Hi, good day, everyone. Thank you for joining us today for Phison second quarter earnings call.

My name is Astoria Chen and I am semi analyst from SinoPac UST. Today we are very honored to have K.S.

Pua, the CEO of Phison. In today's call, K.S.

will first provide us provide us some updates on the Phison second quarter results and business highlight and then market outlook. After that, we will open the floor for the questions.

As a reminder, if you have any questions during the call, you can write down the questions in the chat box, and we will raise the questions for you later. Now I would like to turn over to K.S.

for the comments. Thank you, K.S.

Thanks, Astoria. Good afternoon, good morning, ladies and gentlemen.

Thank you for joining Phison our C Q2 '26 earning call. That's why we start with our C Q2's performance.

Since C Q1, we defined a new category named AI Ecosystem Solutions, which including the enterprise main drive, boot drive, aiDAPTIV+, and AI networking, et cetera. This C Q2 has around 20- 38% of our total business, which 68% Q-o-Q growth.

We still gaining a lot of good demand from the enterprise and AI ecosystem business. The second larger is we named is Embedded.

Mainly is PC OEM, customized drive for the PC, gaming, security, and also the mobile use. This has a 32% of our C Q2's revenue, but we have a over 2x of the growth Q-o-Q.

The third largest is industrial modules, which has a 17%, 18%. Retail is down to the below 5%.

Controllers, we are C Q2 6% just because the revenue growth too much. By net amount of a controller, actually, we are increasing in the C Q2.

You may know the mobile phone, PC OEM shipment actually declining, means Phison we are gaining the unit share of a controller in the PC and also in the smartphones. Phison we are proud that we have the record high in the C Q2, not only in revenue, but also in our market share in mobile, PC OEM, and also group big growth in the AI ecosystem business.

Financial result, I think you just read from the our announcement, I don't want to spend time here. I need to highlight is a gross margin we growth Y-o-Y over 7.5x.

Which is a record high. This is a really show the good performance of our engineers that many years are hard investing in our new projects.

Revenue-wise is a 67.8 billion is a all-time high. The first half, we hit a 108 billion.

Again, this is Actually, Phison, I tried to make Phison 2025 to hit the 100 billion NT, we failed, but we make it in this first half. Gross margin, we hit a 65.3%, but this also we have a company policy.

We try to be more conservative, we have a write-off 1.1% is a inventory write-down. The the component we keep for by policy, didn't use for certain period, we write down and to make sure our inventory is cleaned.

First half average is a 63.8%. Earning per share by C Q2 is a 118, which we including the 14 from our the what's the so-called investing gain.

So 118 is also the all-time high, and the first half we are 187. In our income statement, by revenue-wise, you can see we have big big growth.

This part, I just want to share with you, in special in our R&D investment. Some shareholder may has a concern complaining about the investment R&D, but again, since many years ago, Phison insist, we need to focus in our R&D investment.

You can compare this C Q2 versus the last year C Q2 is around 6.5x, close to 7x of the investment. Why?

Because the AI coming, we need to keep customized a lot of a new design. By second half this year, we have a many new tape-out: controllers for the high-speed PC OEM, for enterprise, for the bridge, cus- some ASIC design, and also the mobile UFS 4, UFS 5.

We gain the market share, why? Because for the last 5 years, when market, no matter market good or bad, we keep heavy investment in R&D.

That's why we can show the good performance here. Again, in the second half this year, Phison will keep same momentum in R&D.

Make sure we heavy invest and we are able to create the new application and try to lead in this market, okay? By the balance sheet, I think, first of all, by cash flow, by C Q1, Phison end of C Q1, Phison we are tight in cash, but come to C Q2, we are now better.

The the AR is coming back, and we have a one CB $800 million by C Q2, which happened. Cash flow, now we are doing okay.

Inventory-wise, I just shared with you company policy: we try to keep equivalent of a exabyte, which we carry every month. I don't know if you believe or not, NAND flash is getting much difficult to ask for supply.

No matter you pay how much of price, the supply now is really a big issue. Our policy is to keep same amount of the exabyte equivalent every month to make sure our design projects we are able to ship.

Unfortunately, by end of this year and C Q- early next year, we got so many new inquiry from the CSP, near-cloud, OEM. We still try to work hard to get upside supply to fulfill what our customer ask.

Inventory-wise, if you really ask me, "Do we enough inventory?" I'm telling you, most likely we still short in the part some part by end of this year and the early next years.

We like to share some TIP and non-TIP different, but basically you can checking here is not much different. This is all about the Phison's our options.

It's by by our financial policy, we're not have a much a stock option to the employee, the different is quite small here just for your reference, okay? We like to show this page.

For many years, I think you guys also expert to know memory cycle, flash cycle, this and this and this. This is all predictable.

Why? Smartphone, you can forecast the unit, the the the content, PC.

From now on, by the way, smartphone, PC is declining by unit this year and next year. Our controller shipment increasing, means we win the major share in this industry.

I think this is about last 5 years Phison keep investing, okay? Never slow down, keep investing.

That's why the PC OEM, the NAND manufacturer, they trust to Phison. They're willing to work closely with the Phisons.

We believe 2 more years, consumer market, smartphone, PC will bounce back. Why?

Because the phone you use may end of life in certain period up to 2 years, demand will back. If Phison '26, '27, we are able to keep winning the unit share of controllers, means '28, '29, definitely, we can enjoy the good return at that period.

What we have to pay is heavy R&D investment. That's why we keep pushing investment into R&D.

Okay. Things keep different is the AI.

