Thank you for waiting. We would like to begin the SoftBank Corporation earnings results presentation for the six months ending September 30, 2025. We would like to introduce today's attendees. SoftBank Corporation President and CEO, Miyakawa. Board Director, Executive Vice President, and CFO, Fujihara. Today's presentation will be broadcast over the Internet. Now, President and CEO Miyakawa will give an overview of SoftBank consolidated financial results and business overview.
Miyakawa, thank you so much for your attending. Now, I'd like to present our financial results for the first half of fiscal year 2025. Revenue reached a record high of JPY 3,400.8 billion, representing an 8% year-on-year increase. Thanks to this strong momentum, we are now in a position to aim for JPY 7 trillion in full-year revenue. Here is the breakdown. All segments reported revenue growth, continuing the positive trend from Q1.
Our distribution and financial businesses, in particular, performed strongly. Operating income also reached a new high record of JPY 628.9 billion, up 7% from the previous year. With a progress rate of 63%, we are steadily on track toward our full-year goal of JPY 1 trillion. By segment, all segments delivered profit growth. In the first quarter, the consumer business saw a 2% decline in profit. Looking at both the first half and the second quarter on a standalone basis, profit increased. The financial business was more than doubled, driven by PayPay's strong performance. Net income also reached a record high of JPY 348.8 billion, up 8% year-on-year. The progress rate for net income is 65%, which also indicates solid progress. To summarize progress against our full-year forecast, revenue stands at 51%, operating income at 63%, and net income at 65%, indicating a strong first half.
As shown, each segment, as you can see, exceeded the 50% progress mark, with all segments contributing to our performance. Not just one of them, but everyone contributed to our performance. So far, we've focused on growing non-telecom areas such as DX, fintech, and AI. As a result, the share of non-telecom revenue, as you can see, expanded from 47%- 63% ever since I took my presidency. As our business has diversified, the total revenue increased by about JPY 1 trillion in the first half alone. From here onwards, let me explain performance by business segment. First, the consumer business. Revenue was JPY 1,475.7 billion, a 3% increase. Mobile revenue also increased by JPY 9.6 billion, continuing the positive trend. Operating income was JPY 330.9 billion, a 3% increase. In second quarter alone, it rose 7%. We had a very robust second quarter.
Smartphone subscribers expanded steadily and increased by 3%, with both SoftBank and Y Mobile brands growing steadily. The increase in short-term cancellation was inefficient from an acquisition efficiency perspective. Through bundled packages such as broadband, electricity, and credit cards, we intend to focus on customers who use our services for longer periods and proceed with the second half of the year. Now, I would like to explain enterprise business. Revenue rose to JPY 482 billion, an 8% increase. Solutions continued to perform well, contributing to 12% revenue growth. Operating income was JPY 104.1 billion, up 10%. Exceeding JPY 100 billion for the first time in the first half, a new record high. Now, the distribution business. Revenue reached a record high of JPY 505.8 billion, up 17%. Operating income also hit a record JPY 22.0 billion, up 36%. There are two factors behind. One is the growth in AI-related products.
The market for AI servers and IoT products equipped with AI continues to expand. We expect this trend to continue for the foreseeable future, so further growth is anticipated. The other factor is the steady growth in recurring revenue, which is one of SoftBank's legacy core business. So the segment has transitioned from a one-time software wholesale model to a subscription-based business model. The distribution business is now on track to surpass JPY 1 trillion in revenue this fiscal year. Although we expected the enterprise business to reach that milestone first, distribution may actually overtake it. Both businesses are growing through mutual competition and learning, creating a very exciting climate. In the enterprise market segment, due to their differing profitability profiles. So we disclose results for two segments, enterprise and distribution. Together, these two businesses are on the verge of surpassing JPY 2 trillion in revenue this fiscal year.
The consumer business generates around JPY 3 trillion in revenue. In our next midterm plan, we aim to grow the enterprise segment to a similar scale. Now, for the media and e-commerce business. Revenue was JPY 822.8 billion, a 4% increase, driven by the commerce domain. Operating income came in at JPY 167.5 billion, a record high. Net income increased by 13%.
