MWC25: The AI Frenzy Sweeps Barcelona, Ushering in the Era of Cloud Phones and "Complete" AI Smartphones

MWC25: The AI Frenzy Sweeps Barcelona, Ushering in the Era of Cloud Phones and "Complete" AI SmartphonesWhile the industry's attitude towards artificial intelligence at MWC 2024 remained at the "interest" stage, the MWC25 exhibition, which took place on March 3rd, 2025, in Barcelona, showcased an outright "frenzy" for AI. Artificial intelligence was ubiquitous

MWC25: The AI Frenzy Sweeps Barcelona, Ushering in the Era of Cloud Phones and "Complete" AI Smartphones

While the industry's attitude towards artificial intelligence at MWC 2024 remained at the "interest" stage, the MWC25 exhibition, which took place on March 3rd, 2025, in Barcelona, showcased an outright "frenzy" for AI. Artificial intelligence was ubiquitous. Samsung's giant outdoor billboards, for example, dedicated half to its S25 series phones and the other half entirely to Galaxy AI. This spectacle confirmed the prophecy that "AI will fundamentally permeate our lives." The Lei Technology MWC frontline reporting team delved into the exhibition, bringing you an in-depth report on the AI frenzy at MWC25.

Cloud Phones: Making AI Accessible to All

The concept of "cloud" is typically associated with enterprise servers, referring to the transfer of local data to the cloud for storage and distribution, gradually replacing local functionality entirely or in part. However, at MWC25, Lei Technology witnessed another possibility of "cloud" in the mobile phone arenacloud phones.

Lenovo officially launched its cloud phone at the exhibition, outlining its four core dimensions: "terminal-network-cloud-AI." 5G-A technology and AI served as crucial catalysts in creating this cloud phone. The key lies in the fact that the high-performance demands of AI are difficult to meet with local computing power; it's not only costly but also presents significant technological challenges.

AI is powerful, but the "on-device AI" performance of many smartphones often falls short. This is due to the vast gap between local and cloud computing power. For example, with the popular DeepSeek model, the parameter limit for local deployment is 32B (32 billion), while the cloud boasts a staggering 671B (671 billion). For ordinary users, leveraging cloud computing power is the only way to experience complete AI services.

This naturally led to the idea: why not move the entire phone to the cloud? Let the cloud run not just AI, but the entire phone system. The rapid development of 5G-A technology makes cloud phones feasible, providing low-latency, high-performance experiences even in wireless mode, thus fundamentally changing the application ecosystem.

China Unicom's booth showcased a typical application scenario of a cloud phone: it can run various large models like DeepSeek and HunYuan, easily switching between them as neededan advantage unattainable by traditional phones. Furthermore, a deeply customized AI system assistant allows users to execute operations with a single command, eliminating the need to consider hardware and software differences across various phones, thereby achieving highly universal AI services.

Universality is the cloud phone's greatest strength. Built on 5G-A technology, the physical terminal is essentially just a display and network device, eliminating the need for high-performance processors and large-capacity storage hardware. This allows everyone to own a phone providing a complete AI experience without the high cost of purchasing a high-end device.

Lei Technology believes that while AI development relies on improvements in hardware performance, future AI will inevitably be applied to more universal scenarios. The tipping point for AI isn't "high performance," but "high accessibility." Further upgrades to 5G technology are sufficient to allow us to place most services in the cloud and obtain the same experience anywhere. Therefore, the complete AI service provided by cloud phones will be a crucial step in the popularization of AI.

Cloud phones have also become another pivotal point for carriers to enter the AI field. In the China Mobile exhibition hall, Lei Technology also saw introductions to cloud phone services, learning that China Mobile had already partnered with Huawei to launch related services as early as 2024. China Mobile's cloud phones are very similar in positioning to China Unicom's, both relying on cloud computing power and 5G-A technology to provide users with a real-time and efficient AI phone experience.

High-performance traditional smartphones remain irreplaceable, but cloud phones in the AI era undoubtedly offer another option, making more people realize that AI is not actually "expensive."

Manufacturers Showing Their Prowess: The "Complete" AI Phone is Imminent?