I think we spend too much time, I don't want to waste the time here. The AI inference, the output was a text, is a kilobyte.

Now user start to go to the image, is a megabyte. Some already have a video, the short video from China, from US, is a gigabyte.

Unfortunately, if you go to the AI agent, okay, if you start to use AI agent in your platform, no matter on-device PC, on-premise service server, or go to the cloud, you are creating terabyte of a content. I don't know how people saying that the flash can supply demand can be balanced by '27, '28.

I don't think so. Why?

Because AI agent is coming. This is defined by the Nvidia the five-level of the AI cake.

What is the Phison's role? I think you have to take look into this: what Phison's our role position in this industry?

We have a controllers. Controllers not only mobile, PC, but we have a lots of aiDAPTIV+, and also enterprise, and some customized controller NPU in the this layer two.

We also provide different kind of a modules: main drive, boot drive, okay, to the so-called the infrastructure. At the end, Phison we have aiDAPTIV+.

Now we transform to the Phison AI Data Platform is a AI applications. Uh 2 weeks ago, our board members also approved Phison start establish a cor- corporate VC.

The first investment is a $20 million, equivalent to the 600 million NT. We are going to start invest to a lot of startup as a ISV who are design their program on top of Phison aiDAPTIV+.

In this five layer, Phison play the role from chips, infra, and also new applications. We are transform- transforming ourselves.

We are now naming Phison we turn to 3.0. I don't want to spend time in today.

Next quarter earning, next quarter earning, we are going to show you what Phison doing in the Phison 3.0. This is to show you.

In the first generation, we are in the controllers, consumer modules. Then we move to the customized modules, enterprise, which we are doing now.

Next, what is our next? Next is mainly in the AI storage and the edge AI platform, which we are to show here show later, okay?

C Q2, this is to show you what our role in the enterprise. This is enterprise boot drive.

In the last few months, our press release, we name our boot drive is like 4- 40 times growth, 60 times growth. Why?

You can see here, AI networking, just only this quarter, we have a six new design-wins program. It's a big program.

In the no matter CSP, OEM, recently we just win the extra two OEM, and the AI storage system partner, the near-cloud, we have a two new design-in, design-win. Infra, we have a one new design-win.

AI chips ecosystem, we have a one new design-win. This is a boot drive.

Move to the main drive. We also cover a lot of design-win: CSP, OEM, AI storage system, infra, we have a two new.

Networking, we just win special one of a big program of an AI networking main drive, not boot drive, okay? In the AI ecosystem, we play the role main drive creating a lot of exabyte, boot drive create a lot of customization.

We are now the key player in this field, okay? Please recognize Phison is no more in consumer.

We are in the AI ecosystems. This is some our new design-win just to show here, okay?

We are in the 5G base station without EMC. We are moving to the medical.

We are going to the a lot of automation. Share something with you about COMPUTEX, what our exposure.

We have enterprise, PCIe Gen 5, already in MP, which create around 30+ percent of our revenue. AI, later I will show you what is our potential revenue in the future.

Gen 6, we are controller start sample back by January. We are going to show our Gen 6 enterprise SSD by November, okay?

This is going to move into the AI ecosystems. By the way, we have a pending order already, unfortunately or fortunately.

We have a pending order, not small amount, but now we have to make sure we can get the supply. Client, E37, is the best performance in the world.

We win the several big OEM project recently. Mobile, this is a 3.0.

Just I want to remind you, Phison is a really key player in the mobile storage controllers. This is we are show in the COMPUTEX our 245 terabyte.

This is a product's good to MP, but recently customer not able to buy, why? The price is a super expensive.

Now customer now move back to the 64 terabyte, actually which which is good to us. We can sell more unit, more controllers, more systems.

This is our Gen 6 controller. First silicon, we're going to MP directly without any ECO.

This again to to to to show that Phison engineers is really solid. How to make it solid?

Again, heavy R&D investment. This project, we are going to show to the November.

Again, I I had to keep repeating: we already have a pending order, but we still don't know if we can get enough flash to ship. This is a 64 terabyte demand.

This is our E37. We show it's a a lot of reviewers saying that this is the best, fastest, best power consumption in the world, okay?

We win several PC OEM and also the NAND company's OEM business. This is aiDAPTIV+.

We keep saying that for many years. I think you If you follow us, you know what about this, right?

We have a products go to the market. We are working a PoC with the MediaTek to use our UFS 4 controllers with the MediaTek to deploy the aiDAPTIV+ in the mobile phone.

PoC pass. PoC doing good.

We are able to let the NPU, which asking for 24 GB LP5, downsize to 12 GB, but still can execute the models to run to run inference, which using our UFS as a cache. This PoC done.

Hopefully, we like to see this go to the market by end of next years. This is also is also Phison inno- innovation.

COMPUTEX. Phison's co-promote the Intel platform with the Phison aiDAPTIV+.

You can check this from the YouTube. You can check what is I don't want to spend time here, but go to check this.

This small PC we name is a AI PC, AI-on-device, can run the 120 billion model with 32 GB of a DRAM. DRAM shortage, right?

We fully believe AI-on-device, means AI on PC, laptop, desktop, is going to happen by next year, but unfortunately DRAM is shortage. AiDAPTIV+ is the only solution.

We also show the mature products to the market with our branding partner: MSI, ASUS, Advantech, and ASRock. This is already go to market.

Not just only a concept, okay? This is our server partners, which enterprise SSD: Giga Computing, AIC, Pegatron, and WD.

Okay. I'd like to show you what is Phison achievement.

