Is Alibaba "fierce" again? A serious discussion on the second wave of cognitive reassessment of Chinese technology

With the announcement of Alibaba's largest parameter model Qwen3-Max, the full product iteration from Qwen 2.5 to 3 has been basically completed. At the same time, Alibaba also announced that the originally planned three-year AI infrastructure capital expenditure of 380 billion yuan will be further increased.

Amid the dual logic of reshaping expectations in the e-commerce battlefield and AI reinventing Alibaba, the revaluation of Alibaba is ongoing.

However, Dolphin Research mentioned the revaluation of Alibaba a few days ago here. This article takes this opportunity to discuss the long-term logical impact of China's AI technology industry, following the revaluation of Alibaba as the main line after the Deep Seek at the beginning of the year— the second wave of cognitive revaluation of Chinese technology and the sustainability of the market.

1. First Catalyst for Revaluation: Not Only H20 is Unusable, RTX 6000 is Also Not Feasible

For Nvidia's old H20 chips, domestic customers need to explain the rationality of using H20 to rule out potential backdoor risks. For the new Blackwell architecture-based RTX PRO 6000D, according to foreign media reports, it faces similar issues and is difficult to deploy.

Of course, the basic premise for doing this domestically is:

a. The RTX PRO 6000 itself does not have very high performance, especially since the memory has reverted to GDDR, severely affecting memory bandwidth. At the same time, the special version for China has been significantly downgraded.

However, since it is a 4-nanometer chip from TSMC, with Blackwell architecture and CUDA, and the price has indeed dropped significantly, there is still some interest in testing it. But with regulatory constraints, this demand has basically stalled.

b. Domestic Competitors are Rising

Domestic AI chip design companies have gradually risen to at least a level that can compete with the RTX Pro 6000 and H20. Many of their performance parameters are not significantly different from the products Nvidia can offer in China.

Although domestic competitors are also trying to be compatible with Nvidia's CUDA, Nvidia still has an unparalleled advantage in the software ecosystem, and there are still hard constraints on wafer foundry capacity (domestic circulation relies on SMIC, while external circulation temporarily relies on Samsung foundry).

However, with Nvidia's products unable to pass smoothly domestically, and domestic cloud service providers increasing their CAPEX, but chip design companies' revenues still being low, the logic of domestic substitution is not far-fetched.

2. Second Catalyst for Revaluation: Overseas Market Reflection

In the latest earnings season of the US stock market, although the two GPU players Nvidia and AMD did not significantly exceed expectations, Broadcom's stunning performance in ASICs and the rapid advancement of Google's TPU and cloud business showed the market a second path to computing power beyond Nvidia's absolute monopoly, a path that Chinese players might be able to replicate.

Next, Oracle's large infrastructure orders and Nvidia's cooperation with OpenAI first confirmed the certainty of AI in application scenarios.

Putting these two together, it is clear that the pattern across the ocean is very clear: AI has certain scenarios and demands, and now CSPs need to reduce computing costs (on the basis of GPUs, ASICs can prosper the computing power ecosystem, bypass Nvidia's tax to build their own inference computing power, and accelerate the reduction of computing costs).

In this context, downstream core players such as domestic large model players like Alibaba (during the Yunqi Conference) frequently update large models and also express their intention to continue increasing capital expenditure on the basis of the original three-year 380 billion yuan.

In this context, if we link , a very clear issue emerges:

a. After a series of US-China games, whether it is H20 or the severely downgraded RTX 6000D, they are not usable; but downstream customers like Alibaba Cloud, ByteDance, and Tencent have given massive AI infrastructure investment budgets at the beginning of the year;

b. Domestic AI chip capabilities, led by companies like Huawei, are also rapidly iterating. In this context, the original 7:3 or 6:4 overseas chip: domestic chip procurement budget of domestic cloud service providers may shift to 5:5 in the future.

The surge in performance of overseas ASIC companies indicates that there is currently a possibility of bypassing Nvidia's technical prohibitions and monopolies in the computing power market; it also brings hope for breakthroughs in domestic chips.

In this context, the revaluation of Chinese technology assets can be said to be interpretable, imaginable, expectable, and realizable, rather than mere speculation of domestic substitution.

3. Third Catalyst for Revaluation: From Tiktok Controversy: The Technology Wall Has Fallen, What Justifies the Valuation Wall?

The Tiktok issue seems to be coming to an end:

1) In terms of business control in the US, through a series of complex operations, it is transferred to a consortium composed of Oracle, Silver Lake, etc., holding 80% of the shares; ByteDance holds the remaining 20%;

2) Localized control and operation, data does not leave the country, stored on Oracle's local servers, and data training is completed within the US.

3) Core controversy: Tiktok's algorithm is authorized (note: not transferred or sold) by ByteDance to the new company for use, with a 10-year authorization period (renewable).

The controversy surrounding Tiktok seems to be resolved through the framework principle of "technology authorization + local operation", with the remaining details being more about the pricing of equity redistribution, specific charging methods for technology authorization, etc.

Dolphin Research does not discuss the details here, but rather the logical derivation of this issue in the capital market: does the resolution of this controversy indicate that the flow of "advanced technology" between the East and the West is not completely blocked, and after a series of negotiations, the Tiktok issue is resolved in principle, and after the details are finalized, there may still be a thawing journey between the high-levels of China and the US:

It brings us a soul-searching question:

a) If Eastern technology can flow to the West, is it possible that Western technology can also flow to the East through similar mechanisms? Ultimately, everything on the negotiation table has a price tag, meaning it is negotiable.

b) If the "technology flow wall" between the East and the West has fallen, and political thawing occurs, what is the point of the "valuation wall" that the investment circle has been adhering to between the East and the West? What is the significance of Alibaba's valuation at around 15X PE vs. Amazon's 22X PE?

Of course, upon closer examination, this logic may have many flaws, but once funds notice this issue, they may make corresponding position adjustments.

Risk Disclosure and Statement of This Article:Dolphin Research Disclaimer and General Disclosure