Good evening everybody. My name is Jeff Ohlweiler.

I'm head of research at Macquarie Taiwan. Very happy to be co-hosting Phison's first quarter '26 earnings conference with CEO K.S.

Pua. Interesting times in the memory market, always very happy to hear K.S.

talk about the NAND. With that, K.S., I'll turn it over to you.

Thanks Jeff. Good afternoon, good morning, good evening, ladies and gentlemen.

This is Phison our '26 CQ1 earning call. In the CQ1, anyway, we changed our present slide here, okay?

We try to make something the highlight in this chart. Now we try to differ our revenue portfolio.

We built the so-called AI ecosystem modules, which including the ESSD module, aiDAPTIV+ solution, AI PC, networking, which we are shipping right now, AI servers, and also the boot drive for the server OEM, ODM, and also enterprise AI. This is 38% of our CQ1 revenue, okay?

Believe in the coming years, AI ecosystem modules will be our biggest revenue cash cow, and we believe sooner or later this will over 50%. Embedded ODM, we have mainly is the mobile ODM, PC ODM, and the gaming module.

Phison, we gained the share in the mobile controllers, and also we gained the share in the PC ODM controller and also in the modules. Little bit surprise to me is the industrial module.

Even though the supply is tight, but industrial module still very, you can see here, very stable, growing, and this can clearly contribute more the gross margin to our business. This also deferred with the Phison heavy R&D investment.

We win a lot of trust from the IPC makers, automotive maker, and also in the space solutions. Controller-wise is around 10%.

Not because declining, actually Y-o-Y we're still increasing by unit, Q-o-Q around flat, just because the revenue grow too much by our AI ecosystem modules and embedded modules. Retail, as we already highlighted, just only that below 10%, and we can see in CQ2 potentially retail modules still declining.

With this kind of our portfolio, we are glad to see Phison is migrate from the consumer, embedded, now we are in the AI ecosystem partners, and we name this as a Phison 3.0. We are going to present you later in the more information.

Financial result, you read the revenue-wise is 40B NT, and the gross profit is 25B. Okay.

This CQ1, every index all 100% all-time high, every index, okay, revenue-wise, all-time high. Gross margin, 61.3% also all-time high.

We also be more conservative. We still have some inventory write-off around 1.5% of gross margin.

We play conservative, and some material which has long long day time, long stay, we just write off to make sure the company has a much better visibility in the future. 這邊呢, 也跟各位中文報告一下, 毛利的部分呢, 我們基本上還是反映得慢, 最主要是因為我們很多訂單呢, 是在三、四個月前接單, 後面才來反映出來, 所以在第四季的毛利是41%, 那第一季呢61%。那各位關心的這種什麼庫存的問題, 基本上在我們這邊已經不是什麼大問題了, 因為除了產品線完整之外呢, 我們在這個供貨上面有很大的一個保障, 那在產品的價值部分呢, 我們充分地表示出來, 我們的這個價值。Earning-wise, the EPS is 68.8.

It's again all-time high. In this 68.8, we have around 10 coming from the investment gain.

We have a JV in China, we have This is coming from investment, but our main business is still doing great, okay? By inventory write-down, we still spend like minus 2.3 something, okay, be conservative.

在這個獲利部分呢, 基本上我們還是採取保守做法了吼, 那庫存部分呢還是有做這個提列。那最主要整體的這個營收需求很好, 毛利率也符合預期, 那整體的利益呢也不錯。那對於未來呢, 庫存部分我們還是盡量呢打呆, 有呆料把它打掉, 放得比較久還是把它打掉, 讓整個庫存呢相對的更為乾淨。Income statement, okay, I think we need to highlight is of course gross margin already we we we we shared to you, Q-o-Q we increase in revenue, profit. One thing I like we'd like to highlight is our operating expense, and this is mainly R&D expense.

CQ1, the number increased that much just because we have to keep investing to the new node. We migrate to the 4 nm.

