The first central enterprise AI unicorn emerges! Backed by self-developed large models, supported by four national team capital endorsements

The first AI unicorn of a state-owned enterprise has been born.

China Telecom Artificial Intelligence Technology (Beijing) Co., Ltd. (hereinafter referred to as China Telecom AI Company) announced the completion of its first round of capital increase, introducing four national-level strategic investors. This move makes the "national team" banner of this AI unicorn even more solid and prominent.

This is a recognition of its achievements. Since the launch of the AI 2.0 paradigm with large models, China Telecom AI Company has become a representative of the national team and has firmly adhered to the self-research route from day one, actively responding to the national development direction of "open source collaboration and independent control," continuously increasing investment in basic capability construction.

Now, this early-deployed AI national team has also become the first AI unicorn among state-owned enterprises, achieving a closed loop from technical recognition to market recognition.

The replenishment of ammunition and supplies is also a sign of a greater decisive battle in global AI competition. The competition for large models has converged after the first phase and has reached the finals, while China Telecom AI Company, with its troops not yet moved, has ample supplies—backed by a solid national strategy.

Major breakthroughs in the technology sector are almost always a precursor to the rise of great powers in history.

Today, AI is undoubtedly that quiet thunder, brewing silently, waiting to explode at any moment.

Against this backdrop, when a unicorn from a state-owned enterprise chooses to enter the fiercely competitive AI arena with self-researched models, how will it leverage the undercurrents of global technological competition?

The First AI Unicorn of a State-Owned Enterprise

AI competition has completely entered a white-hot phase.

In the fantastical year of 2025, first came the emergence of DeepSeek, breaking the existing industrial pattern; then, Silicon Valley tech giants fully engaged in battle, continuously increasing their scaling leverage; ultimately, with Google making a comeback against the wind, surpassing OpenAI, bringing the curtain down.

China Telecom AI Company's entry into the unicorn ranks is like a player suddenly appearing in a national team uniform, charging ahead in the new round of competition, fully pressing the accelerator.

The strategic financing partners of this state-owned unicorn are all well-known:

First, there is the National Artificial Intelligence Fund and the Beijing Artificial Intelligence Fund. Both funds represent a tilt of national-level resources. However, they focus on different directions in terms of specific oversight:

The former is a leading strategic guiding fund that does not pursue short-term monetization exits but aims for long-term investments in AI infrastructure such as chips and algorithms. The existence of such policy-driven capital can provide a safety net for "high-risk, long-cycle, choke-point" projects. It allows enterprises to be slightly "slower" in this impatient market environment, prioritizing focus on the technology itself.

The latter, funded by the Beijing municipal government, focuses on AI ecosystem construction, creating an industrial closed loop.

Unlike the previous "single-point combat" paradigm of technological development, this round of technological revolution emphasizes "concentrating efforts to accomplish major tasks." Taking the GPT moment as an example, this astonishing emergence requires breakthroughs in multiple areas such as the Transformer algorithm, GPU engineering, and high-quality datasets With the highly recognized "business card" of the Beijing Artificial Intelligence Fund, companies may be able to more smoothly embed themselves in the upstream and downstream of the ecological niche, maximizing the utilization of various resources and connecting to vast industrial application scenarios.

Huayu High-tech Traffic Control Fund, this is an investment platform under CRRC Capital. This means that it is backed by substantial resources from the rail transit physical industry.

In the current technological cycle, the landing scenarios for AI products remain highly ambiguous, and truly scalable and sustainable scenarios are rare. Such industrial capital can provide a natural "testing ground" for AI. Once successful, it can bring continuous revenue to companies for a long time.

This predictable monetization capability demonstrates its importance when the market environment tightens and investors' balance sheets are damaged.

After all, Google has proven one thing with the turnaround of Gemini 3 Pro: a viable business model can provide extremely valuable room for trial and error for companies.

Lastly, the CCTV Integrated Media Fund, in collaboration with China Central Television, possesses top-tier promotional resources.

In the AI era, trust itself is a scarce resource. Having the endorsement of a central media undoubtedly makes it easier to gain market attention and trust.

When these four national team investors appear together, the signal they convey is clear.

After all, on one side is the guiding capital that provides insights into policy direction, and on the other side is the physical industrial testing ground that can rigorously test AI. Once refined, there is also authoritative media that can provide "endorsement." The sparks generated by the collision of these cross-resource pieces are exciting to think about.

