China's AI industry has reached a historic moment.

Recently, DeepSeek (Hangzhou Deep Seek) announced a new Series A financing round, with a total funding amount of about 51 billion yuan and a post-investment valuation close to 400 billion yuan.

In this highly luxurious list of investors, founder Liang Wenfeng personally contributed about 20 billion yuan, Tencent about 10 billion yuan, CATL about 5 billion yuan, and NetEase, JD.com, IDG Capital, and others each contributed about 3 billion yuan.

But the most extreme thing wasn't the amount, but the rule set by Liang Wenfeng: giants could pay money, but could only be "pure financial investors," not touching the steering wheel (no voting rights), and their shares had to be locked for five years. At this critical juncture in the AI computing power arms race, giants like Tencent and JD.com are willingly signing these "unfair terms," aiming for a deeply bound future.

| Giants rush to buy: Tencent seeks collaboration, CATL targets "electric meters"

Faced with this hot commodity, primary market investors once scramble to get on board, but Liang Wenfeng's selection criteria are extremely strict. Each of the investors he chooses carries strong industrial synergy attributes.

Compared to giants like Alibaba, which have self-developed large models (Qianwen), Tencent, NetEase, and JD.com hope to leverage investments to create synergies between DeepSeek's technology and their own AI strategies.

And the biggest surprise was CATL. Why would a new energy giant invest in large models? The answer lies in DeepSeek's plan to build its own data center in Ulanqab, Inner Mongolia. The ultimate goal of large models is computing power, and the end of computing power is electricity. CATL's investment in DeepSeek essentially aims to secure future "major electricity consumers" in advance, paving the way for its photovoltaic energy storage and data center power supply business.

| Money over Power: Breaking the Traditional 'Capital Firewall'

DeepSeek actually isn't short on money. Over the past three years, Liang Wenfeng has relied on profits earned from his subsidiary Illusion Quant to support the company's minimalist operation of "no financing and no commercialization."

However, as the global AI "arms race" intensifies, relying solely on its own funds can no longer support the massive computing power consumption. This round of open financing is precisely to bring in more capital expenditure.

However, to prevent capital from interfering with the technology route, Liang Wenfeng designed a very special transaction structure: except for the National Artificial Intelligence Industry Investment Fund (which invested about 980 million yuan and held direct shares), all external investors' funds had to be injected into a limited partnership managed by Liang Wenfeng. External investors only enjoy financial income rights and the right to be informed about information; they have no voting rights, and their shares must be locked for 5 years.

This effectively built a strong "capital firewall" for DeepSeek.

| State-owned Capital: The Long-termism of Industrial Capital

Why are the giants willing to pay for this disruptive architecture?

On one hand, this reflects investors' absolute trust in Liang Wenfeng's technical route and their long-term confidence in China's AI industry for over five years.

On the other hand, this aligns with the underlying logic changes in current AI investment.

Within this structure, the National Artificial Intelligence Industry Investment Fund directly invests, acting as a "stabilizer." Against the backdrop of China-US AI competition, the direct involvement of state-owned capital not only represents the will of the nation but also provides endorsement for corporate compliance and healthy development. With the stabilizing anchor of state-owned assets, other industrial capital is naturally willing to relinquish voting rights and be long-term financial supporters.

Kingtech Perspective | AI Investment Enters the 'Era of Industrial Capital-Led Capital'

DeepSeek's funding list marks the exit of dollar funds that purely pursue financial returns, replaced by domestic industrial capital deeply tied to the AI sector. Future AI investment will not just be about money, but also about the underlying ecosystem resources.

Laying out the "computing-power collaborative" industrial chain

CATL's entry into DeepSeek reveals the next hotspot in AI infrastructure—power and energy storage. With the large-scale construction of large model data centers in Inner Mongolia and other regions, companies with core capabilities in data center power supply, HVDC (high voltage direct), and energy storage systems will see a definite surge in orders.

Focus on domestic computing power chains with "chip-die collaboration" capabilities

The rise of DeepSeek validates the feasibility of the domestic computing power route. With the deep integration of leading large models like DeepSeek with domestic chips (such as Huawei Ascend), the entire domestic AI computing foundation (including servers, optical modules, and advanced packaging) will accelerate its performance realization period.

DeepSeek's 50 billion RMB financing is not just a capital carnival; this huge investment will accelerate DeepSeek's expansion in computing power infrastructure. The "computing-power synergy" and "domestic computing power ecosystem" derived from this investment are the structural main themes the capital market should closely watch.