We have some mask cost, we just write off the mask cost, and we have a lot of AI storage system development and R&D expense, R&D talent recruitment, and to have more incentive to engineers. This is just because we like to have a heavy R&D investment to make sure we can create more value in the coming futures to improve our earning.

I name this is a value creators. The value creators mainly coming from heavy R&D investment.

I need to emphasize again, don't put Phison compare with other module house. We are highly commitment to keep invest to the R&D.

This to make sure we got the good return not only gross margin, in the earning, in the net amount, and also influence power in the industrial. This is a Phison DNA: heavy investment, we get a good future.

Don't keep asking Phison compare with the module house, compare with others, and also don't keep asking about the inventory issue, okay? Inventory good, we still can create the value because we name ourselves is a value creators.

Non-operating gain is contribute around 10 NT just because of the investment and the some JV returns. 這邊呢我們還是要再強調一下吼, 我們由始至終一直在強調群聯很願意在研發上面呢做長期的投資, 最主要呢如果我們不做投資, 只是靠短期的材料買賣, 靠漲價, 這個並不是一個 IC 設計模組公司要做的事, 這是模組公司做的事。群聯電子呢我們還是本著呢我們以一個 IC 為設計的這個根本的一個因素, 然後呢以營收呢從系統來取得, 所以才有辦法創造出呢這個 400 億的這個營收裏面呢有 250 億的一個毛利。那更重要是呢, 我們要確保在未來時間裏面呢, 我們可以有能力取得充足的貨源, 有好的產品線, 可以創造更好的毛利, 當然毛利率呢我們努力來把這個維持好, 最重要是最後的 EPS 呢我們可以有好的回饋給股東跟給所有的這個員工。那我們會持續地強調在研發上面的這個支出。In the our balance sheet, everyone may concern about inventory.

Inventory growth by Q-o-Q is 200%. Again, I need to emphasize, not because we like to buy, just because we foresee in the coming August to December, we may not able to get good enough of wafer, flash component.

We have to, since we design win into our system partner, AI partner, automotive partner, PC partner, we need to make sure we have a continuous supply in the CQ4. That-wise, we need to keep collecting component for our future use.

Again, you may ask about inventory cost. I just emphasize again, we have a confidence to manage this inventory.

No matter the you so-called low-cost or the market price, we are able to use our inventory to create good enough of return for our shareholders and for employees. Okay.

We have some TF and non-TF comparison. Basically, it's not much different.

Phison last year August or September, we issued the stock option to our staff, this create some cost. You can see the different between TF and non-TF is a small gap.

We not issue that much, the differentiation is not that much. Our April, the revenue is 20 billion.

Again, it's all-time high. Again, we still facing the fulfillment to our customer now around 30% to 35%.

We still not able to make the 100% fulfillment. We still go to ask our suppliers to give us a more upside in component.

We also do all our best to gain the more design win in the AI systems in the a lot of new applications. We still need to get more supply for our this second half business.

This chapter we like to share with you how Phison we name ourselves from the Phison 2.0 transfer to the Phison 3.0, and we try to name ourselves become the NAND flash value creators. Again, we no need to spend time to explain why shortage, how shortage, how long the shortage.

I think this already we agree this is a fact. Since it's a fact, from Phison's our revenue portfolio, you already know, we aim more to the AI ecosystems.

What is our product line in this AI ecosystem? We have a retimer, the driver.

We have a eMMC for server boot drive. We have a lot of AI software application.

We have aiDAPTIV+ solution. We have enterprise SSD.

Later, we also like to disclose with you, we are now start to build AI data platform, Phison's proprietary AI data platform for the on-premise solutions. This is all about the AI ecosystem, mainly for cloud, for server, for hyperscale, for near cloud.

Let's go one by one. This category is boot drive, okay?

We already got the two design win, and one new happened in the CQ1. For OEM, we have a three design wins already shipping, start shipping, and unfortunately the the fulfillment still below 35%.