Next, finance business. Revenue rose to JPY 189.7 billion, a 24% increase. Operating income was JPY 38.5 billion, showing steady progress. Turning to PayPay. GMV continues to grow strongly and reached JPY 9.2 trillion, up 25%. EBITDA was JPY 48.3 billion, more than doubling from the previous year. Now, let me explain our next-generation infrastructure initiatives. In our full-year results announcement in May this year, we explained the direction of our cloud business using this slide. Today, I'd like to share updates on those three key areas.
First, our cloud software. Last month, we announced a new Sovereign Cloud service in collaboration with Oracle. The service will be offered starting next April from our East and West Japan data centers connected via OnePort and SmartVPN. Next, our homegrown LLM, or Sarashina Mini. Starting late November or 28th of November, to be specific, we will begin offering APIs to corporate customers to accelerate GenAI adoption within enterprise systems. Third, our AI compute infrastructure. Until now, this was mainly used internally for training Sarashina models, but as of October, we began offering it externally. In addition, to support AI startups. We launched a GPU platform program free of charge. The plan consists of three support tiers tailored to different phases of growth. First, startups building or testing prototypes receive up to 60 days of free access.
Second, those entering the validation or proof of concept phase are offered discounted pricing and development collaboration. Third, companies moving toward commercialization receive financial support as well as introductions to potential customers and sales channels. Last but not least, let me update you on Crystal Intelligence, which we made a press release earlier today about. SB OpenAI Japan is officially launched today, and it will offer exclusively Crystal Intelligence to corporate clients in Japan. As explained before, SoftBank and SBG established C Holdings, and together with OpenAI, formed SB OpenAI Japan as a 50-50 joint venture. We are working toward a service launch next year, and development is progressing smoothly. This is one of the seeds of future growth that we are excited about in the next fiscal year and beyond. In closing, let me summarize today's presentation.
First, in the first half, we achieved record highs in revenue, operating income, and net income, with all segments seeing year-on-year growth. Second, we made excellent progress against our full-year targets across all three indicators: revenue, operating income, and net income. The finance segment more than doubled, and we shared progress on strategic growth initiatives for our next midterm plan, including the Sovereign Cloud, Sarashina, and Crystal Intelligence. That concludes my presentation. Thank you very much for your attention.
We will now open the floor up to a Q&A session. Questions will be taken first from the audience on the floor, then from the Zoom press, followed by Zoom analysts and institutional investors. If you wish to ask a question via Zoom, please access Zoom via the previously announced procedure and press the raise hands button and wait for your nomination.
Once you have accessed Zoom, please turn off the live webcast to prevent howling. We would like to receive questions from as many people as possible, so please limit your questions to two per person. Once you have been nominated, please state your company name and your name. Please raise your hand.
[Foreign language]
My name is Mariah from Diamond Company. I have two questions. One, so the established today, the joint venture between SB and OpenAI, SB OpenAI Japan. I would like to hear the target and goal of this joint venture. I understand that first, this service will be used internally of your company and then to the external sales. I would like to hear what's your plan onwards and also the offering to other enterprises.
You said in next year, exactly when next year, and how many companies you are targeting and what is the sales target. It's related to this topic, but the AI agent services for enterprise, so entity data like a vendor company like that. So the ChatGPT enterprise to reseller agents or NEC or Hitachi, those IT vendors. They are in establishing the environment for AI agents. How you are going to differentiate from them?
[Foreign language]
Our target onwards. The joint venture was established today, and the service in will be next year. Since we first announced in this spring, Crystal Intelligence, seeds of Crystal Intelligence was created by OpenAI. The alpha version we received, and we just started internal verification. From our engineers, the feedback from them, this is something in a totally different world. Comparing to ChatGPT or enterprise, it's unlikely. New. I only glanced, and I haven't used yet.
However, what I was expecting to do has been in shape. Of course, we need to do the fine-tuning onwards. Once this is ready, then the way to work and also the speed of manufacturing products and everything will be completely different. With high anticipation, we are now internally coming up with and working on the details. What we are doing is that we are putting together some systems and also we also feel a risk to directly connect to the company's system. We are now internally working on establishing the certain environment to be able to connect to the system. Once we are ready for verification of our test bed, then we would like to start promoting it to other companies in Japan. As you mentioned, differentiation from the others, but this product itself is completely on a different concept.
