This is Mahendra speaking, and I would like to start the earnings result summary for Q2. 30% growth in net sales, but cost actually grew by 29%, operating income was minus 54%.

At the bottom of the slide, we can see the numbers at constant currency basis, net sales growing at 3%, and ARR growing at 6%, which is an improvement. The focus point is decline in income.

What has happened? Eva Chen will be providing detailed explanation later on.

The biggest focus here is cloud-related expenses. Why has it increased so much?

Other than paying out a lot of dividends, investors have told me in the past many times that we should be investing into the future. Based on that, we have exactly done that, investing to the future.

It's not simple capex. Our investment directly affects the P&L.

This is one big factor behind this number. Eva Chen will give you more explanation later on.

ARR progression, this is an important point. In net sales, we are using ARR to explain the future of the company.

ARR for the enterprise customers, this is improving from 4% from the previous quarter to 6%. In the second half, we expect further improvement in this number.

ARR by region, denominated in US dollars and therefore no currency impact. Americas stands out.

It is struggling somewhat, but Europe and EMEA are growing quite well. Japan is growing only at 3%.

Kevin Simzer will be providing you more explanation. Considering the geopolitical situation, there are headwinds and tailwinds.

In North America, we have some headwinds. This is the cash flow.

There is no comment. This is connected to the operating income.

Number of employees, headcounts is basically flat. In the background, we are acquiring new types of skills, there are some turnovers.

In total, the headcount is basically flat. We believe the situation will be very similar to this in the second half as well.

This is cost, and I'm sure this is something that you're more interested in. Cloud-related expenses stands out.

As I said before, this is investment for the future. Selling and marketing, the purple portion has also increased quite a lot.

For Trend AI and Trend brand, we launched two new brands, we have related expenses and events-related expenses. Headcount cost is increasing, one because of FX, and the other reason is we have internal bonus payments, and we're looking at both operating income and ARR growth.

Operating income was negative, there is no bonus related to that, but ARR, well, you may think that 6% is not enough, but that actually achieved the internal target, some bonuses have been paid based on ARR. This Q2 highlights: highest-ever quarterly net sales, and also as I mentioned before, ARR growth continues to improve.

Moving on to the full-year guidance, at this level of minus income, what do we do? Net sales is on track, there's no change.

In terms of investment, we don't have any temporary items in the second half. Token cost will continue to be incurred in the second half as well.

We expect that at the same level. Non-operating, we don't expect anything major apart from what has already happened in the first half.

FX is still volatile, but our plan is 157, and the current rate is 160, not that different, we'll keep it the same as well. The outlook for the full year: same for net sales basically, but for operating income, and ordinary income, and net income, we have revised them downwards based on the expenses that I've just explained.

This is impacted by the investments that we are making, pushing down all these income lines. That's all from me.

If you require further explanation, please ask me questions during the Q&A. Thank you.

Thank you for coming to our Q2 financial announcement. I think in my 38 years of IT industry, this few quarters are the most dynamic, and the speed of change in the whole industry is the fastest.

As a AI-native cybersecurity company, our transformation is also speedily on, and we must. Let's first look at what is the major change happening in the whole IT industry right now.

I think the most important thing, of course, is AI. In the past 2 months, how many new frontier AI model has been released?

There's so many of them. Most interesting thing for the cybersecurity is every new frontier model, whenever they release, they talked about, "Oh, my god, our model, we cannot control it.

It was doing things out of our expectation. It was doing all this hackings into the other infrastructure."

Cybersecurity is the center of this problems that's happening. I think also hackers is already realizing that.

It's really happening that the AI-native ransomware already happened and is invading into consumer, organizations that is using AI agent. The speed of how this ransomware customize itself into the specific attack is just fascinating.

31 seconds. 31 second, it will change its way dynamically and attack the customers.

With this type of speed of the threat changing, Trend Micro need to invest in how we transition into a real native AI operation to keep up with the speed of the transformation. Actually, about 4 months ago, even before all this frontier model started, we already thinking about this cybersecurity problem in the AI era.

I describe it as like this. In the problem on the left-hand side, it's like if you have thousands, millions of thousands of question and prompt into a big model, then it's like all the different colors all mixed together.

I am a painter. I love painting.

When all the color mixed together, it become what color? Black.

It means that it's like all the prompt throw into a black hole. It will be very difficult, no wonder all this frontier model themselves are saying that, "Oh, our model was out of our expectation.