In hyperscale, we have a three design-in, and according to the forecast, the MP will starting from this CQ3 and CQ4. We again, we need to ask for more NAND flash supply to us to make sure we are able to ship.

For the AI platform, this is mainly for networking, we already shipping, and we keep collecting more material for our second half upside demand. This category is mainly main drive.

Main drive is means high-capacity drive, up to the 122 TB enterprise SSD. We got the two OEM shipment, and in the Phison's our top 10 customers today, we have six are named as AI ecosystem customers.

Hyperscale, the infra operators, we have a two design win, and we're shipping exabyte already from February. In the AI storage systems, we named this near cloud, seven total players, we are there, 100% we cover, three are new, and we ship every month.

This also create and some of them already inside the Phison top 10 customer list. The infra software platform, we also start to engage with the software platform to make sure AI adaptive able to go smooth to the market.

AI enterprise customers, we got two new, two design win and one new. The last density, we ship exabyte in CQ1 not only in the Western country, but also in the Asian country.

Gen 6 controller, we have a silicon ready, first cut, we are going to move to MP, and the first Gen 6 enterprise SSD sample will release by this August, and we are going to show this in the FMS. Computex, we are going to make some demo.

We are the earlier player in the Gen 6 enterprise SSD suppliers. In the industrial IoT, we already ship our eMMC into the very start-up big guy in the robot systems.

We also put our UFS industrial part into the air craft systems, and this is going to MP very soon. This is just to name that Phison, we are very focused to keep develop the new products, to get the new high-tech customers, to bring the value to the company, shareholders, and employees.

The one thing is about space SSD. Phison invest this since 7 years ago.

Again, we got so many design wins, shipment. We have a regular shipment monthly.

Of course, revenue is not that big enough. Space is very niche, but this is to show Phison we are capable, and we have a good supply chain, not only technology, but also supply chain and service.

The most important things, we win the trust of the space ecosystems. One more thing is about during the Chinese New Year, they have OpenClaw in China getting popular, but at the end, the user realized that facing privacy issue and also the expense issue.

Phison, we work with our CPU partners, we bring in the on-premise desktop, laptop, which able to compute the OpenClaw on-premise without using the cloud, and this is proven. In Computex, we have a lot of our partners is going to show this product, and we believe this will be the trend, and we have a confidence, Phison aiDAPTIV+ solution will be the stand-up products in the coming futures in the every on-premise PC solutions.

What the user suffer is about the cost, the the expense. With Phison solution, we provide the on-premise, and with the hybrid cloud, we can reduce 70+% of a token cost.

The token cost can help users more willing to use AI with the hybrid solutions. Okay.

Phison solution is all the new platform with the CPU which they have a iGPU or NPU, they no need to use a high capacity of a DRAM. With Phison adaptive SSD, the hardware itself no change, hardware every everything is a similar hardware, just put Phison adaptive cache with our software, then this convert to the AI PC.

Actually, overall cost down. Why?

Because it reduced a lot of DRAM, and DRAM so expensive, okay? By Computex, we can see this is going to the market.

One things we we we agree a lot of our shareholders, the fund managers, the sovereignty fund, you care about ESG. We made a study about adaptive, how can adaptive matching the United Nations Sustainable Development Goals.

We go to study, we found adaptive matching three of their SDG. SDG 4 Oh, five, we have a five.

SDG 4 is a quality of education. The AI so GPU so expensive.

A lot of third world country, the student they may not able to to to touch AI just because their country may not able to make the fulfillment to deliver the GPU. With Phison adaptive solution, we are able to make education popular in the third world country, to let the student, they are still able to access to this AI systems.

The clean energy, SDG 7, SDG 13, which is a good, healthy to the criminal climate issue. Phison, this solution actually is a distribution compute.

It's not asking one country provide the heavy power supply to supply the data center. This is good to match the SDG 7 and SDG 13.

