A detailed explanation of ZTE's AI landscape: Only when there is a fertile soil of digital foundation can AI applications thrive.

In early March, the 2025 Mobile World Congress (MWC2025) was held in Barcelona, Spain. The ZTE exhibition area at Hall 3 of Fira Gran Via was crowded with people.

Beyond the figures of people of different skin colors, surrounded layer by layer, was the "Tina" humanoid robot jointly developed by ZTE and Sichuan Tianlian. Tina can not only have free conversations with the audience but also invite users to take photos together and present a small gift with its robotic arm... Its flexibility is just like that of a skilled and responsible on - site host.

Compared with the dull and mechanical state of old - fashioned robots, today's robots are more intelligent and agile, or rather, more like "humans". The remarkable leap in the robot industry is inseparable from the technological upgrades in key areas such as communication, algorithms, and chips.

While exhibiting the "Tina" humanoid robot, ZTE also showed off its full - stack AI capabilities without hesitation.

From the underlying network communication equipment, including a variety of new FWA (Fixed Wireless Access) & MBB (Mobile Broadband) products built on cutting - edge technologies such as AI, 5G - A, and Wi - Fi 7; to the innovative terminals at the application layer, including the second - generation small - fold Nubia Flip 2, AI headphones and other smart wearable devices - all these demonstrate ZTE's AI - driven full - scenario smart ecosystem, which also fits well with ZTE's theme for MWC2025: "AI for All".

There was also a corner at the ZTE booth that was equally crowded with audiences. However, different from the onlookers around the Tina robot, in the display area of ZTE's "AiCube DeepSeek Intelligent Computing All - in - One Machine", most of the people who came to inquire were professionals in suits and ties, who had pain points in enterprise AI implementation and were looking for cooperation.

After DeepSeek amazed the industry with its super - strong performance and highly cost - effective training and inference costs, the whole industry witnessed a wave of "connecting to DeepSeek". How to enable all industries to enjoy the technological inclusive capabilities brought by DeepSeek with lower thresholds and higher efficiency has become a common demand among enterprise customers.

The AiCube DeepSeek Intelligent Computing All - in - One Machine launched by ZTE solves the "last - mile" problem of AI model commercialization.

By integrating high - computing - power hardware, an easy - to - use training and inference platform, and mainstream large models, the AiCube DeepSeek Intelligent Computing All - in - One Machine supports the full versions of the DeepSeek series of models, including the full - blooded version and the distilled version. Through the one - click deployment function, users can quickly apply AI models to actual business in a short time, accelerating the entire process from model development to launch and significantly improving the speed of product launch.

At the infrastructure level of underlying computing power, ZTE has formed a device product matrix based on different scenarios such as personal and household terminals and industrial Internet terminals, which includes:

In terms of mobile networks for operators, ZTE continues to lead UBR innovation, launching the industry's first TDD + FDD dual - mode Massive MIMO AAU and 400M OBW large - bandwidth AAU to build minimalist, highly efficient, and green stations.

In the field of all - optical networks, ZTE is committed to promoting the high - quality development of optical networks. For all - optical homes, it increases business revenue through Wi - Fi 7 and FTTR, reduces operation costs through home digital tools, and upgrades scenarios through AI - innovative smart terminals; for all - optical cities, it expands three scenarios of office, hotel, and campus through all - optical ToB to increase customer ARPU; for all - optical infrastructure, it launches intelligent ODN, Full - Band OTN, and AI - Enhanced IP to build a minimalist, ultra - fast, and highly efficient basic bearer network.

In the current wave of the AI boom, ZTE, which has been established for 40 years, is not at all slow or retreating in the face of the industrial leap opportunities brought by AI. Instead, it is forging ahead bravely and actively embracing the transformation opportunities brought by digitalization and intelligence.