We don't know why it do this, how it do this, all this attack," because it's a black hole. Therefore, about 4 months ago, we, Trend Micro, already thinking about for security, for AI, the better architecture is actually smaller SLM, which is all tuned to expertise, different expertise, different customer segment to fit their own needs.

It's like the color is separated, and the controller is controlling that color. Each of those model will be collaborating with each other.

That's what we imagined 4 months ago. Also, this is actually the chart that when I talk to NVIDIA, I send to NVIDIA, I say, "We are going to invest in this type of new model, and we want to use your Nemotron to do all this customized solution for our cybersecurity needs."

Immediately, actually, Jensen Huang himself immediately reply and say, "Yes, that's the right direction. Let's go."

Starting at that time, we accelerated our investment on all the open source and how to tune and perform this type of new architecture. That is the architecture that we draw out, and we believe that three layer of AI security stack.

Layer 1 is all onboard on all the machine, all the server, or the AI appliances. Layer 2 is to protect the AI usage.

Those two need to rely on what we call edge AI model or local AI model, because all this happening in the AI local model. Then the layer 3 is the trust layer that would dynamically choose the right model to submit if there's more complicated problem to be solved.

Those are the three layer, and Trend Micro are investing in this type of architecture. As you can see, just last month when NVIDIA announced their open-weight and open-source project, Trend AI, which is our Enterprise Security division, is one of the inauguration partner in this whole initiative because we believe the future needs will needs both closed and open frontier AI models.

That is why Trend Micro are investing so much into the future AI infrastructure. If we think about this whole information advantage to decision advantage, before, IT industry is mainly focused on information.

Data is the king. Information is the king.

Whoever can get more information is the winner. Now, with AI, information is cheap.

It's very easy. You can gather all this information.

What make the difference is your decision advantage, because your decision is based on your own priority, your own environment, your own needs for the decision. The difference is we need to switch to don't worry more information to understand customer's problem better, understand customer's priority, and help them solve their cybersecurity, make their better cybersecurity decision.

That would be the cybersecurity company's moats. Understand customer more.

If you look at the security gap, we call it, if you look at the red line, that's the vulnerability. That's the information.

Nowadays, you can find whatever vulnerability very fast using the frontier model, and those are very cheap. Now, you can find, but it's piling up so much of that.

However, the customer's decision is depends on the yellow line, which is the remediation capability. Those vulnerability are there, but for your organization, do you need to patch it now?

Where do you need to patch it? How do you want to apply your patches?

Those are the real decision quality that will affect the cybersecurity. That's why Trend Micro believe that we need to invest more not in just collecting all those vulnerability, but understand the customer's environment.

Same time, our customer, we seeing that customer are realizing this, and they say, "We cannot just relying on frontier model." Look, when US government shut down August, everybody suddenly alert.

If my company's operation is relying on those big model, how am I going to operate my company? Therefore, more and more they are investing in sovereign AI.

They want to make sure that decision and the intelligence they can own it by themselves. Sovereign AI is the most important part too.

To enable sovereign AI, you must be sovereign cybersecurity. That's why Trend Micro is investing on AI-native operation.

That is our platform, our engineer, software development lifecycle process, and our business operation, all need to be AI-ready and is the AI edge of the speed and implementation. That's why Trend Micro are doing all this investment.

I say in actually in 2025, when we were talking about Road to 2028, we already indicated 2026 will be our transformation year, and we want to accelerate our AI solution. In the last 4 months, we actually accelerate our acceleration and invest more into the AI solution, because we believe that is the future and that is what Trend Micro need to do to become the native AI cybersecurity leader.

What do we do, and where do we spend all those money? I think this chart is the most telling.

You can see we categorized all this spending into three big category. First one: internal AI transformation.

In 2024 first half, internal AI transformation, our cost is only $0.2 million. This internal transformation means that we need to invest for our employee to train and use AI.

This the internal transformation. If you remember, Trend Micro used to do our AI contest.

In 2018, we already doing this internal AI contest, and every contest we spend about $2 million or $4 million to make sure we have all these engineers that flew in and learning about how to do AI. In the past 6 months, with all this new frontier model and with all this agentic AI, we accelerate those investment in the training.

Therefore, there's internal AI transformation, it grow 6,000 more percentage, which is $12 million US dollar. That is the preparation of the foundation.