The one is SDG 9 is an innovation to reduce the inequalities of the third world countries. We believe AI has to be equal to every human.

Everyone they have a right to use AI, but again expense too expensive. We wish to work with our ecosystem partners, the CPU partners, the GPU partners, the PC partners, software partner, ISV, to deliver the affordable AI platform to worldwide, to help everyone they are able to access to the AI systems.

In this Phison 3.0, what's the Phison's product portfolio? We have a different definition here.

Our main business starting from controller. Again, Phison is a company we are not just a single module house.

All the Phison products from Phison are 100% designed in-house. Controllers, every controller's IP, except the the the the CPU we licensed, the most of the IPs, circuit are in-house.

Phison also doing ASIC. We also provide the design service to some hyperscale customers, and also to the gaming console, and also to the PC OEM, and also to the NAND industrial customers.

We have our in-house ASIC design service capability. We also provide the full turnkey with our firmware.

Of course, with the full system PCB layout, simulation. Other than the controllers, in the system level, we have a innovation aiDAPTIV+ solution.

We provide enterprise SSD, the boot drive, the software application program, retimer, eMMC to fulfill the whole AI ecosystems. On top of this, we provide the service is a full customization, high resilience, fast, flexible response, and the supply chain, we are able to make the supply chain.

We have a much healthy financial status can able to support to the industrial from design, from concept to design to products to shipping as a systems. This is Phison 3.0.

We have a confidence, with this kind of heavy R&D investment, we are capable to fulfill this industrial who need the storage, who need the AI ecosystems, and we can play the role. At the end, the return is we are able to get the good revenue, the better profit, and at the end, the the beauty EPS.

This is a new platform. It's the first time we disclose to the the public and first with to our shareholders, the fund managers.

This we name is a infrastructure platform. It's adaptive concept.

With the GPU partner, Phison provide the adaptive cache to enable using the limited of a GPU card, we are able to exercise fine-tuning and also the good performance in the inference with our KV cache solutions. We also working with the CPU server company as a hyper connection with our Pascari By the way, Pascari SSD is a Phison enterprise brand.

Recently, we get a lot of award. We also win a lot of recognition in the worldwide market.

This Pascari SSD is a component, but we work with the CPU partners, with the server partners, we build our Pascari into a storage server. This storage server is a customized provide to the proprietary application, mainly for the AI adaptive, AI ecosystem, and also for healthcare industry.

On top of a infrastructure platform, Phison, we also provide the hyperconverged infrastructure. It's a software level.

This 100% in-house software development. This is a the the the the AI the software interface which to handle the hardware itself.

Then on top of this HCI, Phison will provide the API, AI gateway. We have a different components to enable the AI application.

We have a different module to help improve the the the the application and performance. This is 100% all from Phison in-house.

We are providing the full system product and service. We start to build also the ecosystem ISV partners to bringing application.

This already they're porting their application on our infrastructure. Phison's our future business, we will have the the the infra as a service, we have the platform as a service, then we partner with the software as a service to provide the full solution to the industrial.

Hopefully, sooner or later, with Phison heavy commitment investment, we can deliver our new AI storage, AI adaptive, and also the storage system to the market. Let me use the Mandarin to explain more.

