Thank you very much for waiting. We would now like to start Fujitsu Technology Strategy Briefing. Thank you very much for joining us despite your busy schedule. The speakers for today are as follows: Corporate Executive Officer, Corporate Vice President, CTO, CPO in charge of system platform of Fujitsu, Vivek Mahajan. Corporate Executive Officer, EVP, Head of Fujitsu Research, Seishi Okamoto. Corporate Executive Officer, EVP, CDXO, CIO of Fujitsu, Yuzuru Fukuda. First, Mahajan will talk about the technology strategy to support customer business growth, and then Okamoto will talk about the research strategy of Fujitsu. Last but not least, Fukuda will talk about the internal practices to accelerate the transformation of Fujitsu and our customers with some demonstration. After a total of about a 30-minute presentation, we'll have about half an hour for the Q&A, and at the end, we'll have a photo session.
We plan to finish the briefing at about 11:10 A.M. After that, you'll be able to look at the demonstrations. For the investors' analysts, once we're done with the Q&A around 11:00 A.M., we'll take you to the demonstration venues. The emcee is Kawahara from Public and Investor Relations. Now, I'd like to invite Mahajan-san to talk about Fujitsu's technology strategy to support customer business growth. Mahajan-san, over to you.
Good morning. I am Mahajan, serving as Fujitsu's CTO. I'd like to talk about Fujitsu's technology strategy to support the business growth. I'd like to give an update. Thank you for this opportunity. First of all, this chart. In September, we had a briefing for investors. We are focusing on three growth engines: Modernization , Uvance, as you should know well, and then the Consulting. These are the three growth drivers for us, the major pillars for our go-to-market.
There are five key technologies to support those. Nothing new here, but recently, for over a year, AI has become a boom. But we've actually started to work on this three years ago, and there's no change on that technology strategy focusing around AI. Now, to remind you the background of where we came to, where we are, and Tokita-san, our CEO, often says, "We want to be the DX partner, the biggest DX partner for customers, to support the customer's DX Computing to be able to process huge data, and then the Network and AI to get the Data and Security. Unless it's secure, nobody would like to use it," and integrating technology. So that's the strategy. So this chart is nothing new, but just to remind you that to you about our strategy. Now, here, now when it comes to enterprise, where do we find Fujitsu's values?
The first point is, for enterprise, we are to offer the AI solution. The issues faced by enterprise is that enterprise-specific data, the business processes, there are lots of data to be processed, and for customers, for the enterprise companies, these include the very sensitive information, and it could be the entire customer's IP and AI to be able to accommodate that, and the second point is that security will be very critical for that purpose: confidentiality, and then the third point, there's customer-specific data or the business processes, then customization, or 80% can remain as is, but then after that, there needs to be a prompt engine or the customization, and that should be done easily, so these are the three major issues. Now, what we offer is a general AI, generative AI framework, and then to support that, as you can see at the bottom, knowledge graph framework.
Knowledge graph is very important for generative AI, as you know. LLM will not be able to use the data and from LLM to be able to recreate the graphs, so graph AI, we have a very high percentage of ratio, and in graph AI, we have many IPs, and we're making announcements at different conferences, and generative AI, of course, using other companies' AI as well, but then generative AI plus different technologies to be converged, for instance, optimization technology, so combining those technologies, we have to accommodate the customers. Another point is AI trust. Of course, there are policies, but the policies need to be used in the AI technologies. How it can be done is very important, so these are the three major focuses for our AI framework. We have our IPs, but combining with other companies' third-party IPs to offer the customers.
So at the bottom, you see several technologies. For instance, root cause analysis. Another example is software engineering, humongous data, as I mentioned earlier, to support DX. That's our goal and mission to support DX. And the systems used by customers now need to be shifted to DX. And there have been very difficult challenges, but utilizing AI with software engineering, we can support this as one of the examples. Another point is to support AI, five key technologies that I introduced earlier. Let me elaborate further. First is computing. Without computing, AI would not grow, as you should know well. So at this point in time, NVIDIA and different computer engines are available, making growth. Now, in our part, we have a long history of computing, so we would like to leverage our strengths to support customers' AI. And that's one point. And there are three points.
First is processors. In Japan, we're the only CPU manufacturer. FUJITSU-MONAKA , 2 nm, first in the world. There will be a launch in them, and we'll be supporting this further, and we've made some press releases, so that's one thing, and this can be elaborated further in detail later. Another one is the servers, platform business with server, combining GPU and CPU to optimize that combination. PRIMERGY CDI and Quantum is another point. This week, I think you saw a lot of news, so in Quantum, I think we do have a top-level technology on a global scale. We believe we are among the global top three. Another point to remind you once again is that the computing and another point to grow AI is a network. That is a critical factor. When it comes to network, photonics and wireless, we are focusing on these both.