Once the similar ones come up, then I also believe that the others might follow once our product comes in the market. The cost competition will be something that we need to look at. I hope I've answered your first question.
For enterprise, what is your target?
We can't tell right now, but within our company, the number of resources who is working on the data. Initially, certain efforts required at the beginning. Tens of hundreds of companies cannot be our target from the beginning. We would like to target one by one thoroughly. That's how I imagine for now.
My next question is about computing infrastructure. Recently, NVIDIA announced that they also participated in an APAC kind of conference in South Korea and announced that they would be providing 260,000 units, and that will be the third largest unit to provide. SoftBank is the most advanced, utilizing GPU of 10,000 units GPU. Comparing the volume with 10,000 and 260,000 that NVIDIA has decided to offer, it's a quite big gap. Considering your initiative, and I believe that Japan is way behind, which you also had a concern on. The government support for GPU implementation, I would like to hear your opinion on that. Also, do you have any plans. To implement more than 10,000 GPUs for South Korea, 260,000 units?
I heard that it will be distributed to some companies in South Korea. Rather than operating individually, by clustering, it would be more efficient. 10,000. The 4,000 GPUs is clusterized. Compared to the large data centers, of course, still in Japan, it's much weaker, less advantageous. That's why we are also working on Sakai data center and Tomakomai data center.
We would like to start the operation of Sakai data center first. With that, considering GPU volume, for Blackwell, 150. We would like to implement Rubin, the next version. There'll be like 30,000 or 40,000 would be the maximum capacity for that scale of data center. With Blackwell, 150,000. I think the computing performance capacity will be the same. Now, this scale is being combined and accumulated to make a bigger. What's coming next is that AI infrastructure scale, and as we call it, next generation social infrastructure, which we started five years ago or so. The key is that computing power would be the power of each nation. I also believe that the government has made a reasonable comment on that. Now South Korea, the 260,000 GPUs, which is quite aggressive. We are thinking about whether South Korea or Japan would be the center of this AI business.
Computing power is the key. I would like to bring the schedule forward to try and see if I can contribute to our new Prime Minister, Taka Ichi. If I could provide some input, I'll be grateful.
What about the government aid? Do you have any opinions on how government should be supporting?
I heard South Korea is providing JPY 1 trillion or so of subsidies. I would say that Japan should provide double of that. However, we need to keep balance in what the government has to do, and also what the private companies like these corporates have to do or can do. We need to keep balance and to establish decent infrastructure.
Thank you.
A ny other question?
[Foreign language]
Suzuki Freelance Journalist. About PayPay's IPO. U.S. government institutes are not working at the moment. I wonder if you have any comment on that.
The process of PayPay IPO, now that U.S. government institutions are not operating, and especially SX reviewing process is being on hold, we can't say anything unless or until the process restarts. Any other question?
[Foreign language]
細田 from Nikkei. First question is about consumer mobile. In the second quarter, I wonder how much confidence you have for the whole second quarter. Document mentioned that computer environment is very harsh and that they are looking at reducing operating income or margin. How do you analyze the market in terms of number portability or MNP? As far as SoftBank goes, we are performing good. However, we saw. More short-term churn users than we thought. We need to come up with an idea to acquire customers more efficiently. That is the lesson learned. In general, we are in good progress. Thank you. Next question is about enterprise business and solution business.
You mentioned before that those two business, you want them to exceed consumer in the next midterm plan. I understand you do not disclose respectively, but what is your confidence level? Consumer, people often ask me whether SoftBank is ready for raising prices. Unless ARPU goes up, we do not expect high liquidity of users. We are looking at a small net increase or net adds, but we do not expect 4 million 5 million users switching from one operator to the other like we saw before. So 1 million net adds times ARPU would be a growth of consumer business. We want to make sure that we still keep a slight increasing trend. We want to utilize AI agents to make our business more efficient. If consumer business grows gradually on the upward trend, enterprise business and solution businesses can follow the trend.
I think at some point, enterprise and solution business will catch up with consumer business in terms of volume. 3 trillion would be the substantial size, and then they would get more momentum in the market. 3 trillion for consumer, 3 trillion for enterprise, that is something that we want to look at in the next midterm plan. Our employees are watching these earning results. I want to sort of vote our employees to stimulate them.