群聯呢一直呢很努力, 也很公開說我們要擺脫所謂你們說的循環的概念。Flash 過去確實是循環的, 那未來呢已經是一個結構性改變, 它很大的成分, 很多很長時間不會再循環了。那群聯電子在這裡面呢, 我們要怎麼取得我們的價值呢？如果我們還是維持著啊做 controller 吼, 賣 controller, 然後呢做點模組, 再過點時間之後呢就不長進了吼, 因為整個 AI 的這個進程跑太快。群聯電子必須要想用我們過去呢努力扎下來的這種資源, 怎麼再開創出一個產業需要的一個產品線。所以呢我們的人員呢也很努力呢自己呢花了時間呢去建構我們硬體的一個 infra 的 platform 吼, AI 的這個 server, 還有 CPU 的 storage server, 吼這個已經, 各位如果注意到話呢, 衛福部的一個比賽呢, 那個, 那個 FHIR 的這個 server 呢, 斐成取得的金牌了吼, 那這邊呢我們也會跟很多的這個 SI 跟醫院來配合, 把我們的方案呢放到全台灣的各個醫院裡面進去, 成功案例之後呢, 我們往外面再走。那在在上面的這個所謂的應用這種中介軟軟體呢, 群聯電子自己開發, 吼我們要 make sure 呢整個系統是是好是好用的。那在上面的各種應用的這種小 component 呢, 我們把它完成。最重要是呢, 以後的 ISV 找了我們, 他只要把他的軟體 porting 上來之後, 我們就可以在我們的原來的 infra 上面呢, 呈現出各種各樣應用出來。那在未來呢, 群聯電子的這個整個系統裡面呢, 我們會有我們自己的 Pascari 的 SSD, 我們自己的 boot drive, 我們的這個 aiDAPTIV+ 的 cache for 這個 KV cache, 我們會有自己的軟體方案, 那我們的收入呢就可以更多樣化, 我們的毛利就可以更好的一個表現啊。所以當大家在問群聯電子說, 啊你的這個循環, 我想循環已經不討論了, 你的這個模組吼這個毛利率, 這庫存, 我想這個再討論也沒意思了啦, 我們努力的投資找出我們的方向。我更相信呢在未來 AI 會一個幾十年時間裡面呢, 是一個需要大量的有能力的人來吸收它的養份, 把它轉成自己的資源, 那群聯會努力的往這邊來做, 也希望呢各位可以理解, 也可以來支持。So this to show you even though our revenue increased that much, we still insist to put around 20% of the expense into R&D. Again, this is commitment.

I keep repeating, don't compare Phison with other controller house and don't compare Phison with module house. They they're not able to invest that much into this heavy R&D investment.

This chart to show you what is Phison's our value creators. Today Phison in Taiwan, we have a R&D around 3,500 engineers.

Last year R&D expense is a 13 billion plus, and I can I I can tell you this year definitely R&D expense may go to the 20 billion plus, maybe go to 23 billion plus. Why?

You you can see we are going to invest a lot new applications. Don't worry, Phison made an investment.

This investment will become the return in the futures. We proof this year, few years ago's investment now we get the returns, okay?

Phison also built a big portfolio of product manager, okay? This product manager mainly is to talk to the customer requirement, the product requirement, technical support.

This is a good skilled engineers, we convert them to the project product managers. We also built a team called business development.

It's mainly to engage with the CSP, hyperscale, AI ecosystem partner, ODM partners, OEM partners. This organization helping Phison to gain to approach to engage more new business, more new application.

Global office, headquarter in Taiwan, we have a big team in the USA. A lot of hyperscale, cloud, near cloud AI systems is coming from USA.

We have a good engagement. We just start our Europe office.

Why? Because Europe, a lot of system integrator, the data center players, they keep approach us in Taiwan to ask Pascari supply.

Instead of this, then we decide to build office, start in the Frankfurt, German, and we start to go to send the people from Taiwan. Today, we have a local hire.

In Malaysia, we have a investment there. Start to ship the the SSD, ESSD, and also the AI service.

Hopefully Malaysia can help Phison to cover the ASEAN country and also the Middle East industrial and the Middle East demand. China JV doing good.

We can see the good return in coming futures and also help us to gain the market share. We have a India JV, starting getting the mobile business, automotive, and enterprise SSD business in India.

With this organization, we aim to AI ecosystem solutions. We try to gain more share, more return from this AI ecosystems.

Phison 3.0, we don't rely on much about inventory, low-cost. We rely on R&D investment, we rely on the new project, we relies on their more engagement with the big guy.

This is a I make it quick to show you how Phison use our adaptive server application, and we have a new team we name is a AI Nexus to build our in-house AI solution to support in in-house team to improve the performance. I just show some results.