So this technology, as I mentioned earlier, in order to process data, computing, and to connect data, the network, these. So photonics with, for instance, Ericsson is doing wireless. And Nokia recently has acquired Infinera and wireless, and photonics, they're handling both. But we've been working on both for a long time already. And in North America and also in Europe, there's a heightened interest on this from customers. So we're going to announce the various technologies around that. And another point is security. Trusted Kunipro was announced. So trustable internet we would like to drive and lead together with different companies in Japan. And AI security framework, LLM and hallucination guardrail, various technologies will be used as well. Security framework will be very important for the customers, especially for the enterprise companies. And another point, and this will be introduced in the demonstration.
Combining the technology and the social digital twin to reduce CO2 emission, this is the technology that we'll be adopting. These technologies focusing around AI that will formulate our strategy. I hope that you will deepen understanding through the demonstration as well. I touched upon Monaka earlier, a very important CPU, ARM-based CPU. NPL is the only replacement, NPL or ourselves. This will be a high-performance, very advanced technology, so it will be incorporated here. The important point here is high-speed data processing. The platform is one thing, and energy efficiency. Energy efficiency is going to be extremely high. These are for data center customers. This will be a process that will be highly interested for them. Security, hardware-level confidentiality. Even if OS is hacked, the security can be protected in this process. Ease of use, because it's ARM software.
But five years ago, we might have been running ahead, but Arm globally is getting a lot of attention, and hyperscalers are offering some. So it's going to be a completely open architecture. This week, Google and others have made announcements on quantum computing. We have been investing in quantum computing very proactively since four years ago, and last year, 64 qubit, and in a few months, 2,056 qubits, and next year, 1,000 qubits. So of course, qubits is important, but as you are all aware, fidelity is the most important. Fault correction, error correction is most important. So there is a hardware, software, and application aspect of quantum, and we are developing the total package together with RIKEN and TU Delft. TU Delft is an expert in superconducting and diamond spin method. Another is the software aspect. This week, fidelity was announced.
STAR architecture, the unique computing architecture, has been released, and so this is a very important point, so we can reduce the necessary qubits with this architecture, so the physical qubit that was required for the customers can be reduced with this STAR architecture, so Fujitsu is not just a leader in Japan, but Fujitsu is a leader globally. This is our message to you, and of course, we are collaborating with other partners in terms of AI and computing. One example, we actually signed various partnerships this year. Not everything is on this slide, but one example is Cohere, so combining our core IP and Cohere's core IP to provide enterprise solutions. Takane is an LLM for Japanese customers, and so this is based on the partnership with Cohere. It's an enterprise large language model, LLM. Another partnership is Supermicro.
They believe in Fujitsu's Monaka potential and have signed a strategic partnership with Fujitsu. So they will be releasing Monaka-based Arm server, and we are going to be deepening our relationship with Supermicro going forward. And also, water cooling solutions are going to be very important. So this is another area where we'll be collaborating with Supermicro. And I talked about CPU. Last month, we did a press release with the AMD partnership, providing a total solution computing to support AI on various fronts. So lastly, there will be presentations from Okamoto-san and Fukuda-san, but a multi-vendor AI agent is going to be progressing going forward, especially in the enterprise sector. We are thinking of providing composable architecture so that various AI agents can be automatically coordinated to provide service to customers. Another is the architecture.
Fujitsu has a computing network, so vertical integration technology reducing the cost for customers and accelerating business transformation through our vertical integration technology. This is the kind of architecture, and this concludes my presentation. Thank you very much.
Thank you, Mahajan-san. Next, Okamoto-san will talk about Fujitsu's research strategy.
Good morning, everyone. I am Okamoto of Fujitsu. I would like to talk about Fujitsu's research strategy. At Fujitsu, to build a sustainable society, we are focusing on three materialities, to support this and to accelerate this, there are five key technologies. The research strategy of Fujitsu is focusing around AI to integrate and converge these technologies to create new values. That's the strategy. Today, I'd like to focus on AI for different areas to give you some progress update.
First of all, when it comes to AI, this year, on July 16th, the partnership with Cohere to develop Takane was announced, and on September 30th, Japanese-specific, or in the Japanese benchmark, Takane has become number one, and we've started to offer that. And business innovation at the company is focusing on generative AI, and at Takane, we are supporting them. And on December 13th, we announced a dialogue type using LLM, Generative AI Next Stage, AI agent itself to solve the solutions to further promote the business innovation. This was announced on October 23rd, Fujitsu Kozuchi AI agent was announced. And today, further advancement of technology since then is something I'd like to share with you. And beyond that, at Fujitsu, multi-AI agent, we've already started to work on this R&D.