Compared to simple financing, this time, China Telecom's AI company seems more like it is using "capital increase" to gather the "masters" scattered across various industries, working together to tackle industry technology and industrial barriers that a single enterprise cannot solve.

In early 2026, facing the impending new wave of technological upheaval, the emergence of such a national team AI unicorn may allow China's AI industry to face the unprecedented changes of the century with greater composure.

Why China Telecom's AI Company?

Capital and identity have already indicated that it is "promising," but the underlying reason still relies on strength.

Since it is an AI company, the first thing to look at is, of course, the quality of the foundational models that serve as the technical base.

The confidence of this central enterprise AI unicorn comes from its "all-modal, all-size, all-domestic" "three-all" model system—the "Star" series large models, which cover four core capabilities: semantics, speech, vision, and multimodal, ranking among the top in several authoritative domestic and international evaluations.

The Star Semantic Large Model is built on a fully domesticated Wanka cluster and a domestic deep learning training framework. Recently, China Telecom's AI company officially open-sourced the MoE semantic large model TeleChat3-105B-A4.7-Thinking, which ranks alongside industry leaders on multiple benchmarks. More strategically significant is that its semantic large model has surpassed 400,000 cumulative downloads, becoming an important force in the domestic ecological open-source landscape Xingchen Voice Large Model, covering three core capabilities of speech recognition, speech generation, and end-to-end voice dialogue, supports free mixing of over 60 dialects and has been implemented on a large scale in China Telecom's WanHao intelligent customer service, handling more than 1 million calls daily.

Xingchen Vision Large Model, achieving internationally leading levels in cross-modal retrieval, supports precise natural language searches for images or videos, and has been put into use in practical scenarios such as flood prevention, emergency command and rescue, and electric bicycle entry recognition in multiple cities.

Xingchen Multimodal Large Model, integrating semantic understanding, speech synthesis, and digital human technologies, lays the foundation for more complex content production and interaction forms.

These breakthroughs not only indicate that China Telecom's AI company's model foundation is quite solid, but more importantly, they represent that the national team is personally taking action to gradually break the technological monopoly of Silicon Valley.

Perhaps this is the reason why the Xingchen Large Model was successfully selected as one of the top ten national treasures of central enterprises for 2025.

However, the true significance of self-developed models goes beyond this.

Mastering technological discourse power allows enterprises to have stronger confidence when facing strategic and product route choices.

In fact, with this confidence, China Telecom's AI company has successfully achieved a dual blooming of both toC and toB, and has reaped abundant results.

First, let's look at the C-end:

The first video generation creation tool of a central enterprise, TeleStudio, supports 2K resolution, long shots over 20 seconds, and precise action generation, maintaining high-quality output through voice-driven lip-sync and music-driven actions.

Tianyi Smart Anti-Fraud provides full-process call protection, including risk reminders for incoming calls, real-time alerts during calls, and risk reports after hang-up.

The imagination in hardware is even richer, including AI glasses that can translate in real-time and AI companion dolls aimed at family parenting scenarios.

The forms are diverse, but the underlying logic is the same: to let the Xingchen Large Model permeate all aspects of daily life, helping users free their hands from the smartphone screen and complete intelligent interactions in a more natural and elegant way.

Now turning to the B-end:

If TeleStudio and Tianyi AI glasses are partners that enhance user creation and quality of life, then the Xingchen Super Intelligent Agent is a "super colleague" deeply embedded in enterprise business.

For the intelligent agent to create value in the industry, the most critical step is to align model capabilities with enterprise business needs. Any disconnect in this chain will significantly reduce effectiveness.

Therefore, rather than merely emphasizing "simplicity and quick onboarding," the Xingchen Super Intelligent Agent places greater emphasis on local adaptation capabilities. It provides enterprises with higher degrees of customization, facilitating finer-grained adjustments to the model around their own business.

However, in real-life work scenarios, there are so many that no single company, no matter how hard it tries, can cover all long-tail fields.

To truly empower various industries, an ecosystem is needed To this end, China Telecom's AI company has created a one-stop intelligent agent "training base," the Xingchen MaaS platform.

With the complete toolkit provided by this platform, enterprises can flexibly assemble their own exclusive "AI legion" according to their needs.

It is worth noting that the Xingchen MaaS platform is not merely a technology provider. To truly grow the ecosystem, China Telecom's AI company also helps clients learn to deploy, manage, and optimize intelligent agents through systematic training.