Last year August, I start to host the team to provide the service to internal every BU, every a unit to use the AI. In the first month, August, you see the workflow just only three, but recently workflow go to close to 1,000 a month.

Monthly users already go to the 1,300 per month. What is a return?

This chart to show you this is Phison's our investment in the human resource equipment. This is a return from the AI.

By early March, already cross over. This is to show you the user user count monthly like increasing, okay?

The input token is 20 times of output token. This means what?

If you are the single user, you use a Gemini, ChatGPT, you input small token, output big token, because you just use as a users. Phison, we use this to read document, to solve internal problem, okay?

Heavy input token, get the small result. This means we using this to improve our efficiency, to reduce the human resource to read the document, okay?

What's our gain? After few months investment, by April, we can reduce a man-month 49 headcount, and this is mainly R&D headcount.

Okay. R&D the the headcount fee is increasing.

By the way, Nvidia is coming to Taiwan to build their worldwide other than US headquarter, they are going to hire engineers. No matter what in coming years, R&D the the cost is going to have a inflection, okay?

We need to start to use AI to improve our efficiency. By the way, Phison is not going to lay off engineers.

We just use more our current engineers to improve efficiency, reduce the cycle time to get the more products to improve our profit, okay? This is our gain after few months investment.

This is some small highlight what we do. It's mainly is engineering.

Actually I also do not know what this here, okay? This chart to show you every item we are able to save some headcount here, okay?

Target target. Right now we are able to save a 49-month headcount, and we already can see the return.

This is my highway cost, this is my monthly investment. Now we start can see the good ROI, will go to the positive.

The goal is in the coming 12 months, we are going to invest more GPU, more development team. We want to improve the manpower man headcount saving monthly, to improve the efficiency, to shorten the cycle time, to win the more project in the future.

At the end, we provide the better EPS to shareholders. Okay.

Let's make some summary here. Again, AI driven the NAND demand, no need to argue, debate, explain, okay?

It's happening. Just the question is, how can you, how can we to gain?

AI 的這個情況呢, 不用再多花解釋了。群聯電子呢, 我你們好奇是我們要從裡面怎麼榨取更多的養分, 吸收更多的資源, 讓自己更為強大。剛剛已經特別在強調我們會由很好的 R&D investment, 而且不只是做 controller 模組, 我們更多呢擴展到 AI 系統的應用, 從軟體裡面去做價值, 以最終目的增加我們的影響力。影響力到最終目的就是我們有這個價格的話語權, 有 premium, 到最後就是好的 EPS 來回饋給各位股東。然後整體的這個 supply 部分呢, 曾經, 我講個例子好了, 前陣子呢媒體說武漢呢長存擴產兩年之後就 oversupply, 我我真的覺得說這個有一點點, 要怎麼說呢？幼稚。這個 NAND 廠呢, 不管未來再怎麼擴產吼, 今年假設他的產出是 1, 他擴產, 兩年後產出不會是 10 啦, 就是 1 點多, 吼也不可能是 2, 因為 2 要兩倍的投資嘛。那各位如果用 AI 的話呢, 兩年內 AI 的使用者會增加, AI 的 token 的增加是幾倍？我我是不曉得了, 有人說是幾百倍啦。那如果 Flash 的產出是小於兩倍, AI token 產出是百倍, 那那我就不曉得說, 這個供給跟需求吼, 怎麼去平衡啊？所以我們歡迎原廠趕快擴產, 不擴產的話呢對產業反而不利。我們也認為呢原廠呢毛利率都很不錯的吼, 70 幾趴、80 幾趴, 確實價格應該要維持住, 不要再漲, 因為漲呢對產業不利啦, 不是對我不利啦, 反正呢米貴我買貴我賣貴嘛, 但是呢盡量的讓它可以維持好, 讓產業的好好的發展, 那群聯電子呢可以從裡面呢吸收更多的資資源跟養分, 來壯大我們自己。So far NAND companies start to invest, but don't worry, it's not going to oversupply, don't worry, okay? I no want to explain the time why it's not going to oversupply.