[Foreign language]
MCA from Bloomberg. I have two questions. First is about Crystal Intelligence. You just mentioned earlier that it is not comparable to ChatGPT in terms of speed. Can you give us some examples of what is so different? As for the concept, it is closer to agentic AI. Would you please give us some examples? I would like to give you some examples, but I cannot answer.
OpenAI told us not to disclose, so I cannot give you details. However, just give you an image. You do not have to manually input. You can just speak to it. Because this is for enterprise, when you do some tasks, you need to establish a workflow. The workflow can be generated automatically. That is the future it has. The ChatGPT or enterprise features are just like rather Q&A features, but it exceeds that level of feature. Once the product is ready, we will hold an orientation session. I would like to refrain from answering by ourselves. The second is about PayPay. You mentioned that right now SEC's assessment is in suspension. Once it resumes, the valuation, I have heard it is about JPY 2 trillion,JPY 3 trillion. What is your expected valuation of PayPay?
I do not want to really mention the actual figure, but I would like to expect a higher. Why? Because looking at the speed of. The business growth, I believe that it will remain. Even I feel that it's too early to go to the public. With the power of business growth of PayPay and the profitability, looking at those two factors together. Also, as for the potential, the bank remains, so I see some future. In Japan, it's a lower country risk. We do have some unique features to look at. I also expect that the higher valuation, but it will be valued when it goes to the public, or it might take a little more time. In general, overall, I believe that PayPay is the company which has the highly valued. More and more means it's more than JPY 2 trillion and JPY 3 trillion.
When I say more, a little more is not too big, but I believe that this company is worth being assessed.
Next question, please. From Yomiuri Newspaper. I have a question about consumer business. Maybe somebody asked before. Mobile business was good in the first half. What do you think is the reason exactly, Miyakawa-san? As of end of September, mobile increased price. What's the reaction in the market since then? What kind of impact that could have on the first-half result or bottom line? Related to that, Rakuten Mobile said they would not increase price. What's your view on Rakuten Mobile from market competition perspective?
For the first half, yes, we are good. The question is, we are better than what? The old segments are growing steadily, and most of our businesses show two-digit growth. To be very honest with you, I wanted them to go more.
In the first half, we achieved net adds and increased revenue. In terms of on track to the plan, yes, we were on track, and we achieved a good performance. Whether or not increasing price contributed from a long-term perspective, since in general, prices are going up, at some point, we may need to consider our pricing. I don't know whether it's tomorrow or one year ahead, but at some point, we would take an action. We are still studying. I don't know how long we can say that we are only the one in the market that is progressing well. About Y Mobile price increase, I was impressed. They did a great job. Reaction from users is pretty good, and we don't see any huge difference in trend. While mobile continues performing well, it's difficult to answer your last question, however, which is Rakuten Mobile.
Rakuten Mobile said they would not increase price because their structure was different from companies like SoftBank. What costs to carriers? In urban areas, you can make money easily, to be honest, because of huge population and users are concentrated. People are using mobiles a lot. So long as you offer services, making sure that connectivity is maintained, then. Urban areas are very good markets in terms of making profits and making business. Especially the last 5%. Geographical area of Japan, when I was CTO, we had a huge challenge because we may need to build a huge high tower, which costs a lot. Optical fiber cables' length would be very long, which costs a lot. The cell site that you build would be used only less than 1% of urban cell site in terms of usage. We need to continue fine-tuning even after we built our cell site.
It costs a lot to build a network in rural areas in Japan. If they would do the same, and if they could say the same thing, I would say that's unfair. I think before, spectrum was allocated across Japan, and that was the rule. We did go outside and built our cell site. Again, the last 5% 10% of area coverage is very, very challenging. If Rakuten would do that, and they would stand on the same stage with us, and if they could say the same thing, I would say it's reasonable, but it's still unfair now because they were given the spectrum without competition. I don't want to say more because otherwise I would start getting excited.
[Foreign language]
My name is Chaima from Toyo Kezai. Two questions. One about GPU.
[Foreign language]
Not only your internal use for SEL, but external sales to start.
The GPU resource, what is the demand you have from other companies externally, and what kind of resource you are expecting? You also talked about South Korea and GPU to be procured more. When you look at the current resource and availability and also the demand by the other companies, if you could share, it would be great. The second one is about SEL, 70 billion parameters and NEC and also other smaller parameters. This is SEL, is closer to the others also. What is the differentiation from the others?
