On the evening of Nov. 10, Alibaba issued an announcement on the Hong Kong Stock Exchange proposing to separately list its wholly owned subsidiary, T-Head (Shanghai) Technology Co., Ltd., on the exchange’s Main Board.

The announcement was brief. It said Alibaba had submitted the spin-off proposal under Practice Note 15 of the Listing Rules, and that the Hong Kong Stock Exchange had confirmed the proposal could proceed. T-Head had already confidentially filed its A1 listing application on Sept. 11, 2026, two months earlier, planning to list on the Main Board through a global offering of H shares, including a Hong Kong public offering and a placement to institutional and professional investors.

The wording was notably restrained. The entire announcement described T-Head in just one sentence: It is Alibaba’s wholly owned subsidiary and “a leading full-stack data center chip provider in China, with chips capable of enabling system-level computing power solutions.”

There was no market-share figure, customer count, shipment volume, revenue figure or description of AI chip performance.

Yet within a day of the announcement, the market had placed the company’s value somewhere between $25 billion and $62 billion.

A1 Filing Came Two Months Earlier

The timeline contains some clues.

T-Head was established in 2018 through the integration of C-SKY Microsystems, acquired by Alibaba, and the chip team of the DAMO Academy. By the time of this announcement, it had been eight years.

The A1 application was filed on Sept. 11, exactly two months before the announcement. A confidential filing means the company does not publicly disclose the application, allowing it to gauge the market and adjust its plans during the listing window. Going public with the announcement indicates that Alibaba had decided to put the plan on the record.

An earlier signal came in March 2026, when Alibaba CEO Eddie Wu was asked on an earnings call whether T-Head might pursue a standalone listing. He said the possibility could not be ruled out, but offered no timetable. Foreign media had also previously cited people familiar with the matter as saying that T-Head planned to pursue a Hong Kong IPO in 2027. That timing aligns with the pace now emerging from the disclosures.

Looked at more broadly, the move fits into Alibaba’s wider restructuring. After launching an organizational overhaul in 2023, Alibaba’s holding group gradually began shifting toward an investment-holding model, pushing its business units to operate and raise capital independently. Chips are the most technically intensive and capital-hungry part of that strategy.

The announcement gave three reasons: to help investors assess the value of Alibaba’s remaining businesses more clearly; to allow the market to evaluate T-Head’s operating performance and growth potential independently; and to strengthen T-Head’s market profile and its ability to access equity and debt capital markets on its own.

Put plainly: When chips and cloud services are sold as one package, no one can say exactly what the chip business is worth. A separate listing would give it a price.

Eight Years to Complete the Three Data Center Pillars

The announcement was sparing with details, but its description of T-Head as a “full-stack data center chip” company carries weight.

T-Head’s product portfolio now spans five categories: the Zhenwu series of AI chips, the Yitian series of Arm server CPUs, the Zhenyue series of storage controllers, the Panmai series of smart network cards, and ICN interconnect chips. Computing, storage and networking, the three core pillars of a data center, are each covered.

That combination is rare in China. China Fund News has said that only Huawei and T-Head currently have full-stack in-house capabilities spanning “computing, storage and networking.” That is a media assessment rather than a statement from the announcement, but it broadly matches T-Head’s product lineup.

The more significant capability is at the systems level. T-Head and Alibaba Cloud jointly launched the Zhenwu supernode, built on their self-developed computing, storage and networking chips, for training and inference of large models with trillions or even tens of trillions of parameters. This is where the phrase “system-level computing power solutions” comes into play: When customers buy individual chips, they still have to handle interconnects, storage and scheduling themselves. A supernode sells the capability of an entire rack.

From 2018 to 2026, T-Head took eight years to build out a footprint extending from individual chips to complete server racks. That is a rapid pace, underpinned by Alibaba’s strategy of sustained heavy investment. Whether the approach works will now be tested in the public markets.

The Market Supplied the Numbers

The figures missing from the announcement have been filled in by the media and the market, but it is important to distinguish between company disclosures and external estimates.

On the product side, the Zhenwu V900 was unveiled at the Yunqi Conference. Public statements say it delivers three times the performance of its predecessor, can scale to 500,000 chips in a single cluster, and is scheduled for mass production in the first quarter of 2027. The Xuantie RISC-V processor IP family has shipped more than 6 billion units cumulatively, according to public statements from the product team.

On the business side, as of April 2026, the Zhenwu series of AI chips had shipped 560,000 units cumulatively and served more than 650 corporate customers across more than 20 industries, including intelligent driving, finance, large models, embodied intelligence, energy and manufacturing. Other reports have said its annualized revenue has reached the tens-of-billions-of-yuan range.

On valuation, the market has produced a range of $25 billion to $62 billion.

But the announcement included one important caveat: The size and structure of the global offering, as well as the extent of Alibaba’s reduction in its stake in T-Head, have not yet been finalized. In other words, whether the valuation is $62 billion or $25 billion, both figures are market estimates derived from comparable companies, not numbers provided by T-Head.

The most revealing figure may be the range itself. The high end is 2.5 times the low end. That does not mean the market has simply miscalculated; it means investors are applying two different sets of assumptions to the same company.

The bullish assumptions are that China’s AI data center buildout is creating a hard demand for domestically developed computing power; that T-Head has completed large-scale engineering validation through Alibaba Cloud; and that its full-stack integration of computing, storage and networking gives it an advantage that warrants a price-to-sales multiple comparable to international peers.

The bearish assumptions are that high-end chip development requires long cycles and heavy investment, competition is intense, external supply chains face constraints, and, most importantly, how much of T-Head’s revenue comes from Alibaba Cloud itself.

The Dispute Is Over the Share of Internal Orders

This will be one of the metrics investors watch most closely after the listing.

If 80% of a chip company’s revenue comes from its parent, it looks more like a cost-center spin-off than an independent chip company. The valuation the market is willing to assign will depend on how much of the business external customers can support.

Public information shows that the Zhenwu series serves more than 650 external customers across more than 20 industries, suggesting that external customers are already using the chips in real-world applications. But the announcement did not disclose the proportion of total revenue generated by external customers, and there will be no definitive answer before the prospectus is published.

Timing is another variable. The Zhenwu V900 is scheduled to enter mass production in the first quarter of 2027. That means there will still be a gap between T-Head’s product narrative at the time of its listing and the point when the chip actually reaches volume production. Over the past two years, capital markets have clearly discounted statements built around “mass production next year.”

Why It Matters

Viewed from a broader perspective, the move marks the end of one phase.

In the first phase of Chinese cloud providers’ in-house chip efforts, the motivation was straightforward: cut costs, secure autonomous access to computing power and avoid being dependent on a single supplier. Chips were treated as part of the cloud business, with their economics buried in the cloud cost structure and largely invisible to outsiders.

If T-Head successfully completes the spin-off, its chip business will move from the cost statement to the balance sheet, with an independent price, financing channels and performance benchmarks. That will create pressure for the company and a reference point for the industry: China’s AI chip companies will have another publicly traded benchmark against which to measure themselves.

Alibaba also left itself room to maneuver in the announcement. The spin-off will take effect only after approval by the Hong Kong Stock Exchange Listing Committee, completion of the China Securities Regulatory Commission filing, and final decisions by Alibaba and T-Head. In other words, the process is not finished yet.

After eight years of development, T-Head is finally approaching an independent valuation. And valuation is never just about calculating what a company is worth. It also forces T-Head to answer a more fundamental question: Once Alibaba Cloud’s orders are taken out of the equation, how many customers will still be willing to buy?