Again, Phison, we start to build our AI ecosystem as I explained just now. This is the things to help Phison to gain more influence power, more profit in the future.

This is a heavy R&D investment. Since we have a good profit Many third-party controller house, module house, you can go to study their R&D investment, even they have a good profit, they keep profit, they never invest that much.

After two years, we believe we can gain more business, more influence power, more product line, more design wins. The rest, Phison is not just a single controller company, we are not single module company, we are a ecosystem company.

We are able start to build from IP, chips, board level, module, service, now we provide the full ecosystem products. Okay.

We issue a ECB going to raise $800 million. Why?

Just because you can see inventory already go to the 70B NT, and we still need to keep collecting more supply for CQ4. We believe CQ4, when Vera Rubin is going to the market, the flash memory supply will be disappear, okay, will be mainly ship to the Nvidia ecosystems.

By the time when that happen, we still need to ship our product to the customer, we need cash to build inventory. 我們呢還是需要呢有充足的資金呢, 也跟各位報告, 不要再誤會說我們要去什麼故意去 build 庫存去賺價差, 不是這個意思。如果我們做了這麼多努力, 這麼慷慨, 這麼多 design-in、design-win 的 AI 系統, 啊到了半年呢, 年底的時候呢發現買不到東西賣不了東西, 那前面做白工了。那我更可以預測呢, 下年底到明年呢, 那個 Flash 供給會非常非常的辛苦, 因為新產品 Vera Rubin 上市之後呢, 他永遠是在最優先權的供貨啦。那時候如果你沒有充足的貨源的話呢跟庫存, 那個你前面的投資都是泡泡湯了。And we have a breakdown in our inventory.

You can see enterprise is a majority, and consumer now less than 3%. Means Phison start to reduce a lot of allocation to the retail, okay?

We aim to enterprise, AI ecosystem partner products, and also embedded, and industrial, which can create the value to the company. We open for the online questions.

Thank you, K.S. If you have questions, please raise your hand, I'll call on you and unmute.

I guess I can start with a few questions. K.S., can you talk about the ASP trend of NAND in the first quarter and how you see it in the second quarter and throughout the rest of the year?

According to the NAND suppliers, in component and the modules, keep increasing. I can share with you, by March, one day component, they increase 50% ASP, and by April, they increase a 20%.

Since the supply demand is a big gap and unbalanced, ASP definitely increased. Again, I wish ASP can keep flat or slightly increased, because that is not healthy to the market.

I know in the past when supply demand got tight, it I guess harder for you to get allocation from the NAND suppliers. Is that different this upcycle, or are you pretty confident on your supply not only this year, but for the next several years?

Yes. When I said, how can I get supply, right?

This is all my story. Then we have a question to the NAND supplier.

Why they wish to ship to Phison? Why?

Why? Because Phison, we prove we are the value creators.

We create this value. The most important thing is the AI going to the edge.

No doubt, AI is going to the edge, and the PC, laptop, desktop, they need to use AI. AI inference, you need a lot of DRAM.

DRAM is a constraint and high price. Phison, we have this innovation, right?

We are going to license our solution to the NAND suppliers. We engage with the two of a NAND vendors already.

Since we license, we work with, develop, initiate this business, then Phison has a better position to ask a more supply.

Great. I guess, obviously you had a big increase in your AI ecosystem revenues in the first quarter as you kind of talked about in the fourth quarter, 38%.

One is, how are the margins there in general versus the other products? Two, where should that percent of sales go to over the next several quarters?

Actually, the gross margin, the highest part is in the AI ecosystem modules, highest, okay, because it's new products, the market tight, and we have a very good flexibility to supply. Customer they're happy, we are enjoying the much better profit margin in this AI ecosystem modules.

