V ecosystem traction, and access to advanced foundry capacity.
  • Global comparables: Global reference companies: Arm, Imagination Technologies, Google (TPU).
  • US export controls on advanced chips and foundry services: Could limit access to leading-edge nodes for AI accelerators and server CPUs; no T-Head-specific listing evidenced.

Milestones and signals

  • 2018—Alibaba T-Head was founded.
  • Listing—NYSE (parent Alibaba Group) · BABA · Alibaba Group ADS

What to watch

  • New product releases with disclosed specifications.
  • Customer qualification, design-win, or volume-use signals.
  • Revenue, capacity, shipment, market-share, or valuation updates.
  • Export-control, supply-chain, or tooling constraints.

Questions

Last reviewed 2026-07-11

Answered questions

Which customers, fabs, or end markets show real adoption rather than policy interest?

Named exposure so far: Hyperscale cloud data centers, Cloud AI inference services, and Edge, IoT, and embedded devices.

What is the measurable gap versus global reference companies?

Against Arm, Imagination Technologies, Google (TPU), and Arm, Imagination Technologies, Google (TPU): RISC-V IP and cloud AI/server chips are credible but trail frontier merchant accelerators; differentiation comes from captive Alibaba Cloud deployment and open-source ecosystem rather than raw performance leadership.

Open questions

  • Which products are qualified beyond pilots in fabless?