To your first question about GPU supply and demand, there are large scale of business demands and was not suitable for what we could offer so far. From the major companies, a big volume of allocation was demanded. We do also have different purposes for use, and we cannot just provide it to one company.
Therefore, we prioritize the use internally. Now we are to start external sales.
[Foreign language]
Some Japanese companies create their own AI and also to create services on outside AI, and also for their own systems, some companies want to implement AI. We would like to also start recovering the cost that's been spent. We are also thinking in a fair way that we are also reusing one for the next version of SEL, so we are going to start leasing more and more, which we do not expect a big return. If we could cover the cost, it would be good enough. The current phase is seeding, and we've just made the field so far, and now we are ready for seeding. Moving forward, when you look at globally, the supply is in short. GPU changes day by day, so it's not worth holding the older version.
Of course, modification is required, but need to focus on newer version as well. To your second question about SEL, our competitors, is the differentiation. With OpenAI model, we are looking at it every day. It requires a lot of time and money, and we believe that technology is not inferior at all, but time and the efforts are the key. A lot of engineers are needed for that, and there is an area which requires some academic level, then we would like to utilize OpenAI things. The call center level, we do not require that high level of engineers. When we also look at the cost, we can also create the decent cost structure. Differentiation is the key, and we would like to also catch up at the end.
The data for learning may be in short in the future, so there will be an opportunity for any AI to catch up. We need to keep moving so that we do not have to give in or give up. We need to stay on the race track so that we do not lose an opportunity. Thank you
[Foreign language]
Yamamoto from Nikkei. I have two questions. First, about mobile business. You are getting customers, but ARPU looked on a declining trend since last year. In the second half, you may want to focus on high-value customers, but in order for you to stay on the upward trend, what kind of initiatives you may have? Thank you.
ARPU looks decreasing, which is true. Consumer business positive, enterprise negative, and all in all negative.
I'm not going to go into details in numbers, but consumers are looking at the steady growth of ARPU. In general, I'm not too much worried about, and enterprise business, which see a very competitive market. Our solution business is growing, so you may get an idea of our strategy if you look at our solution business. We offer our bundled products, solution, and mobiles. Strategy-wise, ours is the right strategy. The second half, I think that you mentioned that you want to focus on high-value customers. What's the reason behind? What's your intention behind? By that, what I meant was to address a worsening churn. Customers who churn in a very, very short term, or those users are hopping around carriers. Users contract, they realize later the contract did not meet their expectations, so they may leave us very quickly.
We want to keep or retain users as long as possible. That's why we want to offer different products like PayPay or electricity. If users can enjoy other products, not just mobile, then they would stay with us. That is something that we discussed internally pretty recently. Next question is about Crystal Intelligence. In the mid-term plan, consumer and enterprise business, you may want to have the same level. I wonder to what extent AI agent contributes to such a growth strategy. I have an impression that monetization may take time because it costs at the initial phase. Yes, joint venture, SB OpenAI Japan. I do not think it will expect losing money even in the first year because structurally, the joint venture can offer consulting services while getting a return from customers.
Resource-wise, we have about 100 employees now, but by mid next year, we want to increase headcount to 1,000, for example. The resource should meet demands. If product does not sell, then those employees could come back to SoftBank. About JPY 3 trillion I mentioned before does not include potential AI agent services. Even without AI agent, we want to hit JPY 3 trillion. Upside is AI agent or Crystal Intelligence. That is something that we can expect as an upside on top of JPY 3 trillion that we want to achieve. If upside is actually growing, what would be the profit structure of the enterprise business? For example, distribution business, even though they increase revenue by JPY 1 trillion, profitability-wise, it is small. If solution goes pretty well, it has high margin business, so not only revenue, but also profit should grow both.
We are calculating assuming that the profit margin stays the same, but if we restructure businesses in a better way, maybe we could expect better performance than we planned in the mid-term plan.
We would like to take the next question as the last one from the venue.
[Foreign language]
I am Ishikawa, freelancer. About network, at this timing, the expedior starting things. What is your new network strategy, and how do you receive the result of OpenSignal?
