On June 5, Tencent Group Senior Executive Vice President Tang Daosheng sent a clear signal: Tencent is adopting an ecosystem-integrated strategy and will introduce more domestic chips in the second half of the year.

He bluntly stated that self-developed chips cannot solve current production capacity pain points, and Tencent is willing to become a "benchmark for computing power demonstration" for more domestic chip manufacturers.

Big companies are currently severely short on computing power; relying solely on their own manufacturing is simply not enough. So Tencent proactively opened its arms and let domestic chip manufacturers "beat the challenges and upgrade." As long as your chip can run Tencent's model, Tencent is willing to provide you with test scenarios and orders. For domestic AI chip companies, this is a once-in-a-lifetime opportunity to "get on board."

| Big Tech Companies' Computing Power Hungry: Securing Big Tech Orders Is Securing the 'Lifeline'

Currently, major internet companies are frantically ramping up AI infrastructure construction, with computing power demand exploding. Giants like Tencent, Alibaba, and ByteDance have demands for AI accelerator cards that even exceed the supply capacity of most domestic manufacturers.

For domestic AI chip manufacturers, entering the supply chains of major companies is key to explosive performance. Large companies not only have massive business scales and strong capital, but once both sides build technical trust through long-term adjustment, their cooperation stickiness is extremely strong.

The cost for major companies to replace their main suppliers is extremely high, meaning that whoever can secure large-scale orders from big companies first will earn a "get-out-of-jail" card in fierce market competition.

| Crossing the "Life-and-Death Line": The Strict "Three-Step" Test and Dual Competition of Big Companies

However, securing orders from major companies is by no means easy. Unlike non-internet companies that value "autonomy and control," large companies' procurement is highly market-oriented and usually involves hardware system testing, model matching testing, and cluster grayscale testing.

Third-party domestic chip manufacturers face extremely fierce competition. On one hand, they have to face ecological barriers from overseas giants like Nvidia; On the other hand, they must also face the "favorite sons" of big companies (such as Alibaba Pingtouge and Baidu Kunlun Chip).

Currently, domestic chips are mostly used for post-model training, fine-tuning, and inference processes. To achieve extreme cost-effectiveness in the high-concurrency, high-throughput internet scenarios on the consumer end is a huge test of chip manufacturers' technical capabilities.

| Window of Opportunity under Capacity Shortage: Who Will Stand Head?

Despite fierce competition, NVIDIA's chip supply constraints and domestic computing power shortages have opened a valuable window for domestic chips. Against the backdrop of limited supply, as long as chips can be produced, the market can basically absorb the entire supply.

Judging from the current commercialization progress, the domestic AI chip sector shows clear segmentation in tiers.

Huawei holds the top spot with nearly 50% market share, followed closely by Alibaba's Pingtouge. In terms of securing actual orders from major internet companies, Huawei and Cambricon have progressed the most smoothly. For example, Cambricon's collaboration with ByteDance has attracted much attention; in just the first quarter of 2025, revenue from "Customer A" (most likely ByteDance) accounted for as much as 96.48%.

However, the strict testing thresholds and extremely high migration costs of major companies mean that only a few chip manufacturers with strong technical capabilities and high ecosystem compatibility can remain. In the future, market share will accelerate to concentrate at the top.

Kingtec Perspective | Focus on leading AI chip stocks

Focus on chip companies that have established substantive order relationships with leading giants such as ByteDance, Tencent, and Alibaba, and have completed hardware-software coordinated optimization in real business scenarios (such as Cambricon, Huawei Ascend supply chain, etc.). These companies have extremely strong performance certainty.

In the era of large models, computing power is not just a competition of hardware, but also a battle of software ecosystems. Chip manufacturers that can provide a complete software stack and help major companies achieve deep adaptation and optimization will receive higher product premiums and deeper moats.

With the surge in shipments of domestic AI chips, demand for upstream wafer foundries, advanced packaging (such as Chiplets), and testing equipment is also set to surge. Upstream core suppliers with domestic substitution capabilities will continue to benefit as "water sellers."

Tencent's statement has injected a strong boost into the domestic AI chip sector. In the current computing power shortage, chip companies can get a share of the market, but "beyond that, it depends on their own abilities."