It's not going to continue to spend that much. It's just initial boost up and make sure all the employee need to get this AI-ready training.

Those are $12.6 million spent in the first half of 2026, that compared with last year, of course, is a big, huge increase. The second part of the cost, which is the largest part: customer AI and cloud consumption.

I'd like to save that part at the last because this is the most important and the biggest portion. Next one is infrastructure expansion and transition cost.

What does that mean? Infrastructure expansion is that because customer are focusing on AI sovereignty, and therefore, they will require all our data need to be located in their country.

That means our Vision One, for instance, our platform for collecting customer data will need to be stored in the local country's zone. Even if it's in AWS in cloud, but there's certain zone, country zone or area zone that we need to expand.

Establish those zone is costly initially. We need to establish those sovereignty zones so that we can expand our business for those countries.

This is the infrastructure expansion, $6.5 million US dollar from zero. Last year, there's no such expansion cost, but this year that's where we spend.

It's once you set up, you set up. That is just one-time cost for setting it up.

The second one is transition. Transition means that we actually are rewriting some of our core module, such as our XDR data lake.

We are rewriting those code onto GPU code. Originally, CPU, GPU is much more efficient and faster.

However, during this quarter, or this half year, we need to run it in dual side, both in CPU and in GPU, so that we make sure the transition and transformation is all complete. That is the double cost of our data lake.

However, those double costs, once we make sure that it's already stable, it will go away. That's $7.2 million US dollar.

Now, the last, the biggest portion, that is customer's AI and cloud consumption, which means customer are using our AI-driven technology or platform such as Vision One. That's why whenever our customer using more Vision One, of course, our cloud consumption and AI consumption will be higher.

However, we carefully monitor this cost. We are not just blindly, and means that Vision One is not making money.

We are investing and cost-consciously monitoring it. The indicator we use to monitor it is we call it cost per SaaS money, which is the gray bar that you can see in there.

Per SaaS ARR money, how much do we need to spend on cloud or AI cost? In first half of 2024, that is $0.19 US dollar.

That's per $1 of ARR, we spend 19 cents of that onto the cloud and AI cost. In the first half of 2025, we actually already consumed, and we optimized it, it reduced down to $0.13 per SaaS ARR money.

We controlled the cost. This half year, it slightly increased to $0.14 per SaaS ARR.

Why? Because we recently launched a lot of new services, such as agentic AI, or virtual red team, or digital twin, or IT security.

All of this are new services that we initially need to set up the infrastructure ready, but our customer number is not that much yet. It's great momentum going to increase, but currently in the first half, because all of this new modules has less customer, and therefore, per customer's ARR cost to ARR money is not scaled out yet.

We believe in the future it will be. It will be scaled out, and then our cost per ARR money will come down.

These are some of the evidence that we are investing for the future, but not blindly. We carefully monitor what is the ROI of that.

For instance, our internal training, internal deployments of all this, we already see 2x employees using all this AI. We already get all our employee ready for the AI.

Most importantly, when they do this, we automated the customer's dashboard deliver in 50% less time. We can service our customer much faster.

AI simulation also deliver become days versus months. Simulation means that we simulate customer's environment, and very easy to get the tabletop cyber simulation for our customers.

These are all getting much better customer services and support for our customer. Also our R&D are deliver more and faster with the software development life cycle AI helpers, we already see two times more code pull requests.

Also, all of this actually, for instance, agentic SIEM, it require to develop on those different connector. Before AI, before this age, then it would take months, even year, to develop those connection.

That's why SIEM, Trend Micro never enter SIEM market. Now we enter it, and we improve all of this, not only shorten to days, and now every 3 hours, we can develop any new connectors, adapter within just 3 hours.

Threat hunting, which is most important for our customer, we shorten those from hours to seconds for customer. Issue reporting and resolution is bit 5 times faster than before.

All of this is we are providing much better, faster security outcome for our customers. That's that part.

Infrastructure expansion, as you can see, no matter all these countries, Japan, India, UAE, German, Australia is already we set up all of this new zone. They are focusing on, and we are already seeing the sign of all this sovereign AI security requirement coming.

We have deal that is processing now already, and there are new that we're just starting to set up in 2026 for Brazil, for South Africa, Indonesia, Canada, UK. These are the new area that we are setting.

This is we investing in sovereign AI infrastructure to unlock a future monetization. As you can see, our momentum is already here.

Vision One ARR, YoY growth, 49%. Continuing seeing customers getting on to use more of those module, and the net retention rate is 122%.