Many quarters, investor, fund manager ask: what is your revenue? What is the revenue?

Again, revenue is a relatively small. Why?

Because this is a totally new ecosystem. We need to have a lot of hardware, software implementation, and we need a lot of ISV.

That's why Okay, what is ISV? Independent software vendor.

That's why we establish our CVC, corporate VC, to start invest to the startup to help them to design their application into our aiDAPTIV+. Revenue-wise, again, I'm still cannot tell you the the the the the net amount, but but this is showcase.

Again, I want to repeat: this is not a PoC, okay? This already in a commercial.

We ship our AI server with aiDAPTIV+, with Phison SSD, with our proprietary service, the software platform, application, customized in Taiwan more than 10 banks, okay? Definitely, this already generate some revenue, but revenue is a small, still very small.

We are expanding this design service products to every bank in Taiwan, every, okay? Which bank if they ask on-premise AI solution, Phison, most likely, is the only capable company to provide.

Hospital: we have a more than 30+ commercial products shipment to the hospitals, which is with Phison aiDAPTIV+ AI Data Platform and FHIR systems. Today, I'm not plan to spend time here.

If you interest, go to check internet what is a FHIR. Next earning call, Phison is going to spend more times to tell you what our achievement details in this AIDP, Phison AI Data Platform business.

We also ship over 20+ of SME. What is SME?

Small, medium-size of a company and enterprise. We ship our server service to them.

We are going to expand this business. By this year, we have a confidence to win commercial business over 100 to bring in Phison AI Data Platform aiDAPTIV+ service to the new our potential customer: bank, government, enterprise, hospital, and education, okay?

We also start to work with the AIDC, AI data center, to put Phison's our software platform into data centers. This is totally our new business.

We take fight- Taiwan, because all engineers Taiwan, right? Now we we will show our performance in Taiwan.

We have a bank We have a shipment: bank, hospital, enterprise. Now we start to move to North America.

Phison already have office in the San Jose, in the Colorado. We new start set up office in the Vancouver.

In Vancouver, start to hire the team, transfer our products to them, our design to them, mainly aim to the US, North American potential B2B SME business. At the same time, Phison has a new foundation in Malaysia, my hometown.

We use them to start look to the Southeast Asia. India, we have a JV called MaiPhi, which transfer this technology to them to help them service the local SME as our new business.

Again, next earning call, we are going to tell you more. This is a go to the AI PC, AI-on-device.

A lot of activities, a lot after COMPUTEX, from now on to CS, a lot activity to promote the so-called AI PC or agentic notebook, okay? This is happening, and Phison aiDAPTIV+ is the only solution.

Investor keep asking about the revenue. Yes, not yet happened that much, but coming.

This is the trend, and Phison has a heavy investment into here. Okay.

Shareholder may again complain, but we keep Even first half good profit, we still have a 22% investment in R&D, which is still by amount, is a record high. We keep to push for more R&D development.

Versus last year, the whole whole year, the R&D investment was a 13 billion NT. Now first half already 24 billion NT.

This is commitment. We are very consistent.

We keep hiring more engineers, again, not for chips, not for firmware, but for our AI Data Platform, because we can see this will be our next future big growth in our revenue. Again, this page to show you we are heavily invested the R&D activities to make sure we are able to growth in the coming years.

Okay. Last, I want to show you.

We define a new business platform called Phison AI Data Platform. This already ongoing, already have a commercial business, a lot of engagement is happening.

Again, I don't want to spend time. Just let you know what we are doing.

Next quarter, you will see more, okay? This is we call infra is a hardware: CPU server from ODM, EMS with the Phison aiDAPTIV+ cache; Phison defined our own storage server with our SSD.

We develop every software. We have a HCI, a lot of function.

We license this to our partner, to the customers, to data center. We develop different kind of functional modules to make sure it's good to use.

Then we start our CVC to engage, develop, support a lot of a new ISV to design application in Phison AI Data Platform. Phison in-house, we are now pushing internally to use our AI on-premise.

We have a AI server in Phison server room. The engineers start to heavily I'm every month review, I push them to use more service from our in-house AI service.

By June, Phison saved 140 headcount engineers, okay? Just 1 month, we are able to take back every server investment.

You can't believe, okay? This is happening.

What we are doing in the our on-premise AI, we use mainly is AI agent and AI copilot, okay? I don't want to go details, but this to show you what we are doing in our AI activities.

By ranking, we have a lots of engineering activity here using AI to improve our cycle design cycle and also the performance. Since I realized Phison our we name is agentic hub, this is so-called agentic hub, okay?

I can save a lot of headcount. We, Phison, placed a big order to our ODM Taiwan to buy Nvidia more than $50 million of a GPU card and servers.

We are going to, step by step, install into our server room, and also we are going to rent data center outside to put the server there. What's the purpose?

The purpose is target, we are going to save close to 500 headcount engineers by end of this year by AI service. We are not going to hire that many ASIC designer, firmware designer, software designer, but we only increase a lot of AI deployment, AI software, AI promotion, AI business development to our new AI Data Platform business.

The all the chips design, product design, we are going to rely on the new AI server, which we just purchased, around $50 million. This page to show you why we insist to go on-premise.

Why not going to cloud? We made a simulation.

We used our June our application, the input token, output token. We compared to the cloud Sonnet and the cheapest China server Kimi k2.

We found if we have a good utility, let's use 100% my server, my cost is here. This is a cost of Kimi, this is a cost of Sonnet, okay?

At the end, you if you are the heavy users, you only choice is go to on-premise. After your server all depreciate, that means the compute power is all you can eat by free.

