Tencent's ecosystem play and CATL's metering play: DeepSeek's financing hides a dual agenda

The much - hyped financing storm of DeepSeek has finally come to an end.

According to foreign media reports and confirmed by multiple domestic media outlets, DeepSeek's first - round financing is about $7 billion, and its valuation has reached up to $59 billion. Tencent plans to invest 10 billion RMB, and CATL intends to invest 5 billion RMB. The founder, Liang Wenfeng, has promised to invest 20 billion RMB himself.

This is the first time DeepSeek has opened its doors to external capital since its establishment. Previously, it was solely supported by Magic Square Quantitative, and Liang Wenfeng never allowed external shareholders to get involved. The list of new investors this time is quite interesting. Tencent and CATL appear on the same list.

In the United States, the investors of OpenAI are Microsoft and NVIDIA, and Anthropic relies on Google Cloud. Everyone is looking for allies within the same technological lineage. Tencent and CATL have the same general industrial direction, both betting on the dividends of the AI industry, but their core business demands and implementation fields are completely different.

Although they are buying shares in the same company, they have different agendas. This "sleeping in the same bed but having different dreams" is what really makes this financing worth pondering.

Tencent: After Winning a Round, It Bets on a Rival

Tencent's Hunyuan Hy3 preview won the weekly championship on the OpenRouter platform on May 7th, with 36.6 trillion Tokens, ranking first in both programming and tool calls. This result shows that Tencent's self - developed technology is not inferior. However, when the financing news came out on June 3rd, the market saw this picture: Even though its self - developed technology achieved good scores, Tencent still chose to invest 10 billion RMB to participate in a peer company, forming a dual - track strategy.

Behind this dual - track strategy lies a real gap in the C - end ecosystem and commercial implementation. According to QuestMobile data in March this year, the monthly active users of Tencent's Yuanbao were only 57.35 million, while Doubao had 345 million and Qianwen had 166 million.

But for Tencent, the more troublesome issue is the growth rate. Yuanbao only added 8.2 million new users in the first quarter, while Doubao and Qianwen added about 100 million and 126 million new users respectively. The top - ranking of Hunyuan Hy3 preview on OpenRouter was influenced by event - driven traffic and concentrated calls, not entirely from natural ecological traffic.

Tencent's capital expenditure in the first quarter was 37 billion RMB, mainly for AI - related investments, of which 31.9 billion RMB was recorded as current capital expenditure. A large amount of investment was poured into AI infrastructure, but the output was obviously disproportionate to the input. The input pricing of Hunyuan API is 1.2 yuan per million Tokens, while Doubao only charges 0.6 yuan. It has no advantage in price and is far behind in scale. Alibaba has established the ATH business group and fully shifted to the Token economy. ByteDance's Doubao quietly launched paid subscriptions on May 4th, with three tiers of 68 yuan, 200 yuan, and 500 yuan. Competitors are no longer comparing who has a larger call volume but who can really make money from the call volume.

Tencent's 10 - billion - yuan investment cannot buy the exclusive right to DeepSeek's technology. DeepSeek is open - source, and anyone can download its code. What Tencent gets is a "pre - emptive right to participate": to ensure that it will not be excluded from the next AI application explosion cycle in the tripartite pattern of ByteDance, Alibaba, and DeepSeek. Hunyuan will continue to absorb traffic in internal scenarios, and Yuanbao, QQ, and Tencent Docs can still be used within its own territory.

DeepSeek can supplement external developers and B - end enterprise customers. The dual - track strategy does not mean giving up self - development but rather completing strategic hedging through "self - development as a backup + external positioning" during the stage when AI oligarchs are taking shape.

CATL: A Battery Manufacturer's Involvement in Large Models Has Nothing to Do with Code but Focuses on Electricity Meters

While Tencent's investment in DeepSeek is still within the Internet circle, CATL's entry is completely out of it. A battery manufacturer's involvement in large models has nothing to do with code but focuses on electricity meters.

The uses of DeepSeek's financing include "increasing investment in computing power, optimizing employee salaries, and accelerating the commercialization process", among which the investment in computing power was reported first by many media. Computing power means power consumption, and power consumption means energy storage orders.

CATL's actions in the past three months form an indirectly infiltrated computing power energy chain.

In April, CATL increased its capital in Zhongheng Technology Investment, the controlling shareholder of Zhongheng Electric, by about 4.1 billion RMB, holding 49% of the shares and indirectly deploying in Zhongheng Electric's HVDC business. The actual controller of Zhongheng Electric, Zhu Guoding, remained unchanged.

In May, CATL invested about 6.5 billion RMB through its affiliated party PJ Millennium to acquire about 38.1% of the shares of Century Internet. However, CATL chose to be a "silent shareholder", and the voting rights are still dominated by the founder, Chen Sheng.

In June, CATL intends to invest about 5 billion RMB in DeepSeek. In three months, it has invested about 15.5 billion RMB. This pace has gone beyond the scope of a financial investment trial and clearly points to a strategic shift in the track.