Not only that, all existing customer that have Vision One attached, go to 35%, and Trend ARR now already 45% come from Vision One's ARR. That means that our business and revenue stream is on the new AI-driven model, and it's much more healthy than those big big deal, multiple-year deal.

It's much more healthy than before. That's why we believe all of this investment is really paying off.

We already also see the pipeline, and if we continue for this investment, then our trajectory of the ARR growth will continue to grow. That's why we invest.

In conclusion, I want to say, yes, we are investing, but we are investing with management action in place to try for economic outcome. We are not just say, "We investing in AI, we will lay off people."

No. We investing because we believe that's what customer need, and it will grow our whole revenue stream.

That's why internal AI transformation, we believe we put in the already. They have full adoption, and now with governance and measurable productivity threshold, we can make sure that AI efficiency will be much better, and we have the operating leverage.

Customer AI and cloud consumption, we will improve our unit economy. Then infrastructure expansion, as I say, once we scale out, be set up, then is unlocking the future monetary.

Then transition cost, it will die down once we move on to the right computing platform. Over like GPU is much more efficient than before, the transition costs will die down after this quarter.

That's Trend Micro, what we are doing, transforming into a AI-native cybersecurity company. I believe Trend Micro have the best opportunity and will be the leader in AI-native cybersecurity field.

Thank you.

SMBC Nikko Securities, my name is Kikuchi. I have two questions.

First question about, I heard that the investment was going to increase, and this is bigger than we expected. I understand that there's going to be a return on this investment later on.

I understand the direction. Unit economics, the revenue, as it increases, the investment will be absorbed.

I understand that as well. When will this happen?

When will the net sales start to increase? ARR growth in the US is still weak.

My question is when can we see the net sales increase enough to absorb this investment? Should we expect that next year?

I would guess so, but is that the beginning of the year or the latter half of the year? When do you think net sales will start to increase, and also what will be the breakdown in terms of types of products and also region?

That's my first question. Maybe you can answer?

Sure, I can start and Mahendra or Eva can also jump in. The way we think about it is that a leading indicator for the net sales is going to be ARR.

Those two numbers are going to get closer and closer together over time. One of the things that we described in both Eva's presentation and mine and Mahendra's is that we're really fixated on growing Vision One.

Our Vision One platform is now pretty close to half of our ARR in total on Trend AI. That's really the focus, is to continue to grow Vision One in the Americas and in the US in particular.

That's why I highlighted the fact that the US is doing so well at growing Vision One, and that is the priority. Over time, in Q3, we see the ARR increasing, in Q4 we'll see the ARR increasing.

As the ARR increases, we will see that translate into some net sales improvement over time. If

I could add, Kikuchi-san, legacy SaaS is slowing us down. That is a type of cloud security that sold a lot several years ago, but it will start to disappear, and then Vision One sales will increase that will accelerate the pace of return on investment.

Thank you. What about the specific timing, though?

Are you talking about next year? According to the situation right now, I cannot really see the ARR growing that fast.

Do you think you will hopefully break even next year? When do you think it will happen?

I want to know the timing. Yes, please respond to that question first if that's okay.

Thank you.

Sorry. I think you already see that we're showing the momentum of the second half.

We expect that Q3, Q4, our revenue continue to grow, and our investment will be slowed down or say already starting to pay off starting the second half. How much is enough to absorb the whole or how much an improvement of the unit cost economy is very hard to say.

All I can tell you is that starting second half, we definitely will be seeing improvement on those unit cost economy already. Thank you.

My second question is also about cost. Customers use your Vision One and that increases the cost, and I understand that the cost will continue to increase in line with net sales growth.

There are also other costs that are increasing. I believe that those other costs are controllable.

If that is true, Kevin, you showed a 2028 target, and AI cost, controllable cost, I'm sure will be controlled by the company. Is that the correct assumption?

That's the question. 2028 is 2 years down the line, but it's going to come very quickly, I'm sure.

Do you still have the same target in terms of performance target? Is it actually possible to continue to reduce the AI costs that are controllable?

I would like to explain first, and I'm sure that Kevin has some additional comments.

As Eva's slide showed today, the customer uses something, then it pushes up the AI cost and cloud cost. For SaaS, $1, it's $0.14 or $0.13, which means that we are still making other profit from there.