You only you pay only the power supply fee, okay? This to show you Phison was a users.

Unfortunately, we can't find solution. We forced ourself become provider.

Now we are a developer, from user to developer. We develop our AI solution to help save our overhead, improve cycle.

Now we turn to provider. That's why next earning call, I'm going to share with more, okay?

What's next? Next step, Phison will aim to small, medium-size enterprise B2B AI on-premise total solutions provider.

Not only for Taiwan, but also aiming to North America, Southeast Asia, India. Hopefully next year, we are able to enter worldwide.

Summary. I think no need to spend time here to tell how difficult to get the supply.

AI is changing everything. Phison is from the controller, consumer, embedded, computing, now enterprise storage, main drive.

Now we are going to as a system solution providers, okay? This That's why we need to keep heavy investment in our R&D.

We are going to build our own ecosystems. Of course, we also need to engage with every key players.

We believe our SME B2B business can attract a lot of suppliers, system integrators, users. This is our next growth and our opportunity.

We invest this over 3 and 1/2 years already. Finally, we can see something is going to happen.

Last: Phison is a innovation company. We keep very consistent, invest, find the right way to help Phison to growth our business and to keep the good profitability.

Few Q&A here. Let me close, then we can go live.

Dividend: we board member decides Actually, we have a debating in the board members, okay, which is good. In the last policy, we try to match 55% of the net profit, but this time we decide not to do that.

Why? Phison, we need a lot of a cash to help to growth our business, to help Phison to bring to get more market share, to help Phison to transform from 2.0 to 3.0, to help Phison from modules, applications to the total system suppliers.

We need cash. We like to use a index as a interest rate, okay?

Based on the yesterday's our stock price, we based on the 60 NT is a half year 2.87. Second half, I believe I have a good confidence we still have a very good earning.

Hopefully, next first half, January, Phison, potentially, I hope we can make the more earning. We are a little bit cash-rich.

We can consider to have a more dividend by ratio at the next January. Again, this no commitment.

Depends on our cash flow and our business and how good we are able to get the cash from the business, okay? Some shareholder may not happy, but some shareholder keep came to me to ask me, "Don't pay that much."

They wish the cash keep in the company to help company drive much faster, rapidly, to get the better return. Let me clarify: in the last many years, Phison, we are good in the earning per share.

We give 55%, everybody happy. Unfortunately or fortunately, now we are turned to special turbo mode is in revenue, in profit, in market share, in applications, okay?

This for what? This to make sure Phison can lead in the market and can enjoy in the last few years business return from the market.

Please understand. We understand the the the the ask from shareholders.

We keep 2.87% half year the the the the the yield, but we need to keep the cash to run the business, to make our shareholder get the more return, not only from the interest, but also for the capital gain. Again, [laughs] analyst, not some, few, maybe one, keep asking Phison as a module house.

Answer is yes, we are module house. What?

Phison module house is a different with what definition of a module house. Traditional module house is a buy-sell, buy, assemble, buy-sell.

Phison as a module house, all activity starting from R&D fundamentally design. Revenue-wise coming from modules, you can name this is a module house, okay?

We use this example for many years. Again, I had to use this, okay?

You look to the Apple and Nvidia: they are chip designer, they design the chips. Most likely revenue not selling chips, they build modules, okay?

Revenue from modules. When I talk to analyst 2 weeks ago, he said, "No, no, no, no, you Phison different different.

You are module house, Nvidia module house, but they are in-house technology." I said, "Hey, Phison controller also in-house technology."

Then he said, "No, no, no, no, no. You controller is only a small BOM cost.

The rest is a flash. All you see the company Nvidia is all about their design."

If you read the report last week, Nvidia new weather brewing platform, memory BOM cost 62%. What's the difference?

Okay? I just want to tell you: we are really consistent develop our controllers, heavy investment, but good thing is our revenue coming back from the modules.

You call me module house, I'm okay, but we are able to generate the value in the market, generate value to the ecosystem, and generate value to the shareholders. Please study more and please understand what we are doing for company, employee, and shareholders.

Phison from everything 100% in-house, going to the products, and we go to the revenue by modules. Now we transform ourself to 3.0, Phison AI Data Platform.

I hope I can show you more number, more visibility by next earning call. Again, I have to use this is a comparison, okay?

How Phison able to keep growth our technology? Simple.

You can see. Shareholder may complain, this is all about our R&D investment versus our revenue.

We just go to the Bloomberg to search, look Apple, Nvidia, they're not small, they're big. They have a big revenue.

If go to net amount R&D spend, they're also big. Phison is a tiny, small company, but we have a much higher percentage.

Means what? Means we aim to very long term and we are able to growth by what?

By revenue of modules, which analyst may not happy to see module, but unfortunately, this helping us to make good enough of profit. Last, please understand: we are module house, but different with the module house you mean.

This is our inventory. Again, we have a by C Q2 90+ billion NT inventory, most likely enterprise, okay?

Then embedded OEM and industrial. I think retail in Phison in the coming two, three, four quarters gone, okay?

Retail may back by 2028 when the market start coming back. Phison is going to use our resource to gain the business.

At that time, if I ask business, means the profit definitely had to be good. If not, why I spend my resource to there?

Back to the ESG, we keep doing all our best, okay? Every index we keep improving.

Thank you. Let's go to the Q&A.

Thank you, K.S., for the encouraging updates and congratulations to the strong results. We are now opening the floor to the Q&A session.

If you have any questions, please use raise hand function and you can also type your questions in the chat box, and then we will raise the question for you. When we are waiting for questions, I will start with some questions from me.

