Toward an Era of Preventing Communication Failures: The AI Base Station Revolution Led by Rakuten Symphony
Toward an Era of Preventing Communication Failures: The AI Base Station Revolution Led by Rakuten Symphony
We watch videos on our smartphones, open maps, send messages, and make cashless payments. We use communication networks every day as if it were second nature. However, the telecommunications companies supporting these networks behind the scenes are now at a major turning point. Traffic continues to increase, equipment is becoming more complex, power costs are rising, and securing talent is no longer easy. The traditional approach of humans monitoring, humans judging, and humans responding to trouble is no longer keeping up. This is why real-time AI is attracting attention. One of the companies at the center of this real-time AI shift is Rakuten Symphony, a communication technology company born from the Rakuten Group. Leveraging the experience of building the cloud-native Rakuten Mobile communication network from scratch in Japan, it has begun providing AI operation technology to telecommunications companies around the world. Based on discussions held at MWC (one of the world's largest telecommunications industry exhibitions) in 2026, the conversation focused on how AI is beginning to be implemented in the telecommunications industry. It was a panel discussion where executives from Google, TELUS, Vodafone, and Rakuten Symphony gathered to face the question, "Is AI really useful in the field of telecommunications?" The answer was "Yes, it is useful," but they also honestly shared that many companies are not yet using it effectively. This is not just a future blueprint; it is a reality that has already begun.
Video Description: How can AI contribute to helping telecom operators run smarter, greener, and more autonomous networks? At an MWC26 panel discussion hosted by The Network Media Group (NMG), leaders from Vodafone, TELUS, Rakuten Symphony, and Google take the stage. They explore how real-time AI and closed-loop optimization are transforming RAN (Radio Access Network) operations. From performance improvements to power consumption reduction and the path toward end-to-end network autonomy, we bring you the latest discussions from the front lines.
Key Topics:
• Application of AI and machine learning to real-time RAN optimization
• Closed-loop automation for continuous network adjustment
• Energy-efficient network operations and sustainability goals
• Dynamic traffic management and capacity balancing
• Roadmap to end-to-end autonomous network operations
The Telecommunications Industry Has Begun to Graduate from the Era of "Fixing Problems After They Occur"
Until now, operations in the telecommunications industry have been passive in a sense. A failure occurs. Communication quality deteriorates. Congestion occurs. Then, the cause is investigated, and humans respond. This flow has been standard. However, now that networks have become massive and 5G and cloud migration are progressing, this method is no longer sufficient. It is too late to respond after a problem has occurred. The important change discussed on the panel is the shift to a world where AI constantly monitors the network, proactively detects signs of anomalies, and automatically corrects them if necessary. For example, the moment AI finds signs that could lead to a decline in communication quality, it cross-references data from multiple domains to identify potential causes and performs optimal adjustments. By the time humans gather in a conference room saying, "It's a bit slow, isn't it?" the response is already complete. Such an era is about to arrive. The panelists stated that these mechanisms are already running on actual networks. From "responding when a problem occurs" to "acting before a problem occurs." This shift is progressing quietly but surely in the telecommunications industry today.
AI That Reduces Base Station Electricity Costs Is a Very Realistic Story
What is particularly interesting about this discussion is that the role of AI has become a "solution to management issues" rather than a "convenience feature." For telecommunications companies, the power cost of base stations is a very heavy burden. As mentioned in the panel discussion, more than 80% of the power consumption of the entire communication network is used by base stations (equipment that transmits radio waves). In other words, if base stations can be made energy-efficient, the impact is significant. It is the area where the effects are most visible, directly linked to both cost reduction and energy saving. Real-time AI reduces the load on unnecessary equipment during times with few users, restores performance during times of congestion, and aims for overall optimization including cooling equipment and processing loads. Humans would need to monitor 24 hours a day, but AI works without complaining and without taking breaks. If there were an AI employee who decided at 3:00 AM, "Let's give the base stations in this area a rest," it would not be surprising if they were awarded for it.
