In 2026, for the first time, competition in the AI field will no longer be just a contest of model capabilities.

It began to evolve into a series of collisions of deep paradigms:

The United States vs China, closed source vs open source, super capital vs extreme efficiency, AI empire vs AI industrialization.

If you pay attention to this industry every day, you will find an increasingly obvious change:

The AI world is splitting into two completely different development routes.

On one side is the United States.

OpenAI, Anthropic, and xAI are using huge amounts of financing, super data centers, and GPUs around the world to build an unprecedented AI infrastructure empire.

On the other side is China.

DeepSeek, Zhipu AI, Kimi, and MiniMax are beginning to break through in another way: lower cost, higher engineering efficiency, stronger open source ecosystem, and faster domestic adaptation.

On May 8, 2026, a revelation from The Information brought the core players of this "China route" to the surface and made the world aware of fundamental changes.

DeepSeek financing, why did it shock the global AI circle?

It is reported that DeepSeek is planning to conduct its first round of external financing, with a single amount of up to an astonishing 50 billion yuan (approximately 7.35 billion US dollars). What’s even more striking is that its founder Liang Wenfeng may invest 20 billion yuan himself, accounting for 40% of the entire round of financing.

This company, which has always been independent, did not raise funds, and was not eager to commercialize, suddenly opened up the road to capital. This in itself is a strong signal: for the first time, a basic model star company with truly global influence has emerged in China.

But what is more important than the 50 billion valuation is DeepSeek's unique rise. Over the past year, it has violated almost all the growth logic of Silicon Valley AI companies.

ReviewOpenAI’s path: super financing → super GPU cluster → super data center → Microsoft support → global capital promotion.

AndDeepSeek's path is: long-term no financing → extremely low training costs → prioritizing engineering optimization → large-scale open source → strengthening domestic chip adaptation.

It is not like a traditional AI startup company, but more like a "technical efficiency monster". According to multiple media reports, DeepSeek’s latest valuation has reached US$45 billion to US$50 billion, and the capital behind it includes the National AI Fund, the state-owned assets system and semiconductor industry capital. This shows that DeepSeek is transforming from a pure "technology laboratory" to "a core member of China's AI national team".

What really deserves attention is not valuation,

But "route split"

Many people still understand AI competition as "whose model is smarter." But now, an increasingly obvious fact is that there have been fundamental differences in the paths of global AI.

Let's look at the American route first. Its core can be summarized as: Super Capital + AI Empire.

American AI companies are entering an extremely exaggerated stage of development. They are becoming less and less like "technology companies" and more like "digital infrastructure countries."

• OpenAI, with a valuation of US$852 billion and its latest financing of US$122 billion, is backed by SoftBank, Microsoft, and NVIDIA, and is promoting a super AI infrastructure plan called "Stargate" with a total scale of US$500 billion.

• Anthropic, has a valuation discussion approaching US$1 trillion. With the support of Amazon and Google, it is wildly expanding the Claude Code, Computer Use and enterprise agent ecosystems, and has begun to bind Google TPU to the AWS cloud.

• xAI, valued at US$200 billion, relies on X (Twitter) real-time data, Grok Connectors and Agent ecosystem to start developing towards an AI operating system.

You will find that the core characteristics of the American route are very unified:

1.Super financing:AI has become a purely capital-intensive industry.

2.Super computing power: The core competition revolves around GPU, data center, power and cloud infrastructure.

3.Closed source ecology:OpenAI and Anthropic increasingly emphasize private models, corporate barriers and platform control.

4.AI Imperialization: Their goal is no longer to "train a model", but to establish a complete set of AI infrastructure.

However, on the other side of the ocean, a change with completely different logic is taking place.

China’s route: AI industrialization and ultimate efficiency

Chinese AI companies are taking another path. The biggest feature of this path is "efficiency driven" rather than "capital driven".

We can summarize the core characteristics of the "China route" as follows:

1.More emphasis on open source: A large part of DeepSeek's global influence comes from its successful open source ecosystem.

2.More emphasis on cost: What Chinese model companies are best at is not "burning more GPUs", but "achieving similar results at significantly lower costs".

3.More emphasis on engineering optimization:Including MoE (mixed experts) architecture, inference optimization, scheduling optimization, model compression, etc.

4.More emphasis on domestic adaptation: In particular, in-depth cooperation with domestic chips such as Huawei Ascend has become a key variable in China's AI route.

However, it should be noted that China’s AI industry is not monolithic. DeepSeek and Zhipu AI have actually began to represent two different directions.

Many people tend to confuse China's large model companies with one another, but in fact:

• DeepSeek is more like an idealistic hybrid: It's a bit like "early OpenAI" combined with "Linux" and "Android". It represents technological revolution, open source diffusion, extreme cost optimization and developer ecology. Its goal is to become the basic model capability layer of global AI.

• Zhipu AI is more like a pragmatic platform player: It is becoming more and more like the "Chinese version of Anthropic". It places more emphasis on enterprise agents, reasoning capabilities, localized ecology and enterprise-level platforms. In particular, its GLM, AutoGLM and Agent systems have clearly pointed towards the "enterprise AI operating system".

The coexistence of the two constitutes a completely new situation in China's AI.

The global AI power map has changed, and the next stage is

"Infrastructure War"

This is very important. In the past, Chinese Internet companies mostly copied the American model.

But in the AI era, we have for the first time "two completely different Chinese routes" : the DeepSeek route (global developers, open source, technology diffusion, extremely low cost) and the Smart Spectrum route (enterprise market, Agent platform, domestic ecology, enterprise closed loop).

This means that China's AI has begun to form a real ecological layering.

Looking back to the global map, looking at valuations, the current pattern is very clear:

The biggest significance of this inventory is not that "China also has big model companies", but China has begun to appear players who can truly enter the global AI core poker table.

Many people are still debating whether GPT-5 is stronger, Claude 4 is stronger, or DeepSeek-R2 is stronger. But what is really important in the future may not be "whose model is the strongest at a single point", but who can establish a complete AI ecological closed loop.

The next stage of competition will evolve into a comprehensive "AI infrastructure war". What really determines the outcome will be control over the following areas: Chip, cloud, data center, Agent, API, developer ecosystem, operating system, enterprise workflow...

Because, top AI companies have become more and more like "digital countries".

They are building computing power, building an ecosystem, cultivating a developer system, laying out a payment/API network, and weaving an Agent system—in essence, they are competing for the right to define and dominate the next generation of digital infrastructure.

Write at the end

China’s real opportunities for AI may have just begun

In the past 20 years, China's Internet has been catching up with the United States. But in the AI ​​era, for the first time, "different technical routes" can develop in parallel.

• American route: Super capital, super GPU, closed source ecology, and AI empire.

• China route: Open source, extreme engineering optimization, localization, low cost, and AI industrialization.

Who will win in the end? No one knows the answer now. But one thing is becoming increasingly clear:

The next stage of AI competition is no longer just a competition in model capabilities, but a comprehensive competition in ecology, energy, chips and industrial organization capabilities.

DeepSeek's valuation of US$50 billion may be the clarion call for this new era to truly kick off.