My first question is about the NAND demand and supply dynamic. As you just mentioned, NAND supply remains quite tight and you were actively adding supply sources.

Can you give us some update on your current supply condition? For example, how many suppliers do you now have LTAs, and how it compared to what you see in the last quarter?

What can we expect for the incremental bit allocation for maybe second half this year or 2027?

Thanks, Astoria. To be honest, in the last 9 months, I spent too much my time by interview, by TV, by everywhere, keep talking about supply, demand, supply, demand, okay?

I don't want to go deep, go details today. Supply, demand, no solutions for many, many years, okay?

Back to Phison, we still able to get supply, but unfortunately supply not enough to what we need. Last week was the FMS in San Jose, we meet every suppliers there.

They recognize Phison's value. I'm telling them Phison is not going to take their flash, make the modules, and sell the modules.

We are now converting ourself to the NAND users. Just now I mentioned, we are having our business in the so-called AI Data Platform, and this is a number we show to them.

They are very surprised what we are doing. I'm telling them the NAND what we purchase is not only for selling to others, but we build to the systems.

This can bring more values. We can't disclose what kind of LTA, how many exabytes, because if I say something here, my competitors may go there to ask the same amount.

I'm not going to do that. Just a commitment here: we have a supply, but supply still not enough what we need, okay?

Thank you very much. My second question is about the NAND price in the second half.

How do you see the price trend in the second half compared to the first half? I remember our peer, Sandisk, they said in their earnings call that price increases will moderate.

Does that match what you are seeing now? How should we think about the gross margin in second half and onward?

I remember by March, I already saying that the memory company, the gross margin bottom line 80% easier, easier. 85% may happen.

Now I'm telling you, if 2000- 2026 90% gross margin may happen, but I don't think this is healthy to the industry. My sincere advice: don't do that.

If you ask component for 90% gross margin, industry potentially will be killed. You can see the IoT, you can see the consumer die, okay?

I don't want to comment what price is going. Supply tight, demand too strong, price definitely was still going up, but 85% more than enough, okay?

Thank you. I already see some investors online.

Maybe next question is from Charlie Chen, Morgan Stanley. Charlie, you can-

Yes. I unmuted myself.

Hi, K.S., and thanks Astoria. Definitely want to congrats for very strong results.

Let me probably just get straight to the point, right? For your AI-related business, right, it's at 38% of revenue.

May I know whether eSSD is kind of make the most of that business? How big can that eSSD exposure into next year?

This 38%, main drive is around 30+%,

okay, because this are more in the revenue generator. Boot drive, we win a lot of program and share, boot drive play a big role, no matter in the AI system or the networking.

Okay. Again, aiDAPTIV+, direct solution still small as I mentioned, right?

This is a totally new, we need to spend a lot of time. Now we convert this to the AI Data Platform.

Hopefully, hopefully, from C Q4, C Q1, C Q2, we can see the significant number in the revenue.

Okay. I guess two follow-up questions, right?

It seems like this eSSD business continue to see meaningful growth. Do you have a kind of roughly market share, I'm not sure how big is your TAM, but do you have a kind of rough market share for your eSSD business?

My another follow-up question is actually to Asturia's question about the NAND price trend. I just want to get a sense whether the consumer NAND and the AI NAND price has the same correlation.

I ask this is because I admit that Phison is no longer the consumer-related player, but my concern is that if a consumer NAND market price shows some weakness, whether it would also impact your current AI storage price. That's my two follow-up questions.

Thank you.

Phison, we are not a NAND manufacturer. We are not aiming to the enterprise SSD market share because we are not NAND supplier.

By the way, the NAND is shortage, right? We take our limited supply, limited is quite a lot, mainly go to the customized, okay?

Even go to the main drive, most of that is a customized products. That's why we can enjoy the much better gross margin, okay?

I read a report is a pay report, US report. Phison is around 1.9%.

This is what I read, okay? 1.9% of enterprise SSD.

This is what I can share with you, you can go check the report. Back to the so-called consumer NAND, AI NAND, or I'm like to more easier, downgrade NAND and a good NAND.

Downgrade NAND, the price also increased a lot and their supply declining a lot. Why?

Back to the history: when the 2D NAND, every 18 to 24 months, they have a new NAND generation. The NAND generation not mature yet, N+1 coming, every period they have a lot of downgrade.

Now 3D NAND, the longevity getting much longer. You can't believe the NAND produced by the six NAND vendors, the yield rate can be 90+%.

Or my people report to me they have a some wafer with 95% of a good yield. Means what?

Means the downgrade part declining a lot, okay? I don't think this downgrade NAND will impact to the AI NAND.

Who is going to buy consumer downgrade to the AI system? Again, you keep saying that the spot market dropping.

Go back to check YouTube, which I already take by March. The time I already hint: after June, Shenzhen module house is going to dump their wafer.

Why? Because their business 90+% are in the consumers.

Consumer dead. They need revenue.

If they're public, they need revenue. They buy wafer, they sell wafer, okay?

This happening. I like to share with you: in the June, July, Phison, we spent our cash more than $450 million to buy the NAND wafer from Shenzhen module house.

Of course, the price is much cheaper than the NAND contract price. If not, why I'm buying that, right?

Why they sell? Because Phison has a solution.

I buy. Don't take that spot price is a regular price, because NAND guy is not going to decline their price,

okay? Understood.

Thank you.

Yeah.

Thank you. Next question is from chat box, Bruce.