While cloud phones offer great promise, widespread adoption still faces challenges, with 5G-A network coverage being a key issue. In contrast, high-performance terminals have higher penetration rates. To provide unique AI experiences, terminal manufacturers are actively developing product functions and building proprietary AI ecosystems.

Samsung showcased dozens of S25 series phones, displaying its latest AI achievements. A key feature allows users to summon AI from any connected device and receive results on another. Through a hybrid AI system using local and cloud power, Samsung's AI can understand work plans and act as a personal assistant; it can also adapt and recommend phone interface layouts based on user habits, making the phone increasingly user-friendly.

Samsung also showcased a series of ecosystem products, including smartwatches, smart rings, and smart XR devices. The XR devices, a significant collaboration with Google, wirelessly connect to phones, enhancing the interaction between phones and XR devices within the AI ecosystem. Samsung envisions utilizing phone computing power and the virtual reality experience of XR devices to achieve real-time scene analysis and high-precision gesture recognition, enabling users to interact with virtual content without controllers, for a more realistic digital world experience.

In a way, Samsung's Project Moohan is similar to Apple's Vision Pro, but the difference lies in Vision Pro being an independent high-performance hardware device, while Project Moohan is a distributed product whose performance only meets basic needs, relying on the phone to supplement its computing power. This approach allows Samsung to provide rich AI virtual reality experiences at a lower cost.

Honor's AI phones also attracted many foreign visitors, reflecting the desire for truly useful AI assistants. While space constraints prevented demonstrations of AI tasks like "ordering coffee," the ability to trigger a series of instructions with a single voice command was impressive. Lei Technology attended Honor's Alpha strategic launch conference on March 2nd and provided live coverage (searchable online), offering a deeper look into Honor's ambition in the AI field. Their investment of $10 billion over five years is a significant commitment in the smart terminal sector, driven by the accelerating pace of AI ecosystem adoption. Honor possesses several advantages in the existing AI ecosystem, justifying this substantial investment.

The product matrices of Samsung, Honor, Xiaomi, and others demonstrate the accelerating pace of AI adoption. Samsung has expanded AI to wearable devices like earphones and watches, even launching a dedicated AI ring to integrate AI operations into daily life. Xiaomi's Mijia ecosystem and human-vehicle-home AI ecosystem also put immense pressure on other phone manufacturers. Few companies can drive adoption across so many fields, a testament to the high costs involved. However, once successful, this locks users into the ecosystem, creating brand loyalty.

For Honor and other phone manufacturers, open collaboration is the best way to overcome these challenges and secure a favorable position in the future AI ecosystem. Google's Android AI ecosystem also adopts this approach, but its superior system ecosystem status prioritizes the adoption of the Google ecosystem rather than increasing system and product market share.

The level of competition among smartphone manufacturers is unmatched by other technology companies, even more so in the AI era. Every manufacturer seems to have its own strengths, whether it be new product forms or entirely new applications, leaving one wondering: when will a truly "complete" AI phone finally emerge?

AI-Assisted Fitness Gains Traction: Health Remains a Constant Need

Besides various AI phones, MWC25 showcased several companies' AI devices and algorithms for human motion capture. China Mobile developed a game called "AI Eight Brocades," where players stand before a camera, mimicking the movements of an on-screen instructor while the AI scores them based on posture and range of motion.

While essentially a gamified imitation exercise, it reflects a new trend AI fitness. Health has become a key focus for younger generations, making fitness a booming industry and sparking discussions about personal trainers. While effective, the high cost of personal trainers is prohibitive for many. AI presents a new alternative. With enough human body posture data, multimodal AI can judge user movements purely through vision, using visuals to guide corrections, essentially providing a portable personal fitness trainer.

At least four companies, including China Mobile, have introduced similar tools or games, with one startup already offering a phone-based solution. In essence, the phone acts as a "camera" and display, leveraging local or cloud computing power for real-time motion tracking. However, current solutions are limited to pre-recorded fitness programs, lacking true all-scenario adaptability.

This is reminiscent of the "fitness smart screens" popular during the pandemic, now a selling point for many TVs and smart devices. However, we may soon no longer need dedicated devices for AI fitnessa phone and a display may suffice.

Conclusion: The Future of AI Has Arrived

From AI in phones to practical applications, it's clear that the past (The conclusion seems incomplete in the original text.)