An 8-Year Strategic Play That Added $100 Billion to Alibaba’s Market Value

Alibaba’s share price has been closely tracking anything tied to its chip story.

By the close of Hong Kong trading on January 23, Alibaba’s stock was up about 3.25% versus January 21, adding roughly HK$101.2 billion in market capitalization in just two days. On January 22, Huxiu reported that Alibaba’s chip unit, T-Head (Pingtouge), is expected to pursue an independent listing.

T-Head traces back to the chip project team within Alibaba’s DAMO Academy. In 2018, the team was spun out and established as a wholly owned Alibaba subsidiary.

What’s especially notable is how T-Head fits into Alibaba’s broader AI blueprint, which now spans four major pillars—where the chip business functions as a “tactical foundation”:

Alibaba Cloud, launched in 2009, serves as the “mothership” for Alibaba’s AI efforts and the starting point of its AI journey.

T-Head, established in 2018, supports Alibaba Cloud by supplying chips that help lower costs, reduce reliance on external chip vendors, and improve “coupling” through in-house silicon.

The Qwen foundation model, first released in 2022, leverages the computing base provided by Alibaba Cloud and T-Head, and acts as the model backbone for Alibaba’s AI products across both enterprise and consumer scenarios.

Qwen’s consumer-facing app and AI hardware, officially launched in November 2025 (building on earlier products such as Quark, founded in 2016, and the Tongyi app opened to consumers in 2023), represent Alibaba’s core AI-to-consumer push—alongside devices like Quark AI glasses and other AI hardware initiatives.

Sources told Huxiu that internally, T-Head has long been treated with “saturation-level investment.” Since 2018, Alibaba has consistently invested heavily in chips—and has done so steadily rather than episodically.

Importantly, Alibaba did not require T-Head to finance R&D through external fundraising. Instead, the mandate was clear: focus on product refinement and research, with the parent company absorbing the cost.

Over the past few years, Huxiu learned that Alibaba’s internal thesis for building T-Head has centered on one core differentiator: “full-stack capability.”

That means expanding across the full spectrum—CPU, GPU, storage chips, edge chips, and even general-purpose silicon—while also building end-to-end in-house capabilities spanning architecture, IP cores, chip products, software stacks, and ecosystem development.

So far, across an eight-year runway, T-Head’s product portfolio includes the Hanguang 800 AI inference chip, the Yitian 710 server CPU, a general-purpose GPU chip (PPU), the Zhenyue 510 SSD controller, and the Yuzhen IoT edge chip line.

This diversified approach differs from many domestic chipmakers that concentrate on a single track. Recent IPO newcomers like Moore Threads and Biren Technology, for example, largely focus on GPU development, becoming key forces in China’s push for homegrown GPUs—but their playbook is fundamentally different from T-Head’s multi-front chip matrix.

Veteran industry observers often point to Huawei (HiSilicon + Ascend) as the most representative company in China with truly full-stack chip capability. Huawei’s biggest edge, however, is that it has successfully penetrated nearly every major scenario—from smartphones and servers to cars.

Among China’s internet giants, T-Head is widely viewed as the closest in strategic posture to Huawei’s model. The difference is emphasis: Huawei is scenario-dominant across a wide device landscape, while Alibaba’s chips place greater weight on ecosystem-driven, cloud-centered full-stack integration.

The strategic differences among major domestic internet players’ chip approaches can be summarized as follows:

Tencent: Its self-developed chips to date are specialized chips, focused on AI inference, video transcoding, and smart NICs—tightly aligned with its own business and AI roadmap.

ByteDance: Similar to Tencent in direction, concentrating on chips for AI inference, video encoding/decoding, and head-mounted displays—though market chatter about ByteDance exploring more general-purpose chips persists.

Baidu Kunlun: Focused on AI-specific chips, without fully committing to broad general-purpose silicon in the traditional sense.

Alibaba T-Head: A full-stack chip matrix, with fully self-owned IP cores and chip design, and an open ecosystem orientation.

With Alibaba now framing AI as one of its two core group strategies (the other being “big consumption”), and with CEO Eddie Wu explicitly calling out a three-year plan to invest RMB 380 billion into AI hardware infrastructure—alongside Alibaba Cloud’s long-term “ASI” ambition (Artificial Super Intelligence) as an ultimate technical goal—T-Head’s strategic value to Alibaba is hard to overstate.

Some analysts see T-Head as a true “first move” for Alibaba: chips, cloud, and models together form the fertile “black soil” that supports Alibaba’s future AI imagination space.

Another important internal theme is Alibaba’s long-discussed “one cloud, multiple chips” strategy—designed to ensure supply stability by avoiding reliance on a single vendor. In that context, T-Head directly reduces Alibaba Cloud’s dependence on external chip suppliers.

From a cost perspective, sources also told Huxiu that because “compute scarcity” remains a real constraint for leading model players, in-house chips can be an effective cost-control lever when capacity is tight.

At the leadership level, the most important benchmark for T-Head is whether it can reliably support Alibaba Cloud’s service delivery. Internally, the relationship between cloud and chips isn’t treated as a simple “supplier” arrangement. Teams closely monitor customer-side testing data when clients run workloads on T-Head chips, evaluating real-world compatibility and coupling.

If T-Head ultimately succeeds in spinning off and listing independently, it could bring three layers of impact:

First, it may ease Alibaba’s funding pressure to a degree by opening a new “supply line” through public market liquidity.

Second, it could free up internal resources that can be redirected toward foundation models and consumer AI products—the “black soil” where new AI species can grow—helping Alibaba concentrate more firepower in the consumer AI arena.

Third, as an independent company, T-Head may be better positioned to expand its external customer base and broaden its market influence.

Even so, the challenges remain persistent.

Since 2024, China’s GPU market has been red-hot, and talent competition has intensified sharply—meaning T-Head will likely face an even more aggressive war for top engineering talent.

On the competitive front, Alibaba is also up against strong players such as Huawei, Xiaomi, and Baidu. In some cases, rivals are effectively engaging in an all-out “cloud + chips” confrontation.

And we can’t forget the other battlefields. Alibaba’s consumer AI fight with ByteDance and its on-demand retail and delivery war with Meituan are still ongoing.

In that context, a T-Head spin-off is being viewed as a strategic “breakthrough move” amid intersecting battles—a way to loosen a relatively stagnant situation, relieve pressure in a measured way, and concentrate resources on the most critical points of attack.

Because chips are the definition of a slow-burn engineering project. This is a long campaign—and it will be won through endurance.