The investor wants to know, "Longsys recently co-exhibits a general DRAM offload solution for edge devices with UNISOC and is also collaborating with AMD on HPLC + SPU. How about Phison?"

This is the first question from Bruce.

In the March 20-something, they have a so-called China FMS, the Chairman Longsys was my previous speaker, and the next. I surprised, he said he got this, something, something.

When I on stage, I say, "Hey, don't copy. You are a big company.

You are innovative company. Why you just copy?

Come to talk." This is my comment.

I don't care. Anyone want to copy, we have our weapon, we have our solution, but wait, see how Phison react.

I don't care, but we win everything because we are the innovators. We are the inventors.

The worldwide know. We start promote this since 3 years ago.

No hurry. We wait and see.

At least in the world, many countries still respect to intellectual property, okay?

Thank you, K.S. Next question from Bruce I think is a follow-up.

He said, "Apple build its own SSD offload specifically to ship Apple Intelligence, but most laptop OEMs clearly don't have that kind of capability. AMD choose to partner with Longsys, while Phison choose to partner with Intel and MediaTek?"

Asturia, how about going this way, okay? I think the audience, okay, we no need to spend time here,

okay?

Apple is Apple. It's all proprietary, no comment.

You name AMD chose Longsys, wrong, wrong. This only happened in some show AMD China, Longsys.

You go to ask AMD Austin what happened today, okay? I don't want to comment here.

Phison chose partner Intel, MTK? Okay.

Don't worry, no rush. A lot of NDA there.

You wait and see, okay? I just saying that don't copy.

Be a man. Then we go to the next

Next question is from Simon Woo, Bank of America.

Okay. Hey, Simon.

Hello?

Oh, hi, K.S. Can you hear me well?

Yes.

Great. Congrats.

Great result, great number.

We work so hard. We act so much.

When I'm talking, nobody listen.

We are listening to you, sir.

Thank you.

That's a joke,

that's a joke.

Yeah. Maybe number one, how long you think the NAND price, it's still rising, right?

When it will stop? You mentioned that

memory

makers are very profitable.

Try to call you to ask me, call to your Korean friend, okay? I wish 85% gross margin is really good.

Don't ask to 90.

Okay. You think it is a time to see more stable NAND price these days?

No, not stable. What I like to mention is a saturate.

Still raising, but the ratio getting smaller, and eventually saturate.

Saturate means

Means hit a certain point, gross margin there, supplier is happy, then they start to pick up who they are going to ship.

Okay. Sorry.

NAND price still some room to see the NAND price increase for second half this year, but how about next year, sir? Stable, right?

No? Hey,

let me ask you: if next year nobody doing things, right, the gross margin of a memory company go to 95%, do you believe this is going to happen or not?

Theoretically possible, but I think like Taiwan foundry, shortage is there, but they are not raising the price aggressively anymore.

Memory is different.

Different. I see.

You are feeling that the shortage of NAND will continue into next year?

Many years, not next year. Many, many years.

Oh, many years. You don't see quite significant NAND bit growth per wafer with 200, 300 layers, sir?

No, no,

no. You go to ask the manufacturers, okay?

I met the one NAND, the top-level guy, I said, "Hey, from investment to production 2 years?" He said, "No.

It's a 4 years. 4 years."

Not easy. I realized, not easy.

Okay. Then just in case I'm double-checking, sir, your inventory value, 90 billion

NT- It's a lot.

-is just a book value, right? You have some Book value,

we never revalue, okay? We, Phison policy, never revalue.

We only write down.

Only. Okay.

Following the GAAP. That's a great.

Then would you recap your recent convertible bond issuance? Because your competitor also issued CB last nights.

Why? Why?

Why? Looks like that

Because the consensus indicates that maybe your competitor need extra cash to buy more NAND chips for the enterprise solutions, working capital increase.

No, they are the controller company. Why buy the NAND?

Because enterprise solution need the NAND chips. Because it's a module business.

Enterprise solution

Charlie is here. Then Charlie can ask, "Hey, they buy lots of NAND, how about gross margin?"

Gross margin usually the controller should be higher because only the fabless business model. Module theoretically should be lower, but it is growing very strongly.

Anyway, I think my question is would you recap your previous funding? It was what?

ECB, $500 million, well executed?

Fully used.

Already used. Used,

already already turned 1 and 1/2 times, already turned 1 and 1/2 times.

It was a $500 million US dollar equivalent? Sorry, would you remind us

800. Huh?

Sorry?

$800 million, ECB.

How much?

800. 8.

800? Phison raised the $800 million US

dollar CB, ECB, $800 million,

yes.

CB. Oh, so that implied what percentage of the dilution?

Not yet, it's not happened.

The equivalent-wise.

You are some by NT$300-something NT, right? 10

10% or

No, no,

no, no, no, no, no, no,

less, less, less, less,

less. Low single-digit.

Less, that's single-digit, low

5.

Okay.

If I remember, was 3-something, 3.

Yeah, 3% something because your conversion price

Yeah, yeah,

the market level.

Yeah.

Anyway, you used up the cash proceeds, $800 million from issuing the ECB, you already bought a lot of NAND chips, that's why your inventory is very high.

To me, it's too low.

Oh. Too low.

Okay. In that case, you believe in a second quarter revenue already strongly up and then September quarter Oh, no, no, no, no, no, no,

no,

no,

no. I have to manage my inventory to design in customer.

Design in customer not buy everything 1 night, they buy 1 year, okay? Next year, supply is getting much, much, much, much difficult.

You mean the NAND supply will be tighter?

Yes.

