DeepSeek's valuation soared to nearly US$60 billion, and Internet giants split into two camps

DeepSeek's first round of financing split Chinese Internet companies into three clearly defined factions. Behind each faction's choices are their own financial backgrounds, anxieties, and different judgments on the future of AI.

"Investment School": Three Abacus of Tencent, JD.com and NetEase

The most active "investment faction" chose to pay for the ride.

Tencent became the largest external institutional investor with 10 billion yuan. But what is less known is that before Tencent made this decision, it had just experienced a "self-research victory" - Hunyuan Hy3 preview won the OpenRouter platform single-week championship on May 7, with 3.66 trillion Tokens, ranking first in programming and tool invocation. Even so, Ma Huateng's words at the shareholders' meeting in May are still heartbreaking: "A year ago, I thought I was on the boat, but then I found that the boat was leaking; now I feel like I have stood on it, but I can't sit down."

The C-side gap behind this is the core anxiety. Data from QuestMobile in March this year showed that Tencent Yuanbao’s monthly active users were only 57.35 million, while Doubao’s monthly active users were 345 million and Qianwen’s 166 million. Even more fatal is the growth rate - Yuanbao only had 8.2 million new users in the first quarter, while Doubao and Qianwen added about 100 million and 126 million respectively. Tencent’s AI-related capital expenditure payment in the first quarter was 37 billion yuan, a large amount of which was invested in AI infrastructure. The output and investment are obviously not proportional. The Hunyuan API input price is 1.2 yuan/million Tokens, and the bean bag is only 0.6 yuan. There is no price advantage, and the scale is left behind by orders of magnitude.

By spending 10 billion to invest in DeepSeek, what Tencent got in exchange was not technology exclusivity (DeepSeek is open source, and anyone can download the code), but a "priority right to participate" - ensuring that in the top three of Byte, Alibaba, and DeepSeek, it will not be kicked out of the table in the next AI application explosion cycle. Hunyuan continues to digest traffic in internal scenarios, and DeepSeek complements external developers and B-side enterprise customers. The strategic hedging of "self-research and external blocking" is Tencent's most pragmatic AI survival strategy.

Jingdong has a different plan. JD.com's AI never intended to be a universal base from the beginning. Its goal is "the world's largest physical world operations center." This year, JD.com’s AI R&D investment has increased by more than 200%, and its applications cover more than 3,000 scenarios including retail, logistics, and industry. The self-developed JoyAI large model only serves vertical fields such as e-commerce supply chain and intelligent customer service. Spending a small amount of money to buy top-notch general-purpose technology is much more cost-effective than burning money from scratch.

NetEase's choice can be called the shrewdest "free ride". NetEase's core businesses are games, music, and education. What they need most is AI content generation and intelligent interaction, not the underlying large model. The "Yuyan" large model that was once launched has long been dormant. Later, it simply turned to the vertical field and developed the mining embodied intelligent model "Lingdig". Investing in DeepSeek is equivalent to getting the world's most advanced AI tools at the lowest cost.

"Self-research school": The bottom line of Alibaba, Byte and Baidu

The "Self-development School" who watched with cold eyes chose to build their own ships. The three companies have one thing in common: they all regard AI as a life and death battle. Alibaba's Tongyi Qianwen, Byte's Doubao, and Baidu's Wenxinyiyan have invested tens of billions in research and development funds - the general large model base is the lifeblood, and they would rather burn money to maintain the bottom line of self-research. Alibaba has established an ATH business group to fully shift to the Token economy. ByteDoubao quietly launched three paid subscriptions (68 yuan/200 yuan/500 yuan) on May 4, and Baidu Wenxin is working hard to catch up.

There are also "outsiders" like Pinduoduo and Meituan who choose to wait and see for the time being.

CATL: Battery factory invests in large models, focusing on electricity meters

The most unexpected participant in the entire financing is CATL - Why would a battery company invest in AI? The answer lies in the energy consumption requirements of computing power data centers.

CATL's actions in the field of AI infrastructure in the past three months have formed a clear computing power energy chain: in April, it increased its capital to Zhongheng Electric's controlling shareholder by approximately 4.1 billion yuan, indirectly laying out the HVDC business; in May, it invested approximately 6.5 billion yuan in 21Vianet through related parties for approximately 38.1% of its shares; in June, it invested approximately 5 billion yuan in DeepSeek. About 15.5 billion yuan in three months, which is beyond the scope of financial investment testing.

The ceiling of the main power battery industry is visible to the naked eye - in April 2026, the retail penetration rate of domestic new energy vehicles exceeded 60%, and the era of high growth rate officially ended. The power demand of AI data centers is becoming a new growth pole. Ningde Times clearly mentioned at the performance briefing that as the popularity of AI accelerates, power and energy shortages have become core bottlenecks, and the demand for data center storage has increased exponentially. Under the high and stable power supply demand of ultra-large data centers, the corresponding energy storage scale can reach 15-20GWh. CATL’s net profit in the first quarter was 20.7 billion, and it had more than 410 billion in cash on its books. The 5 billion invested in DeepSeek only accounts for about 1.2% of its cash reserves. However, DeepSeek’s computing power expansion plan (such as the Ulanqab data center) is the energy storage order growth curve of CATL.

The valuation jumped from US$45 billion to nearly US$60 billion in one month, with the entry of Tencent and CATL providing industry endorsement for this valuation. The head-to-head confrontation between the investors and the self-researchers will soon begin at the AI ​​application layer. This financing is not only a coming-of-age ceremony for DeepSeek, but also a mirror that reflects the different judgments of Chinese Internet companies on the future of AI—paying for a car or building a ship yourself. A choice must be made in 2026.