According to IFR, a publication under Reuters, Alibaba Group's wholly-owned semiconductor chip business entity T-Head is planning to list on the Hong Kong Stock Exchange next year, aiming to raise USD 2-3 billion.
Earlier, in an October research report, JPMorgan provided a cautious judgment framework for T-Head's independent listing. Based on its calculations, referencing peer EV/Revenue multiples and aggressive 2026 revenue assumptions, T-Head's valuation could range between USD 25-62 billion if listed separately.
Even earlier, in January 2026, when news of the spin-off first emerged, Alibaba's plan was to restructure T-Head into an independent entity partially owned by employees before exploring the possibility of an IPO, with "the specific timeline yet to be determined."
T-Head announced its establishment in2018 and has offices in Zhangjiang, Shanghai. The company offers a full-stack product portfolio covering end-to-cloud integration, encompassing data center chips, IoT chips, and more, achieving full coverage of the chip design chain from end to end.
As an AI chip developer within the Alibaba ecosystem, T-Head provides a "backup solution" for Alibaba, reducing dependence on external suppliers. To date, T-Head has launched products including the Yitian processor chip, Zhenyue SSD controller chip, Hanguang artificial intelligence chip, and Yuzhen RFID chip.
The Zhenyue 510 is a high-performance enterprise-grade SSD controller chip. Its IO processing capability reaches 3400K IOPS, data bandwidth reaches 14 GByte/s, and energy efficiency ratio reaches 420K IOPS/Watt.
The Zhenyue 510 adopts T-Head's proprietary chip and firmware architecture. Through excellent hardware-software co-design, it achieves performance breakthroughs while reaching optimal energy efficiency. The chip utilizes T-Head's proprietary Low-Density Parity-Check (LDPC) error correction algorithm, with coding efficiency approaching the Shannon limit and significantly improved error correction performance, resulting in a bit error rate as low as 10^-18. Additionally, the Zhenyue 510 employs integrated hardware-software media application algorithms capable of accurately predicting media level drift, greatly improving long-tail latency and providing applications with highly consistent performance experiences.
The Hanguang 800 is the first data center chip released by T-Head. It is a high-performance AI inference chip based on a 12nm process, integrating 17 billion transistors, with peak computational power reaching 820 TOPS. In industry-standard ResNet-50 tests, inference performance reached 78,563 IPS, with an energy efficiency ratio of 500 IPS/W.
The Hanguang 800 adopts T-Head's proprietary architecture, achieving performance breakthroughs through hardware-software co-design. T-Head's independently developed AI chip software development kit allows the Hanguang 800 chip to deliver high throughput and low latency high-performance experiences when developing deep learning applications. The Hanguang 800 has been successfully applied in scenarios such as data centers and edge servers.
The Yuzhen 600 is a low-power, high-performance UHF RFID electronic tag chip compliant with the EPCglobal Class-1 Generation-2 UHF RFID protocol. This chip features high read sensitivity, omnidirectional reading, and strong environmental adaptability, making it particularly suitable for application scenarios such as smart logistics, smart warehousing, smart retail, and asset management.
The T-Head Yuzhen 611 is a low-power, high-performance, general-purpose UHF RFID electronic tag chip that meets EPCglobal G2V2 and ISO/IEC 18000-6C protocols. Ultra-low power circuit design combined with automatic impedance tuning makes this chip particularly applicable to complex scenarios such as footwear and apparel, fast-moving consumer goods retail, smart logistics, supply chain management, and dynamic asset management.
In May 2026, T-Head released the training-inference integrated AI chip Zhenwu M890, equipped with 144GB HBM. Officially claimed to be three times the performance of the previous generation, it also released the accompanying ICN Switch 1.0 interconnect chip, which can enable a 128-card super-node architecture, already applied in Alibaba Cloud Panjiu AL128 node servers.
In July 2026, T-Head open-sourced SAIL, a full-stack artificial intelligence software platform for Zhenwu AI chips.
On September 22, 2026, T-Head released the new generation training-inference integrated AI chip Zhenwu V900, with performance tripled compared to the Zhenwu M890, capable of meeting the training and inference needs of trillion-parameter large models.
The Yitian 710 is the first Arm server chip released by T-Head, independently designed and developed by T-Head. It features an advanced architecture, high energy efficiency, and high bandwidth, compatible with the Armv9 architecture.
Notably, accompanied by the explosion of agent applications, Agentic AI drives the differentiation of CPU demand. T-Head will launch two server CPUs, the Yitian 720 and Yitian 730, in 2027.
Among them, the Yitian 720 will achieve comprehensive improvements in single-core performance, core density, and energy efficiency; the Yitian 730 will be the first to be designed based on T-Head's fully self-developed CPU microarchitecture, with single-core SPECint2017/GHz performance increasing up to 1.4 times that of the Yitian 710.
What truly gives T-Head the confidence for an independent listing may be the acceleration of its commercialization progress. Alibaba CEO Eddie Wu disclosed for the first time during the Q3 FY2026 earnings call that T-Head "has successfully commercialized," with chip scale exceeding 470,000 units and annualized revenue surpassing the CNY 10 billion mark.
From an industry perspective, T-Head's pursuit of an independent listing coincides with the dual window of sustained heating up of AI computing investment and accelerated domestic substitution.
If T-Head successfully lists on the HKEX, it will become another heavyweight target in the HK stock market's semiconductor sector following SMIC and Hua Hong, also providing a reference for the capitalization path of subsequent domestic chip companies.
Editor | Liu Chengxing Source | Comprehensive compilation and internet