Right now, our network is a step behind KDDI when you look at the result from different sources, but we do not think that we are inferior to others, but network design is superior to them, I believe.
Other than communications, we are expanding to non-communications area, and when the population is declining and when we concentrate on telecommunications business alone, then we have to consider getting more from the users, and we need to add more products or services on top. New technology is the seed for a new product. At this timing, now AI emerged. That is why we are slowing down or we are shifting the priority from telecommunications to non-telecommunications area, but we are not going to slow down that. About the OpenSignal, that standard of the assessment does not make sense. That is the feedback I hear from the people on the field. Having said that, it is true that. There are some areas that we are still behind the others. We do not want to make excuses. Of course, it is not only that.
What is most important is the customer satisfaction indication, which is called SBF, and we are also calculating that. We are still seeking a higher level to achieve and gain customer satisfaction. We are not going to compromise, and I am also discussing with the one on the field. I would like to make this company so that everyone says that we are number one. Where do you think it is different from what the actual is? What I heard from the field is that some items that the staff feel the higher score is different from what they believe, which does not contribute to customer experience. It does not really refer to or reflect the actual customer experience. I am not criticizing the result, but to motivate the ones on the field and to achieve higher customer satisfaction. Some areas should be fairly assessed. Thank you.
Now, we would like to take questions from participants on Zoom. First, Kikuchi-san, SMBC Nikko Securities.
[Foreign language]
I have two questions. First, about Crystal Intelligence. I heard that OpenAI owns copyright. I wonder if it stays the same. 。 Or IP. 。 So royalty is not that big, and SoftBank's, OpenAI Gain is bigger. So royalty, OpenAI gets more, or SoftBank gets more SBC Group。 And SBC pays $3 billion a year, and you are the biggest operator. I think most of that will be paid by SoftBank. I had that impression. Miyakawa-san says that pay for use. You do not want to pay that much, or you do not intend to pay that much. The usage fee, how would that impact your bottom line? Thank you for your question. I said pay for use. That is our intention. In fact, we started paying just a little bit.
We just started paying just a little bit. For this fiscal year, we have already taken account of the fee that we pay. About IP, the original one was created by OpenAI, so OpenAI definitely holds the IP. When it comes to the joint venture, whatever is created in the joint venture and what is generated in the process of consulting, then the joint venture would own that IP. We are expecting major enterprise customers. Whatever enterprise customers would create with our capability, then the ownership would belong to that enterprise customer. It is case by case. The payment we have made so far was pretty small. Again, that is pay for use. That is the current status. Would you expect, or should we expect, that would remain the same? For example, would that have any impact on next year's financial results?
I would say we could expect positive, but negative capability. If outsourced costs JPY 1 billion a year, then. We don't have to spend that much if we replace such outsourced resources with AI or AI's product. In other words, we don't start anything that could cost more than we pay for existing operations. Again, we don't expect anything negative from that pay-for-use scheme. Second question is about mobile business. Yesterday, some carriers made an earnings results announcement, and they would say that, they would criticize others or use excuses for their poor performance. Miyakawa-san, you don't want to spend money on short-term churners. My question is, going forward, quality matters more than quantity. Instead of spending a lot of money to acquire consumer customers, you may want to focus more on enterprise or solutions.
Carriers don't have to compete with each other, saying that they are losing their customers to others or vice versa. Compared to last year or looking back six months ago, what's your position in terms of market competition? Sorry, it's a very vague question, though. Thank you. Carriers, if they feel that they were attacked by others, I think they indicated SoftBank. My people said the other way around. SoftBank is aggressive, and I think SoftBank is actually aggressive and may be considered as an attacker. If we put the brake, put a foot on the pedal, they might stop, and then we would stop. Again, if you stick too much on quantity, but you don't see any business growth, I would rather shift the focus to AI than just getting more numbers of users. Our royalty users would pay more than others.
That means we expect a higher ARPU, and they are very important users. We should spend money for those royal customers and customers who stay with us long. Our people understand that direction, and we are on the same page internally. Even though we see a small decline in numbers, we don't want to be too concerned.
Masan from Nomura Securities, please unmute and ask your questions, please.