The percentage of a shortage next year, bigger than this year?

Yes. Yes.

Yes, no doubt.

That's why

Simon, you have to agree, I'm a really good, accurate fortune teller, okay? Since last year, I keep saying a lot, nobody listen to me, right?

Now regret.

Okay.

I think we need to spend some time to others,

okay? Okay.

Sure.

Okay.

Thank you.

Thank you, K.S. and Simon.

Next question is from the chat box from Qing from DIGITIMES. The question is in Chinese.

I will translate it in English first.

She

wants to know the update about the SFIPC, the Investor Protection Center. 請問KS,這個有關投保中心的這個進度是否能跟大家更新一下？謝謝。

是這樣子的,我尊重投保中心的訴求,那如果他有任何的證明我們過去經營的方式讓股東,當下股東,那時候股東有損害,我負起全責,我賠償。但是如果沒有辦法證明,拿出子虛烏有一直這樣拖,這個對整個社會,對司法的資源,甚至對現在的股東是不公平的。我呢誠懇的希望大家呢回歸事實,如果有證據我們就把它給做一個賠償,而不是在那邊拖拖拉拉。那在過去我被告知因為呢有這個官司,我們在MSCI一直有障礙,那也很欣慰,之前呢有媒體願意呢把這個東西寫出來,那我們也順利的進了MSCI的這個裡面。那我要呼籲是投資保護人中心,你要保護的是每一個時間點的保護人,不是只有那個時間點保護人,用那個時間保護人來拖累現在的保護人的投資,其實我覺得這是不正確的。謝謝。 Thank you, K.S.

I will rephrase that

Astoria, we know run of time, no need to say, okay? Yeah.

Sure, sure,

sure. Okay.

Anybody interested, you can email me, I can talk to you here.

Sure, sure. As we are running out of time, we will conclude the call yeah.

Asturia, how about

let me take over, okay? Let me take over, save time, okay?

Don't trip, that is Silicon

I don't want to spend to comment Silicon Motion. I'm just telling you, many years ago, I'm saying controllers is no hope.

No believe, right? Phison so lousy, and Charlie say we are Phison is module house.

What's wrong with module house? We're doing good.

Looks like SIMO now transform themselves to module house. If you are the investor, you are the analyst, you have to start to ask what is their gross margin rate in the futures, okay?

Next, 記憶體廠庫存 LTA 影響記憶體毛利率是庫存. We have our own LTA, SCA, no matter what, I'm not going to disclose because sensitive NDA.

Uh 毛利率是庫存會下來. 哎呀,這個簡單數學啦吼,你一直買一直買,你的毛利率不可能一直等比例上去啦,所以毛利率本來就應該要下來,只是是怎麼下來。有模組公司一個季度下十幾趴,那我們在漲嘛。那後面我覺得我沒辦法跟你預告,但是至少我做一個工作就是,我們一直投資我們的研發,希望呢大家都買一樣的NAND的價格,人家加3%賣掉,我可以加30%、40%、50%做成不同的產品,這就是研發投資的價值,OK？那這點呢,我沒辦法跟你證明什麼,那你就等時間,我講過,等等看,wait and see。如果你不相信,去看過去10年來,我們跟所有的模組公司的毛利差別,你大概就可以理解了吼。下一題, claim to be using SLC to high pure per peak.

Okay, 這個是那個叫StorageX啦吼,我們也在裡面啦,所以時間關係不comment了吼,我們也在裡面。Uh then, Boris, English, sorry. "Strong result, superior product, superior margin, difficult to tell how much comes from your value add and how much using-" Okay, again, back to the so-called low-cost inventory.

Again, I'm I'm not able to describe what is that. Agree, definitely, we have a low-cost inventory, no doubt, but you can see, if left hand I have a low-cost inventory, good profit, right hand, I spend this to R&D.

I move this to R&D, okay? This R&D is going to generate the future momentum for our business.

Again, what Phison has to do, if price keep going up, the gross margin definitely cannot be linear, but we need to have a new product. I keep saying that we are moving to so-called AI Data Platform.

This next quarter, I'm going to tell you more, okay? This would be our next growth, and we have to do that.

If not, the things will follow what you said, okay? Gross margin, I fully agree.

Thanks.

What's the next?

I think Charlie Chen also have questions, he raised his hand.

Yes.

Oh, no, actually my questions were all answered and yeah, yeah, all good for me now. Thank you.

Thanks, Charlie. I think again, Phison keep telling the market one really consistent message: I know price going up, eventually gross margin rate will drop.

This is simple mathematics, I know. Only thing to move the gross margin is new invention, new product.

This is what we are doing. Nvidia many years ago from cheap, they sell GPU to gaming car maker like ASUS.

10 years ago, they sell the car to the mining. Now they sell the full rack, okay, and they mark up their gross margin.

We name this is a total systems. Phison from controllers many years ago, modules, now we move to AI Data Platform as a systems.

Good luck to Phison. Hopefully, I can show you our big transformation within 12 to 24 months.

Thanks, K.S. That definitely good luck for that.

Also comes to me on this question because I think Astoria asked the question about Sandisk, right? You also sort of agreed that gross margin would be better to normalize a little bit.

Memory is taking too big cake from AI CapEx, right? What are thinking going to happen in the coming one to two quarters?

Do you think does the CSPs or, for example, Nvidia, Google, AMD, will try to de-spec the memory or they will try to pressure the price?

They have to. No, Charlie, sorry, they have to,

okay?

Definitely. This happened.

I'm saying that since end of last year, right? Now you can see a lot of systems that cut half of DRAM, right?