So today, there will be a press release on multi-AI agent, which I would like to explain as well. So first of all, Fujitsu's AI agent, on October 23rd, we announced a dialogue type of generative AI, and on top of that, three technologies, AI in an environment to learn the environment on its own. So how would they memorize? It's going to be a very important technological aspect. Now, our agent, only the important contexts are memorized. And also, when it comes to learning, to proactively learn. So we have made much advancement on that. And then the behavior of control with the guardrails to control the AI behaviors. So these are the three technologies. So we are making a press release today, and there's going to be a demonstration. So this is a workplace operation support agent.
Based on the documents, such as safety rules and regulations, the video cameras for a long time will be autonomously or proactively learned by the AI agent, and without people telling, AI will make a proposal to make an improvement, or they can produce work reports or incident reports. For the context memory, tight accuracy analysis function to process the long-duration video content, so there is a long-duration video analytics benchmark, and we achieved the world's top record, and you can take a look at that later at the demonstration. Next is the multi-agent that we are focusing a lot now. Fujitsu actually, we've been doing research on multi-agent for three decades, so the 30-year-long multi-agent research and in a distributed environment, data and process to be interacted in a secure way.
And furthermore, what is included in the generative AI or the learning technology? We do have all those technologies. So what we only can do, this is what we'd like to achieve. The first thing is the co-creative learning. So among multiple agents, can learn either in a cooperative or the adversarial way, and then among the agents to provide the security or the reliability in their information exchange. That's the second one. And the third one is for AI to optimize the overall performance, AI workflow control. These are the three technologies. Today, today's press release is mainly focusing on co-creative learning, multi-AI agent. Application to security is what we are making announcement today. So in the cybersecurity, the agent who is attacking and agent who is defending, and then the video analysis test agent, there are three agents.
The attacking agent will actually initiate the attack, and then the defending AI will, of course, try to defend. They are to learn in an adversarial way. What will be the impact? There is a testing agent to that. These three agents are working, these three agents for a new threat that can be detected automatically, and the action can be taken. Please enjoy the demonstration on that later on. For multi-agent, cross-industry areas, why should we do to apply this technology? COCN. We made a proposal, and this is the innovation of engineering, which was proposed to Council on Competitiveness-Nippon to enhance the competitiveness globally and to enhance the resiliency and to enhance the social importance. What kind of issues shall be solved? What kind of proposals, the policy proposal recommendations, can be made?
So starting from here, I would like to touch upon the evolution in some of the technology areas, starting with data and security. The disinformation and fake information from GenAI is considered to be one of the largest risks globally that was announced at the World Economic Forum this year. So we have an all-Japan team to counteract against disinformation through the K Program. K Program is key and advanced technology R&D through cross-community collaboration program, and Fujitsu is a leader of this NEDO K Program. In terms of the benchmark for text, we have been able to mark the highest globally score. So in terms of text, we have the world's best technology already. And going forward, we will be expanding this to image and video and location information and such multimodal information. And this will be expanded not just to Japan, but also globally.
Next, in the area of computing, solving the power problem caused by GenAI. This is becoming a big social issue. In April this year, single GPU version was released, so having the number of GPU with the same AI computing performance through the AI computing broker. And we will have a multi-server version next January, so according to our analysis, we can halve the number of GPU and reduce the power consumption by up to 59% through this AI computing broker. As you can see on the right, we already have trials with customers and are ready for full-fledged deployment. Next is quantum computing. As Mahajan-san said earlier, quantum hardware, software, and application, we are doing R&D across all fields. In terms of hardware, 64-qubit was released last year, and by the end of this fiscal year, the next generation 256-qubit quantum computer will be released.
We are working with various customers and partners and STAR architecture, as mentioned by Mahajan-san. This architecture will advance the realization of practical quantum computation by up to 5-10 years. This year, with AIST, we have received a contract regarding this quantum computing system, and we have a lot of inquiries from Japan and globally, and in Kawasaki, we will be building the quantum computer center by September 2025. Lastly, I'd like to touch upon converging technologies. One major target is ocean, especially blue carbon ecosystem. It is said that blue carbon ecosystem is decreasing by 2-7% annually. Blue carbon can absorb 1.5 times the amount of CO2 compared to green carbon, so it's a very important part of our ecosystem, and we are creating the underwater drone that can be managed stably even under different ocean currents.
To be able to take precise measurements of seaweed, even if the water is murky. So this kind of a drone image is being developed. We are incorporating marine ecology to accurately measure CO2 absorption through seaweed. This technology now covers 50% of Japan's seaweed bed. We are working with BLUABLE , which is a company that spun out from Fujitsu in protecting biodiversity and promoting blue carbon. The demo that will be shown today are shown on the slide. There are also two press releases. Four are going to be new announcements today, and two are powerful updates to what we have already announced prior to this event. This music that you are listening to, this is through the collaboration between Fujitsu and Amadeus Code, which is a startup. Amadeus Code has a lot of training data, 40,000 songs.