At the ZTE booth, Chen Zhiping, the vice - president of ZTE, told 36Kr that ZTE's approach to embracing AI can be described as "both internal and external", where:

Internally, ZTE has widely adopted AI in code generation, equipment manufacturing and other processes, which has greatly improved efficiency in multiple internal enterprise scenarios;

Externally, ZTE provides customers with complete computing power equipment, end - to - end communication, large - model platforms and other technological capabilities, helping customers achieve a complete set of services for joint innovation from "infrastructure - software platform - application scenarios".

"We need to build a distributed implementation of computing power for enterprises, including the edge side, the cloud side, and the end side. In fact, this is to build a carrier for intelligent implementation. So infrastructure first is definitely a prerequisite for us to let AI create value." Chen Zhiping emphasized to 36Kr.

"To get rich, build roads first" in the AI era

Remember the "Tina" humanoid robot mentioned earlier? To enable her to recognize commands and know how to shake hands like a human being, this seemingly simple action actually involves the reconstruction and upgrading of a whole set of cloud, edge, and end transmission networks.

First, there is the consideration of cost. For robot applications, if all the computing power is deployed on the robot body, not only will the cost soar, but the battery life will also be significantly reduced by about 50%;

Second, there is the requirement for low latency. If every reaction of the robot needs to be sent back to the cloud, the lag in human - robot interaction will be particularly obvious. To improve this situation, edge computing power is needed to plan complex tasks with lower latency and generate action sequences to be assigned to multiple robots, making the robot's reaction more agile;

Finally, there are also security requirements. The collaboration between the edge and the end side can better prevent robots from encountering malicious commands and causing uncontrollable consequences.

Therefore, at the Barcelona exhibition, ZTE provided Tina with language ability and motion decision - making ability through the Nebula multi - modal large model deployed on the computing power base station, and empowered Tina with edge computing, thus achieving the combined performance requirements of high upstream bandwidth, low downstream latency, and high reliability.

It can be said that the vivid reactions of the robot are just the results that users finally see. In fact, in the areas of chips, computing power, and algorithms, ZTE has been quietly accumulating strength for many years.

Facing industry innovation, ZTE provides a full - stack AI comprehensive application solution from chips, hardware, software to integration services, aiming to achieve the best application effect and create the most cost - effective AI applications for customers. Among them, the Digital Nebula 3.0 has the Nebula large model built - in, integrating tools for the entire process of large - model training, inference, and application development to help industry customers use AI more quickly, better, and more economically.

In Chen Zhiping's view, ZTE has been operating in the ICT industry for 40 years. In these 40 years, ZTE's understanding of ICT and its ability to integrate the software and hardware of digital and intelligent infrastructure have enabled it to build a solid foundation for quickly carrying AI.

"As the saying goes, 'To get rich, build roads first'. To build a good road for AI development, digital infrastructure comes first." Chen Zhiping emphasized to 36Kr.

If the robots at the ZTE booth are still in the concept implementation stage, let's shift our focus to the Nanjing Smart Binjiang 5G Factory put into operation by ZTE in 2020.

The flexible production line of the Binjiang Factory is a 100 - meter production line composed of 28 workstations. Through 5G CPE technology, the pictures taken by the high - definition cameras on the robotic arms are transmitted to the background for centralized calculation and analysis, and then fed back to the robotic arms to perform actions, realizing the flexible installation of accessories. It can accurately and efficiently assemble products according to different order requirements, solving the problem of low production efficiency in traditional production lines that need to stop and change lines when processing different orders, and endowing the machines with a "brain" and the ability to respond similar to that of humans.

Nanjing Smart Binjiang 5G Factory

Compared with traditional production methods, the flexible production line can achieve personalized and continuous production processes. The automation rate of the entire production line reaches over 95%, the production quality is improved by over 50%, the production efficiency is increased by over 125%, and at the same time, costs and material losses are effectively reduced.

With such a strong digital foundation, it has become particularly smooth for the Nanjing Binjiang Factory to integrate into the AI - enabled production efficiency scenarios.