Let me ask this way: do they cut storage? They cut DRAM, yes.

Did they cut storage? No,

right?

Storage is income. Agree?

Inference two part: power, storage,

okay?

DRAM is here, you can use algorithm or you can just sacrifice the performance, but storage is a money, okay? AiDAPTIV+, what's the difference between aiDAPTIV+?

aiDAPTIV+ using flash to complement DRAM.

Yeah. My only concern is that during this sort of normalization or kind of negotiation, yeah, there could be some sentiment impact to memory stocks, right?

I agree with you that the fundamental remains to be very strong, just capital markets, right? Especially for momentum investors, they are kind of looking at those kind of price negotiation very closely.

I guess that's why people really care about your view.

Yeah. I think we are at 1:00 time, but still I think since still 170 audience there, right?

Simon, you have something to ask, Simon? Hello, Simon?

Yes, sir. Just a very very quick question.

Second quarter results very good, 65% revenue growth. It's more price increase driven, right, rather than any meaningful volume increase Q-o-Q?

By exabyte we growth, and of course price we growth.

Yeah. Another question is your volume growth trend in terms of the bit, how it can be meaningful?

Because NAND chip makers are saying 20% bit growth or a little bit lower, how about your overall the product volume growth? No,

product volume actually declining.

Declining, I see.

Modules declining. No, depends on Consumer decline almost to nothing, okay?

I have to clarify. Don't mis-understand me, okay?

Enterprise module increase because we have nothing last year. Controllers increase.

All combined together, Phison's own bit growth Combine everything,

controller unit increase, exabyte increase. Yeah.

Maybe third quarter, September quarter, December quarter visibility, the gross momentum still based on your blended ASP upside?

Demand is there, strong, but again, I think I need to keep repeat: I have to manage my inventory. Why?

If I design in one program for 24 months, I have to make sure I can supply. I can't sell components tomorrow.

If I sell, I'm not able to buy back next year, I die, because the CSP will kill me.

Okay. Oh, sorry, one last question from some investors, they are asking this.

ECB issuance already done, but you are saying inventory still not enough, that means additional funding needed or you have enough cash to build up the further inventories, sir? Thank you.

Okay. By end of March, Phison has a big big debt from bank.

By July, we pay back almost everything.

No debt.

Really healthy. Means what?

Means coming 6 months, if I need to get cash, I can use bank. If business keep going strong, second half next year, yes, I need to raise fund.

Okay. All clear, sir, and we appreciate your valuable time today, and congrats, great result.

Thank you, sir.

Thank you. Since still 160 here, I want to 我用中文向各位在座個簡單報告哦。公司狀況我想這個不多說啦,都都這樣子的。那研發的費用確實是高,但是為什麼要高？因為美國公司來台灣挖人啊,我們培養好工程師,如果沒有辦法留得住,對公司長期是損害。那現在反過頭看誰有能力留得住,所以我們研發費用呢會強力的把人人好人才留下來,確保我們有更多更多新的生意,可以順利拿到。我們同時呢也用了大量的AI工具,自己開發的,把這個 我的政策是在整個IC跟產品設計上面已經不再找太多新人了啦,基本上就是用舊人配上AI,那提升效率,倍數的成長,那費用不再增加,那創造更好的營收跟獲利。有加人的部分,絕大多數在AI的這個開發、生意的等等的,那剛看到下個下個季度我會好好的跟各位表述我們在做什麼事。

那對於呢股東的期待呢,當然有兩有兩種聲音,一種說啊我就要55%。那董事會呢也在辯論55%我們發出去,銀行來借錢經營,那讓公司因為這個資金不足成長動能有限,那大部股東是說我們應該讓公司繼續成長。 那另外再表表述啊,我們這次是上半年哦,我們一年分兩次,上半年還有個下半年。如果我們下半年有更好的獲利,現金流更強,我們再評估下半年,因為這個股息是講股東會是一整年啦,一整年吼。

所以我們期待有更好的一個現金流,下半年看看有沒有機會呢把比例提高,但是這不是承諾,要看明年的整體的這個市況。但是基本的一個依據呢,就是我們希望整年的based on我們這個董事會的這個殖利率,我們基本上還是希望可以看到有5%啦,這是大部分股東的需求。 那更多的良心股東希望公司不要配太多股息,他希望公司把錢留著,讓公司可以更好成長。最後呢我要表述的是群聯電子呢這段時間會轉型成快速成長的一個企業,而不是一個穩定獲利配股息的企業。我認為這是這輩子只有一次的機會,我們要趁這個機會抓住快速的成長,然後呢在這產業裡面呢把這個最大的這個生意把它吃下來。

那剛剛呢投資人提到說我們的競爭對手也去這個弄ECB找錢。 這個只是呼應任何群聯電子呢過去講的IC很難賺錢controller,發生了模組,但模組呢馬上分析還是要講說毛利率,這個我們都經歷過了啦,都經歷過了吼。所以剛進來做模組毛利率不用太久啦,你們等著看吧,好不好吼。

那以上呢是跟各位呢這個台灣的朋友呢做一個說明,那也希望呢大家可以理解,我們的目的就是希望群聯電子長長久久穩穩定定一直的成長。謝謝各位。 Asturia, I think we run out 10 minutes, if no any question, I think we can end up.

Or in case because still audience there, if you have any question, you ask, I answer, okay?

Thank you, K.S. As we are running out of time, we will conclude the call.

Thank you, K.S., for your time today, and thank you, everyone, for joining us.

Okay. 感謝各位, thank you.