TodayToday's edition
Alibaba told the Hong Kong exchange on Nov. 10 that AI chip unit T-Head has applied to list. IFR, cited by Reuters, puts the raise at US$2 billion to US$3 billion; Alibaba has not set the size.
Chips & Power DeskEdited by Ege Erten
3 min readAI-assisted
Alibaba confirmed on Nov 10 that T-Head applied to list in Hong Kong; the raise range is IFR-reported and the V900 figure is a vendor claim.
Illustration (AI-generated)
Alibaba Group said on Nov. 10 that its AI chip unit T-Head submitted a confidential listing application (Form A1) to the Hong Kong Stock Exchange on Sept. 11, 2026, according to Alibaba's announcement filed with the SEC as a Form 6-K.
The announcement says the exchange had first confirmed that Alibaba may proceed with the spin-off. The planned Global Offering of T-Head H shares would combine a Hong Kong public offering with a placing to institutional and professional investors. Alibaba describes T-Head, a wholly owned subsidiary, as a provider of full-stack data center chips for system-level compute in China, and expects it to remain a subsidiary after the listing.
For readers who track China's AI-chip supply, the filing shows T-Head preparing to raise outside capital on its own. Alibaba gave no revenue, loss, shipment or customer figures for T-Head in the announcement, and the application itself is confidential. Alibaba also told the exchange that the spin-off is insignificant to the group when measured against its total assets, revenue and net income, which is Alibaba's own statement of T-Head's scale relative to the parent.
Alibaba said the size and structure of the Global Offering, and how far its shareholding in T-Head will fall, are not finalized. The International Financing Review (IFR) reported earlier that the share sale could raise US$2 billion to US$3 billion, according to Reuters, as carried by The Standard of Hong Kong. That is IFR's report, not an Alibaba figure.
The listing still needs approval from the exchange's Listing Committee, completion of a filing with the China Securities Regulatory Commission, and final decisions by Alibaba and T-Head. Alibaba said there is no assurance the spin-off will happen or when. It also obtained a waiver from the exchange rule that normally gives existing shareholders an assured entitlement to shares in a spun-off company, citing the burden of a US public offering and PRC restrictions.
T-Head's current data-center accelerator is sold as the Zhenwu 810E. The company's product page lists 96 GB of memory per module at 2.7 TB/s and 700 GB/s of chip-to-chip bandwidth per chip, and T-Head said in April that more than 100,000 Zhenwu chips were in use for autonomous-driving work on Alibaba Cloud. These are vendor claims.
According to Reuters, T-Head unveiled the next-generation Zhenwu V900 in September. Alibaba claims it is the most powerful AI chip in China, with three times the performance of the M890 and clusters of up to 500,000 chips for frontier-model training. Alibaba expects the V900 to enter mass production in the first quarter of 2027. No power draw per chip or per cluster has been published, so the electrical load of a 500,000-chip cluster cannot be sized yet.
The filing arrives as Chinese buyers look for domestic training silicon. Voltgrid reported on Nov. 3 that DeepSeek reportedly plans to train V5 on Huawei Ascend 950 chips, a plan neither company has confirmed. T-Head's parent is also a large memory buyer: as Voltgrid reported on Oct. 28, Alibaba, Tencent and ByteDance reportedly seek 10-year DRAM supply deals.
T-Head's financials, shipment volumes and customer list are not public, because the application was filed confidentially. The offering size, the structure and Alibaba's post-listing stake are open, and the US$2 billion to US$3 billion range comes only from IFR via Reuters. The Zhenwu V900 performance and cluster-size figures are Alibaba's claims, and the Q1 2027 mass-production date is an expectation. The next step to watch is the exchange's Listing Committee hearing and the CSRC filing; Alibaba said it will make further announcements as appropriate.