[Foreign language]
Regarding the next mid-long-term business plan, I have two questions: how you can make profit, and as for Japan, what is the demand of AI? In January, Japan's entire GPU supply and demand balance has lost its balance due to the initiative by this company. When we look at this climate, when you look at the U.S., the super companies like in the U.S. making AI or data centers, and Japan, many companies are utilizing U.S. or foreign AIs.
When there is not enough demand, where is the demand for the data centers? What is your thought on the demand of the GPU demand in Japan? 6,000 GPUs are now operating Sarashina. That is possible. That is feasible. Right now, 460 billion parameter, that is the Sarashina's, which I would like to bring up to 1 trillion parameter. To do so, if we just utilize the current. Infrastructure, then it will take years. That is the disadvantage that we hold. We want to do, but we cannot do so because of the shortage of power and also the GPU supplies, in short. We, as a public company, need to also consider our earnings. Compared to GAFA, we haven't been able to make enough investment in this particular field.
Japan, as one of the economic powers, to remain in this position, then we need to face this and tackle this situation. What we think now and a year later will be very different. We need to have more GPUs. Considering that, we have presented our earnings results today. This is being monetized gradually and next fiscal year onwards. We have no plan to present our result that is dragging the result due to the data centers and GPUs.
[Foreign language]
Only SoftBank is taking initiative in this, but not other companies are taking the similar initiative, which I feel sad about or disappointing about. SoftBank is taking initiatives in AI and also coming up with the intelligence. Even just building a data center, the structure, it's not enough. As you mentioned, compared to two or three years, three years ago, GPUs were not sufficient right now.
What do you think about the business model of data centers? I also explained in the board meeting what I would like to do internally so that the cloud business. Right now, the cloud in Japan, even the try to catch up with the overseas cloud, it's too late to catch up. When we look at the sovereign, even though we call that sovereign, and this is only within the contract. That is how I explained to the Japanese government. We would like to also hold the technology verification capacity. When we look at GPU as a service and the GPU cloud, it is going to be made completely differently. When we start from scratch, we can accumulate knowledge. Once the time comes, then we would like to become a cloud company. That is how our commitment is.
We did not have any intention to just build a data center structure. It is not profitable. Because of our business scale, what we can do is that we can do this related business horizontally, comprehensively. That is why we started from zero. It has taken a little time, but we have nurtured engineers, and we have nurtured some resources. We would like to materialize this. It means that in Japan, domestically, your business model is this scale. You do not see any competitors right now. Of course, how much GAFA can do in the Japan market, but including Japanese companies, how do you see the competition? It depends where you look at. Of course, there is a competitor anywhere you go. AI on GPU, so I believe, therefore, the platform has to be versatile, a platform. I mean, the companies will be creating its own platform.
Telecommunications companies like us will be the one to establish the environment for them to be able to easily use or make a platform. We all have to compete and learn each other. There will be one like OpenAI, GPT, or Gemini. For us, we would like to make sure that we are the one to offer.
Thank you. We'd like to take the last question. Tokunaga-san from Daiwa Securities, please.
I have two questions. First, about the shareholder return. You want to return to shareholders maybe more than before because of the inflationary environment? I think that's good from a minor shareholder's perspective and your group perspective. Any update on your view about shareholder return?
We shared with you the first-half results, and usually second-half we could achieve better. Since first-half was good, are we going to quickly increase return to shareholders?
At the moment, maybe an easier decision to make is share buyback as opposed to more dividend. The midterm plan that will start next year would be even more important. We want to grow further in the next phase. So long as we achieve the target that we have envisioned, I think we an definitely be aggressive about shareholder return. I think it just depends on how we go in terms of our business performance against the target. We will make a decision accordingly. Current level of dividend plus something is, I think, something that we could afford. I would just say today that, please, expect more for next term. Next question is about retention of customers. Maybe stop excessive cashback to acquisition, or maybe you may want to preferentially treat long-time customers. You may want to combine a lot of initiatives to keep customers.
Are you looking at only acquisition plan, or you may also consider a price plan? We look at both. Short-term plan, and we may make an announcement of a new price plan or change price plan at some appropriate time. Thank you very much. That concludes Q&A. That was all for the earning result announcement for FY 2025 Q2. This presentation will be available on demand on our corporate website. Thank you very much for joining us. The earning results for FY 2025, second quarter, SoftBank. Thank you.