CATL's net profit in the first quarter of this year was 20.7 billion RMB, and it has more than 410 billion RMB in cash on its books. Investing 5 billion RMB in DeepSeek only accounts for about 1.2% of its cash reserves, which is not a big deal. However, three consecutive investments related to AI infrastructure have gone beyond the scope of casual allocation by the finance department. The ceiling of the main business of power batteries is clearly visible. In April 2026, the retail penetration rate of new energy vehicles in China exceeded 60%, marking the end of the high - growth era of the industry. The energy storage market is also in a severe price war, and second - tier manufacturers such as EVE Energy, Hichen Energy Storage, and CMIG are rapidly eroding market share.

The power demand of AI data centers is this new field. At the earnings briefing on April 15th, CATL clearly mentioned that with the accelerated popularization of AI, the shortage of power and energy has become the core bottleneck, and the demand for energy storage in data centers is growing exponentially.

Industry estimates show that under the high - stability power supply demand of super - large data centers, the corresponding energy storage scale can reach 15 - 20 GWh. As large models shift from training to inference and with the deployment of multi - modal models, the power consumption of data centers is growing exponentially. The traditional power grid cannot meet this stable power supply demand, and energy storage systems, backup power supplies, and HVDC have become necessities.

The energy storage system of the Lingang Intelligent Computing Center jointly developed by CATL and SenseTime has a capacity of 17.888 MW and can save nearly 10 million RMB in electricity costs per year. The math adds up. Investing in DeepSeek is equivalent to locking in a certain and high - growth end - customer in advance. DeepSeek posted job openings for its Ulanqab data center on its official website in April. Judging from the recruitment information, the Inner Mongolia data center is in the project - planning or construction stage and has not been put into operation yet. Its computing power expansion plan is CATL's order plan.

Liang Wenfeng's Steering Wheel

At the beginning of April, the market rumored that DeepSeek's valuation was about $10 billion. The $59 billion valuation in the financing plan half a year later shows an amazing increase. This leap is difficult to attribute to a single factor. The expected release of DeepSeek V4.1, the expansion of the open - source ecosystem, and the overall enthusiasm of the Chinese AI primary market are all driving forces. The entry of Tencent and CATL provides industrial endorsement for this valuation and magnifies the valuation premium. It is not that capital entry boosts the fundamentals, but that capital recognizes the fundamentals and thus magnifies the valuation premium.

Tencent brings the imagination space of application scenarios, and CATL (through its affiliated party) brings the feasibility of energy infrastructure implementation. However, the fundamental reason for the rising valuation is always DeepSeek's own open - source ecosystem and the expected implementation of multi - modal products.

Ultimately, it is because DeepSeek has reached that position that it has attracted these two industrial giants.

Tencent endows DeepSeek with application - layer value, and CATL (through its affiliated party) brings energy - layer value. However, there is an implicit vulnerability in the "hub premium". Tencent needs DeepSeek's model capabilities to support super - scenarios such as WeChat, games, and cloud services. CATL needs DeepSeek's computing power expansion to digest its energy storage production capacity.

The demands of both companies are reasonable, but their directions are not the same. In the future, if Tencent requires DeepSeek to give priority to serving its ecosystem and CATL requires exclusive energy cooperation, the tension between the shareholders' interests and DeepSeek's independent commercialization will emerge.

The enterprise - level multi - modal model V4.1, scheduled to be released in June, is a crucial step for DeepSeek to shift from open - source demonstration to independent paid services. This step requires the support of shareholders but also needs to break free from the binding of shareholders.

Liang Wenfeng has invested 20 billion RMB himself, more than the combined investment of Tencent and CATL. The number of external shareholders is controlled within 10, and they are all industrial parties that can bring strategic resources, with no pure financial investment institutions. This structure is not an alliance but a loose collection of "each getting what they need".

The core purpose of Liang Wenfeng's large - scale investment is only one: to hold the steering wheel in his own hands. The completion of DeepSeek's financing is just the beginning, and the real test is yet to come. When the shareholders start to cash in their "right to participate" and "right to orders", no one can be sure whether Liang Wenfeng's absolute control can really hold the situation.

Conclusion

ByteDance and Alibaba are taking the "imperial vertical integration" route, building their own models, clouds, chips, and applications, and swallowing the entire industrial chain. Tencent and CATL have a completely different approach. They do not build everything themselves but choose to bind to an independent hub through capital and jointly bet on a third - party node.

Liang Wenfeng is the one holding the steering wheel.

This is not a matter of superiority or inferiority but a different layout path spawned by the different maturity levels of the AI industrial infrastructure in China and the United States. American AI companies are competing for GPUs, while Chinese AI companies are already competing for power, scenarios, and industrial chain support.

DeepSeek's financing is essentially an expensive ticket to enter the "infrastructure arms race". The ticket has been bought, but whether the show can go on depends on what each of these "audience members" wants to hear.