There are no issues with copyright regarding this 40,000 songs of training data. We're working with professional creators. Inoue-san is president of Amadeus Code. He has worked on 55 works ranked number one in Oricon Chart by Arashi and KAT-TUN and so on. Then the Takiyama-san on the right is B'z tour guitarist, and Koenma-san is a hip-hop composer. We're working with such musicians to create new music. With that, I'd like to hand over to Fukuda-san. Thank you very much.
Thank you. Next, Fukuda-san will talk about the internal practices to accelerate the transformation of Fujitsu and our customers.
Hello, everyone. I am Fukuda, serving as a CDXO and CIO responsible for Fujitsu's own digital transformation as well.
What was introduced so far, the new technologies, how those are developed and delivered in our own transformation is what I'd like to talk about. In introducing generative AI in Fujitsu, it was back in spring 2013 that we started to work on that. About 18 months have passed, and we are now working so that we can change ourselves. Nowadays, more than 30,000 employees are using GenAI on a day-to-day basis. About 130, excuse me, 170,000 times AI was used. That's more than 10 times compared to one year ago. A very primary GenAI leverage, I think, is taking root. Fujitsu Kozuchi's cause and effect discovery function is used to measure the efficiency of the business. Now, annually converted, 920,000 hours have been saved. Efficiency has improved that much.
In terms of the labor cost rate not reaching a three-digit OEM, but it brings about a significant impact. What is unique is, I think this is very specific to Japan, someone providing instruction to do top-down rather than that. With 356 organizations, more than 1,000 staff are in a task force way working, promoting GenAI. This is what we tried. This is the impact that we got. We created this application, so why don't we do this? Of course, the clear top-down strategy formulation is important, but at the same time, from the employee side, almost like a bottom-up so that the people get engaged with each other to work on this. I think this is the kind of activity. Also, one year ago, we started with the chat AI, and the API was made open one year ago.
And nowadays, a very simple is creating applications, embedding a verse. And what was introduced today, AI agent is what is now happening. Now, let me show you what we are leveraging internally. Please show the video. So this is Fujitsu Kozuchi, which is being used internally. There are various AI apps in Kozuchi, and one of such apps is the customer proposal agent. So let's say the sales organization has an appointment with a potential customer. And rather than just saying hello, in order to be able to discuss customers' issues and solutions, sales wants to create a material to be able to show the customer. So this is a prompt that I had prepared in advance. Since many of you in the audience are from the media, let's say you have an appointment with a customer in the news media industry.
Let's say that the customer is interested in data utilization. Analysis and incorporating such data into business use. This is the kind of presentation material that you want to prepare in advance to the appointment. I've asked the AI agent to prepare the document. With this prompt, on the backside, the AI is identifying the issue and making a proposal. That agent is one. Then another is an agent that knows Fujitsu's offerings and services very well. This agent is looking at the potential solutions that Fujitsu can propose. Then there is another agent that is putting together the presentation based on the other agent's information. Multi-agent proposal is what you see here. If you are in the media space, maybe you can tell whether this is a good proposal or not.
So let's say you want to put this into a document. Now, this is based on Fujitsu's format. So you ask the agent to create the proposal based on a PowerPoint format, and you get about 60%-70% level presentation. I was in sales, so I know what it takes. So you go to the customer's website, you identify their potential challenges, and then I think it took several hours to prepare a presentation considering Fujitsu's solutions and the customer's challenges. So this is what you have here. So improving engagement with the audience, improving revenue, making business process more efficient. These are some of the topics. And then as possible solutions from Fujitsu, these are the offerings. And then if you want to delve deeper, you can interact with the agent to enrich this document further. Currently, Fujitsu is still in the beginning phase of utilizing the multi-agent approach.
But going forward, there are various new technologies, as mentioned by Okamoto-san, and this multi-agent approach is going to evolve further. As CIO overlooking the entire IT of our company, we don't want to be fenced in by a certain mega vendor. And from the perspective of national economic security, there are some data that you don't want to upload to another company's server. I believe that AI data should be held and stored internally rather than depending on another company. So the multi-LLM, multi-agent of Kozuchi, allowing for collaboration with different players and having the autonomy as a company to control one's AI data gives a lot of benefits. And in terms of penetrating the AI agent that I just demoed in Fujitsu, for example, CRM data that this customer visited this seminar and left these comments or feedback.
If there's an agent that can analyze such data, then that agent may be able to propose what kind of topics to discuss with the customer going forward and offer what kind of solutions. Technologically, this is already possible, but whether this is acceptable to humans or to current organizations, that may be a bigger challenge. Humans need to change our mindset and behavior, and sometimes, as an organization, it is difficult to change from the status quo, so change management and consulting may be the key to further implementation of this technology. This concludes my presentation. Thank you very much.