Through the Nebula industrial large model, the Nanjing Binjiang Factory can speed up the generation of process documents by 10 times, reduce the quality inspection labor cost by 70%, and at the same time empower external parties to create innovative applications in industries such as industry, transportation, and energy to realize the monetization of AI value.

The more fertile the soil of AI infrastructure, the more prosperous the applications will be

Under the successive waves of AI set off by ChatGPT and DeepSeek, for enterprises with a clear understanding, blindly chasing and learning is not the way to go, as this often leads to the inefficient situation of reinventing the wheel.

In this regard, ZTE's thinking is: Excellent large models should not be isolated from each other and fight alone. Instead, they should be like a think - tank composed of various experts, jointly providing AI services for users.

Therefore, ZTE's terminal business proposed the concept of "AI Together" for the first time in the industry. Through the Nebula Engine, it can intelligently call multiple expert large models, including Nebula, Doubao, and DeepSeek, to achieve system - level integration of AI large models, allowing the mobile phone to automatically select the most suitable expert to solve problems.

What really excites the industry is that ZTE's "AI Together" is not just lip service. The Nubia Flip 2 it launched has become the first small - fold new product with a full - size embedded DeepSeek. It supports a new voice interaction method of softly waking up the AI pet and has functions such as anthropomorphic chatting, Q&A, AI call translation, and AI office, providing users with more emotional value.

Based on such an open and efficient ecological cooperation concept, users of ZTE's terminal products can enjoy the most cutting - edge technological capabilities.

During the Barcelona exhibition in March, ZTE's Nubia Nebula AI pushed an update with a full - size embedded DeepSeek version. The new embedded DeepSeek has taken the lead in the industry to realize the intelligent online search function, which can intelligently judge whether online search is needed according to the scenario without manual switching by users, providing a smoother and more intelligent search experience.

The new embedded DeepSeek has also updated the voice interaction function. Users can easily activate it through voice for in - depth reasoning and analysis. In addition, the generated content supports easy viewing of historical records and can be saved to the notepad with one click, facilitating users to review and organize at any time. Currently, the new embedded DeepSeek has been pushed to the Nubia Z70 Ultra model and will be updated on more models in the future.

It's not just mobile phones. As the combination ability of AI + hardware models becomes more in - depth, terminal products are also given more possibilities by AI.

ZTE has extended its AI capabilities to more diverse terminal forms. It has released the world's first 2 - in - 1 5G + AI cloud computer product and multi - form AI terminals designed specifically for household users.

For the market of companion - type AI toys, ZTE has launched a companion - type AI companion solution integrating a multi - modal AI large model and 4G + Wi - Fi dual - network fusion based on its self - developed chip platform. It can be flexibly configured to access AI large models such as Doubao and DeepSeek, realizing real - time and efficient voice interaction experience and learning companionship capabilities in all scenarios.

So, what kind of revolution will AI bring to terminal applications? We also posed this "problem" to Chen Zhiping.

Chen Zhiping did not give a hasty conclusion to the currently changing AI hardware market, but amidst the changes, there are always constants.

"For terminals, in the end, we still need to focus on people's needs and experiences. It's difficult for us to define what the final product form will be, but the exploration process itself is full of imagination." Chen Zhiping told 36Kr.

On the Barcelona stage, ZTE also had its moment of glory: Thanks to its outstanding performance in promoting mobile communication technology innovation and global digital transformation, it won three awards at the GSMA Global Mobile Awards (GLOMO): the "GSMA Foundry Innovation Award", the "Best Mobile Smart City Innovation Award", and the "Open Gateway Challenge Award". This is also the best proof of Chinese companies' core competitiveness on the world communication network technology stage.

The 2025 Mobile World Congress (MWC2025) has come to an end, but ZTE at the Barcelona exhibition is just a microcosm of its full - stack AI capabilities. From computing power infrastructure, software platforms to application scenarios, ZTE's 40 - year accumulation of communication network technology and customer service experience enable it to quickly find its position in the surging AI wave.