The reason that income is declining is, for example, there's duplication and also internal education. Those costs can be flat or can be reduced over time so that can be effective.

We don't think the headcount cost will increase that much either. That's what I wanted to say.

Oh, one more thing. We do have a 3-year plan.

At the end of today's event, we will talk about the IR Day event. In the IR Day event, we will provide more explanation about our future outlook.

Right. Kevin, can I ask you about the target, 2028 target?

What is the current status of progress against the target in 2028? Kevin?

Like we've been talking about through Eva's presentation, Mahendra and mine, we don't see any change to the Road to 2028. That continues to be our North Star business model that we are moving towards.

There were two areas, our cost of goods sold and our sales and marketing expense, both did increase and we've talked about both of those in terms of we see that as temporary and that those will gradually become more aligned with what we expect from a Road to 2028. The only thing I would add in addition to what Mahendra said around the unit economics, I like that calculation a lot and one of the things I don't know if everyone picked up on it, but our Vision One platform grew 49% and our cloud cost grew substantially less.

While we will make improvements in our overall cloud utilization and we will get leverage from some of these investments, but the top-line growth of the platform is much greater than the costs increase, so.

Thank you very much. That's all from me.

Thank you. Thank you very much for question.

I would like to move on to the next person.

This is Sato from Jefferies. Can you hear me?

Yes, we are hearing you.

I have one question. This is just to check once again.

This fiscal year, ARR, 15%? 15% ARR, I believe that that is your full-year guidance.

Can I confirm that there's no change to that target?

Our full-year ARR target, what we are saying is that we would like to aim for double digit, but our expectation, we are just saying that we are planning to increase ARR, but we cannot say whether it's going to be 15% or not. So 15% is the number that Kevin has mentioned when we announced quarter 4 results.

Kevin? Yeah.

Formally we do not guide on ARR. We guide on our net sales and our net income.

That comes from Mahendra. I try to give you an idea of what our objectives are and from an internal standpoint and where the business is headed.

Yes, we are targeting to continue to increase as this ARR momentum continues to build. We see it over time that it will hit double digits.

That's what we are moving towards.

It has been improved to 5%, about 6% at the moment. Your second-half scenario is that there will be some gradual increase towards the end of the year.

There's no change to that thought process. I'm not sure whether it is going to be 10% or 15%, but can I understand that this growth is going to be U-shaped?

Yes, exactly. So 10%, 15%, those numbers we don't give you officially.

However, going forward, our idea is to aiming for the growth, double digit. Yes.

That's all from me. Thank you.

Thank you. Moving on to the next question.

JP Morgan,

I am Henderson. I have two questions.

The first question somewhat overlaps with the earlier question about the outlook. Revision, cost increase includes the factor related to cloud.

I also understand that there is transition cost, infrastructure expansion cost, and some other one-off costs for the second half as well. Can you please explain how much you have increased against original plan in the second half in terms of cost?

For next fiscal year and beyond, what is your outlook on cloud cost? Is it going to be higher than what you had expected in the original plan?

Can you please give us some information about the outlook of the expenses costs, including what will happen in next fiscal year?

For next fiscal year, please wait until we start talking about next year's expenses. The SaaS sales, if that increases, the cost also increases for cloud.

Eva's slide showed that, there's a relationship between the two. There is also duplication and also investment for education.

Those will decrease over time. Eva, do you have anything to add?

Yes, I think AI internal transformation cost, $12 million originally was not in the plan. Because we see the need, and therefore that the whole company starting to all use AI.

I want to indicate one thing that I'm very confident that that will go down, because Mahendra mentioned about we have this cash performance bonus scheme, which is linked directly to the company's net income. That is the incentive that all the employee, including major spending of the tokens, our engineers, they're all looking into that.

It's like a automatic cost control. Once they seeing that token cost is too much and then it will affect the cash performance bonus, they starting to control it.

We already seeing that at first, you probably want to use the easiest one, the coolest one, and once we realized that those token cost is too high, we already starting to see engineer have much smarter way which module, which type of questions should use what type of smaller model or bigger model, frontier model, when you only use it when you really need it. I'm very confident on this type of self cost control by our whole employee and whole organization.

Thank you very much for the explanation. I have another question about headcount.

Q1, well, against Q4, Q1, we saw an increase. Now we are seeing a decrease.

The company is investing a lot into the product, which is very good. What about sales capabilities, or expanding partnerships, or acquiring new customers?