As of now, the public cloud of the Xingchen intelligent agent platform has served tens of thousands of developers, with the total number of intelligent agents created exceeding 100,000; on the privatization deployment front, related solutions have successfully landed in more than 50 large state-owned enterprises.

Ultimately, the national team is inherently skilled at promoting industrial implementation.

There are three layers of background behind this:

First, state-owned enterprises themselves have a vast number of implementable scenarios.

In this training ground, new products can quickly receive feedback through internal testing and then iterate based on that.

What is ultimately delivered is a complete product that has been rigorously refined and backed by a wealth of verifiable data.

Second, China Telecom is already experienced and has rich marketing channels in terms of industrial implementation.

In 2024, its industrial digital revenue reached 146.6 billion yuan, accounting for nearly 30% of the total revenue for that year. Through deep cooperation with major state-owned enterprises and municipal governments, it has accumulated valuable experience.

Additionally, China Telecom has over 1 billion public customers, tens of millions of government and enterprise customers, and a nationwide channel network, providing a natural soil for the application and implementation of AI.

Third, and often underestimated: state-owned enterprises have very strict requirements for security and compliance.

In the context where intelligent agents need to be deeply integrated with business architecture to create value, this is not just an added bonus for enterprises, but a prerequisite that must be included in decision-making considerations.

In summary, state-owned enterprises are naturally suited to serve as the bridge that "connects cutting-edge technology with industry."

This may gradually become a division of labor in the future AI industry: foundational model manufacturers focus on researching models, tech giants emphasize ecosystem building, and startups delve into vertical scenarios...

And the national team is responsible for pushing technology towards large-scale industrial implementation, allowing more ordinary people to reap the benefits of technological development.

In the face of the fiercely competitive AI life-and-death race, China Telecom's AI company, which could have only controlled the macro direction, has resolutely entered the technical front line, fully committed to deepening the research and development of large models.

This is both a response to national policy guidance and a choice made by China Telecom's AI company based on its own mission and responsibility.

After all, only by having the discourse power of technology in its own hands can it avoid the passive situation of the Android ecosystem being "shut down" by Google back in the day. It can maintain an autonomous pace and maximize the value of AI.

Slow down, go deeper, look further

When a central enterprise sits at the global AI unicorn table, it may bring a more tranquil yet long-term landscape to this fiercely competitive industry.

The AI company of China Telecom, representing the national team, is also speeding down the highway of AI development, but at its core, it always adheres to the four words "long-termism."

Taking data labeling as an example, the AI company of China Telecom is not just a "technical party" providing a platform, but also plays three roles: ecological planner, capacity operator, and new profession cultivator.

As a planner, it combines local industrial characteristics to assist the government in introducing upstream and downstream enterprises, building a data ecosystem.

As an operator, it allocates platform resources to provide stable and sustainable orders for labeling bases across various regions, ensuring "there is work to do."

As a cultivator, it establishes a standardized training system to continuously incubate new talents such as data labelers and AI trainers.

Ultimately, what the national team pursues is not short-term project profits, but rather to drive industrial development through technological empowerment, cultivate a prosperous AI ecosystem through industrial development, and ultimately achieve long-term value that benefits multiple parties.

The positive externalities brought by ecological prosperity far exceed the short-term benefits of any single project.

Now, with the AI company of China Telecom joining the ranks of AI unicorns, it also represents that its persistent "long-termism" approach has gained genuine recognition in the capital market.

And when the AI industry sees a unicorn characterized by long-termism, we may finally have the opportunity to cool down this overly heated machine—

Sometimes, going a little slower may lead to going further.

The global AI competitive landscape is becoming increasingly complex, and China is beginning to reshape the global AI competitive pattern. The technological revolution is a prelude; what truly brings about significant changes in the global competitive landscape is often the industrial revolution that accompanies the update of technological paradigms.

This means that for the country to seize this opportunity, it must not only break through technological bottlenecks but also simultaneously develop related industrial infrastructure. This is a larger-scale, longer-cycle, and more challenging task.

In this context, the AI company of China Telecom, as a national team entity, first broke through the existing technological hegemony structure with self-developed models, and now further increases its investment, collaborating with four cross-resource national team capitals to comprehensively coordinate the industrial chain and accelerate the promotion of AI empowerment across various industries.

And when AI truly achieves large-scale implementation, ultimately permeating every aspect of everyone's daily life, we will know—what the national team + self-developed models can truly bring in value.

"The first AI unicorn of a central enterprise" may just be the starting point.

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at their own risk