DeepSeek raises RMB 150 billion — what capital really wants to buy isn't a large model.
DeepSeek raised ~100 billion yuan, over 150 billion in 100+ days, with Tencent, CATL, SMIC Capital and Geely joining—not chasing a model but AI's entry into real industry: plugging AI into production, controlling equipment, creating value. Answering questions isn't doing work; robots need perception, control, data loops, engineering validation.
DeepSeek has raised 150 billion yuan – what capital truly wants to grab may not be a large model at all, but rather –
the entry point for AI into real industry.
According to a media report on October 22, DeepSeek's new financing round is approximately 100 billion yuan, with a pre-money valuation of about 500 billion yuan. Combined with the previous 51 billion yuan, the cumulative financing over more than a hundred days exceeds 150 billion yuan.
What is more noteworthy is not how much money there is, but who is investing.
Tencent and CATL continue to increase their stakes, and industrial capital such as SMIC Capital and Geely have also joined in.
Why are players in chips, automobiles, and batteries all rushing to invest in an AI company?
Because what they see may not just be a model, but the intelligentisation opportunity of the entire industry in the future.
In the past, AI competition was about whose model was smarter. But as model capabilities gradually become widespread, the real challenge has become:
Who can connect AI into production processes, control real equipment, and create actual value?
01.Competition Around AI Is Extending into the Real World
Liang Wenfeng's recent moves into embodied intelligence are also worth attention. He has not only invested in Unitree Robotics but is also reported to be planning to participate in other embodied intelligence startup projects.
This means competition around AI is extending into the real world.
But there is a key question here:
AI being able to answer questions does not mean AI can get things done well.
For robots to truly work, they still need perception, control, data closed loops, and engineering verification. For enterprises to truly use AI well, they also need business knowledge, system integration, and reliable delivery.
This is precisely what many people tend to overlook.
02.As Code Becomes Easier to Generate, What Is Truly Scarce?
For the software industry, AI programming is lowering development costs, but lower development costs do not mean the value of software will necessarily drop to zero.
Quite the opposite – as code becomes easier and easier to generate, what may truly be scarce is:
an understanding of the industry, and the ability to turn technology into business results.
So what this round of financing for DeepSeek reveals is:
In the future, the most valuable people may not be those best at writing code, but those who most clearly know what problem should be solved and can make AI truly solve it.
As AI becomes cheaper and cheaper, where exactly does your value lie?