Maybe you are not investing as much into people or headcounts. Should we be focusing on headcount to think about that?

Kevin, you want answer that?

I think one of the comments that Eva had made upfront was that we are making a number of people changes across many functions, including sales and marketing, in order to become much more AI-native and AI-literate and leveraging AI. That comment applied to all functions, including sales and marketing.

We are making investments in both Trend AI and Trend Life in sales and marketing specifically. There's investments from a branding perspective, there's investments from a vertical marketing perspective, there are investments from a channel.

We're really working very, very hard on reigniting the channel ecosystem. We are making sure that we have the right people in the right job.

There's a lot of go-to-market transformation going on right now. We feel very good at the changes that have been made and the changes that are planned.

Just make sure that we are confident, and therefore we will not give up any opportunity to invest in increasing our ARR and our revenue. As you can see in this quarter, Mahendra's number shows our increase in sales and marketing.

Not necessarily in people headcount, but we are increasing our spending or investment in sales and marketing because of our confidence in we can solve the right problem for our customer.

Thank you very much. Yes, I do trust in your product.

I am looking forward to accelerated ARR growth. Thank you for your answers.

Thank you very much. Move on to the next questioner.

Is there anyone who would like to ask question? Can you wait for a while?

Please go on.

My name is Hideaki from BofA Securities. There's one question.

2028 target, the net sales growth, CAGR 8% to 10%. Against this number, if with a constant currency, it's 3% growth for the first and second quarter.

Take a look at this table include you are showing with the changes in currencies. You mentioned that your target is on track, I don't understand your explanation.

Can you talk about that?

I would like to first give you an explanation about this year. No changes to the full-year forecast.

Excluding the currency changes, it's 60%, net sales growth itself is going to be growing in the second half. For the 3-year forecast, there might be some comment, but 8%, 10%, this year is a transformation year.

If we refresh it, then this growth will go up, the 3-year forecast is remains the same. Kevin?

I would just echo what Mahendra said, and maybe tie it back to a comment I made earlier and that is that we are very fixated on ARR, and we feel like as that ARR momentum continues to build, it will get closer and closer to that net sales number. That's why we're using that as our leading indicator: ARR momentum.

Thank you very much. Understood.

That's all from me.

Thank you. Any other questions?

Sumitomo Mitsui DS Asset, my name is Watanabe. Can you hear me?

Yes.

Sales cost, 12 billion for the full year. Specifically, where do you see the increase against the original?

For cloud, I know that you already had an increase in the first quarter. But 12 billion, which other specific items that have increased or are increasing?

The second-quarter sales cost up by 10 billion and by quarter, but for the first half, it's 3.5. For cloud, I understand the token usage was inefficient in the beginning, maybe used in areas where it was not supposed to be used, because you tried to apply this across the company.

Now, the profit level actually affects their bonus, and individuals are trying to optimize that. In that case, you're leaving it up to the employees, and we cannot really see how the management is getting involved in this kind of control.

Are you trying to control or have you tried to control the amount of token usage?

Second-half forecast, I'm sure that Habara can explain this further. It's not that simple.

They cannot just look at the bonus and the cost and optimize individually. Yes, the management is overseeing this as well to optimize this.

We knew that if the situation continued, we would have to make a downward revision, and we accepted that. We have decided to go ahead with it.

Habara-san, can you please talk about the second half? Maybe give us some more comments.

Sorry, I do want to comment on this. Yes, we do putting some control of all the token usage and all those.

A very important thing is that we believe we don't want to suffocate innovation by putting those too strict of the control, because we are competing with hackers, the bad guys. We need to make sure that our solution really solve the customer problem.

We do putting the control, but it's not like you trying to suffocate the engineers' innovation and say, "Oh, you cannot do anything." Please understand.

I think that is a very important management principle of our cost control.

I would like to give you some granularity to your question, Watanabe-san. First, 12 billion cost increase, 80% of the cost increase is cloud-related, as already explained.

In terms of breakdown between first half and second half, well, a lot of costs were incurred in the first half. I said majority of 12 billion is cloud-related.

We have already actual cost spend in the first half. In the second half, we will have smaller duplication, therefore the cost level should be lower.

Out of the 12 billion, majority is cloud-related. Compared to the actual of first half, second-half cost increase will be smaller.

We will not see accelerated growth of the cost in the second half. The factors for cost increase were stronger in the first half.

I hope that this answers your question.

Thank you.