The Real Difficulty of AI Implementation Is Not "Intelligence" but "Connectivity"
However, this panel discussion did not end with just praise for AI. This is important. It is a reality that there are many cases where expected results are not achieved even after implementation, and the reasons for this were discussed frankly. The cause cited was the problem of "AI that only looks at its own area of responsibility." A communication network is a world where many parts are complexly connected. Wireless, core networks, clouds, power equipment, etc., all influence each other. For example, suppose AI detects a problem of "weak radio waves" in a certain area and automatically increases the output of the base station. However, because of that judgment, the radio waves interfere with neighboring base stations, and the quality drops in the neighboring area instead. Such things happen normally in communication networks. In other words, it is a whack-a-mole state where fixing one place breaks another. AI that acts by looking only at its own area of responsibility will end up holding back the entire network, even if it has no malicious intent. What was discussed in the panel was, "Because many AIs are designed that way, they don't have as much effect as expected even if introduced." So, what should be done? The answer given was a "phased and cautious implementation approach." Test the AI in advance in a virtual space that mimics the actual network, and move to actual operation after confirming that there are no problems. It is also important to leave a mechanism for humans to check and intervene, rather than leaving everything to AI. It might be similar to the image of not handing the steering wheel to AI immediately, but letting it practice in a large parking lot first, and letting it run on the highway while humans watch over it. This was emphasized in the panel discussion not as an overly cautious stance, but as a design philosophy for continuing autonomy safely for a long time.
Telecommunications Companies Will Change from Line Companies to AI Operation Companies
As this trend progresses, the competitiveness of telecommunications companies will no longer be determined solely by rate plans or the number of base stations. How well can they grasp the situation in real-time? How autonomously can they protect quality? How stably can they operate while keeping power consumption down? How much can they shift human resources to advanced tasks? Rather than the network itself, the competition will be about how smartly they operate the network. Telecommunications companies will change from "companies that provide lines" to "companies that optimally operate social infrastructure with AI." This panel discussion showed that change quite clearly. As a leader in the discussion of this change, Rakuten Symphony's position is worthy of attention. The words "This is the next standard for the telecommunications industry" carry a persuasiveness unique to the company, which has accumulated unique knowledge in building Open RAN and cloud-native communication infrastructure.
It Actually Matters to Us Users Too
From the perspective of a general user, you might not really get it when people talk about base station energy-saving control or AI automation. But what happens as a result is the stabilization of communication quality, faster response to failures, maintenance of price competitiveness, and expansion of capital investment capacity. It is directly linked to the smartphone bills we pay every month and our daily connectivity. If AI can reduce the waste in electricity bills and failure costs for telecommunications companies, it will be easier to maintain price competitiveness accordingly. In the future, there is a possibility that it will have an indirect effect on our smartphone bills. Rather than flashy new rate plans, AI working in invisible places might actually be more effective for our lives in the long term. It doesn't show up on smartphone screens and is hard to turn into a commercial. But that kind of subtle and important evolution is what determines the future of communication.
Personal Opinion and Observations: The Day AI Becomes a Design Philosophy Rather Than a Tool
The Debate of "Whether to Use It or Not" Is Already a Thing of the Past
What impressed me first in this panel discussion was that the question of "Should AI be used in communication networks?" was not even on the agenda. The speakers' talk was already at the next stage of "how to design it correctly" and "why there are cases where it doesn't work as expected." From an era of questioning the effects of AI to an era of questioning the quality of AI. This is a shift in perception that is visiting not only the telecommunications industry but all industries. However, the fact that it is being verbalized and discussed so clearly is what the telecommunications industry is like right now.
Energy Saving Has Become the Most Important Management Issue, Not an "SDGs Topic"
The figure that base stations account for 80% of power consumption is just trivia if heard from the outside, but if heard by a telecommunications company's financial officer, it immediately leads to the story that "that's why the cost structure is heavy." Electricity bills occur every month, and costs accumulate as more equipment is added. Now that base station density has increased with 5G, this problem has become even more serious. Energy-saving control using real-time AI is a story that cuts directly into this structure. It is often talked about in the context of renewable energy utilization and carbon neutrality, but the essence is the sustainability of the business. If environmental consideration and improvement of management efficiency can be achieved with the same measure, the priority will naturally rise. I believe that the fact that energy-saving control using AI is an area where cost-effectiveness is easier to explain compared to other investments is also the background to why its introduction is progressing realistically.
Siloed Companies Will Have Siloed AI
The problem of "AI that only looks at its own area of responsibility" discussed in the panel is a story of technical design, but it is actually also a story of organizational theory. A structure where siloed departments each improve only their own metrics, creating inefficiency as a whole, is not rare in companies or government. Since AI is often created in a way that inherits the organization's design philosophy, the situation where "our company is siloed, so the AI became siloed" can be said to be a technically inevitable result. Conversely, an organization that can use AI introduction well has a structure that can integrate data and systems across the board. To create an AI that can overlook the entire network, a system that can overlook the whole must exist first. If AI is a mirror of an organization, the real difficulty of AI introduction may be the problem of organizational culture, not technology.
Phased Autonomy Is Not an Overly Cautious Stance, but the Only Correct Route
The implementation approach of "testing in a virtual environment before going into production" and "leaving a mechanism for humans to check and intervene" is obvious when you think about it. But there is meaning in properly verbalizing the obvious and sharing it with the entire industry. Amidst the extreme reactions of excessive expectations or excessive skepticism toward AI autonomy, there is value in clarifying the middle path of "proceeding in a phased and verifiable manner." Communication infrastructure, in particular, is an area where the cost of failure is high. It functions as social infrastructure, not just at the level where general users are troubled by "I can't use my smartphone," but for emergency calls, medical equipment communication, and logistics control. In places where high reliability is required, a design philosophy that positions AI not as "amazing technology" but as "part of social infrastructure that is carefully operated" is being questioned.
The Meaning of Rakuten Symphony Being at the Center of This Discussion
The fact that Rakuten Symphony sat at the same table as global players like Google and Vodafone in this panel discussion and led the discussion is worthy of attention once again. There was a time when the cloud-native, virtualized communication network that Rakuten Mobile built in Japan was viewed with skepticism by experts around the world, asking "Does it really work?" However, now, because it has that architecture, real-time AI integration is possible, and it is in a position to lead industry discussions. While many telecommunications companies around the world are being forced to modernize while holding onto legacy equipment, the advantage of Rakuten Symphony (Rakuten Mobile), which was built cloud-native from the beginning, is gradually taking effect in the context of AI utilization. Rakuten Symphony's name has come to be mentioned as an answer to the question of where the standard of the telecommunications industry is heading. This is no longer a coincidence.
The More Truly Excellent an AI Is, the Less Anyone Notices It
Finally, let's take a step back and think. At the stage where AI is talked about as "amazing technology," flashy features and dramatic results are inevitably emphasized. But in the context of autonomous operation of communication networks, the maturity of AI appears in the opposite direction. When AI is functioning really well, the proof is that "nothing is happening." Failures are prevented, quality is quietly maintained, and power is optimized. Without anyone noticing, problems are handled before they become problems. It doesn't show up in commercials, and it doesn't make the news. But I think that is proof that AI has changed from a "tool" to a "subject that makes decisions." An AI that keeps working quietly is actually far more difficult and far more valuable than a noisy AI. What this panel discussion showed is that the telecommunications industry has finally begun to aim for its true ideal state. The future of communication may be decided by AI working quietly in places no one can see, rather than flashy new rate plans. A revolution that doesn't show up on smartphone screens has already begun.
How Real-Time AI Is Driving Network Autonomy and Energy Efficiency in Telecom
https://symphony.rakuten.com/blog/how-real-time-ai-is-driving-network-autonomy-and-energy-efficiency-in-telecom