Maybe one more question for me before I turn it over. I think Charlie raised his hand, I'll call on him after this last question.

You had 40%+ gross margin in the fourth quarter, 60%+ in the first quarter. I guess where can that go?

What's going to be a normalized margin once prices kind of stabilize? What do you expect for the next couple quarters on the margin front?

Phison is going to issue the ECB, okay, I already got our banker cannot have a guidance, cannot have a number, okay? I better to follow what they ask.

I have a confidence to make sure we can keep increasing our revenue. We can try to do our best to maintain the gross profit margin.

Overall, we have a confidence to increase our gross margin. When Flash price going stable, I always ask myself, what is the value of Phison?

If I just build a module, you can expect the margin go to single digit. Phison, we provide controllers, we provide AI ecosystem service, and we already start to have a software application business.

We have this investment since 2 years ago because we understand the day is coming, but we can had taking this as our compensation to our gross margin rate. I'm keep saying that we are willing to invest to R&D just because to create something good value to us in return in the coming years.

Great. Charlie, you want to unmute yourself and ask a question?

Sure. Thanks, Jeff, for hosting this call.

Hi, K.S. Congratulations, and you totally impressed us, and so beat my numbers a lot.

Some questions were already asked by Jeff. I think have a step back, right?

You keep talking about NAND or AI storage is no longer a cyclical industry. Do you think your business model can be less cyclical in the future?

Because I feel like second half, you just talked about, right, when the NAND or module price has start to stabilize, there could be some margin pressure. Is that the right way to think about this?

Yeah. If Phison at the last few years, we still treat ourself as a controller house or as a module house, we enjoyed the every years of profit, then we enjoyed life, then the answer to you is I pessimistic in the future for our future business.

You can see for that many years, Phison keep heavy investing, and I'm working very hard, right? We believe before talking about Phison, why don't we use some example in the history?

TSMC, when I was student 30 years ago, how's the TSMC business margin? They keep heavy investing R&D every year, every month, every day.

Now they are the king. If I'm telling you I have a confidence in future gross margin rate, but with small R&D investment, then means I'm just a blabbing, okay?

This is to show you we deliver what we talk. Okay.

I agree when price going highest, then the gross margin rate definitely is challenged. This is a simple mathematics.

We need to create the value, we're working hard to build the new ecosystems, we're working hard to provide the so-called adaptive for this platform. This is the new products.

Today, if you go to challenge Nvidia, 10 years ago, they are in the mining, in the GPU for gaming, right? You ask, "Hey, how about you?"

Jensen Huang definitely, "I'm working hard, I doing this, this." Now he proved.

Okay. I use the same wording, we're working hard.

If you believe us, then support. If you still have a question, my advice is wait, wait, see.

等等看啦, 不急啦. Because I try my best to tell you I'm working hard, working very hard, and the number we are putting in today is happening, right?

Just explaining, time will tell. Okay?

This is the answer for myself, and this is commitment from myself. We are going to deliver what we talk.

Great. Thanks so much.

I think yeah, that's very, very helpful. My second question is really about your progress in ESSD.

I think you show several design wins as a major CSP, OEM. Again, congratulations for those project win.

I'm curious about how Phison is going to work with NAND flash supplier, because that's just from my kind of humble opinion, I feel like the NAND fab that's still be mostly important product, right? If they can supply NAND, you can get a business, right?

I'm not sure how Phison can work with those NAND flash OEM to penetrate this ESSD market, or another way to ask this question is say why not those kind of NAND flash fab, right, to do this business by their own?

Actually, most of every ESSD, every solution to the CSP to users are highly customized, to be honest. They don't have the standard part.

Of course, they have a standard part, but every customer, they need to have a service and customize the solutions, okay? AI is changing so fast.

At the last 40 years, semiconductor did by Intel. Intel every year give you the 5 years roadmap.

Everyone know what is going to happen in 5 years, you can prepare. In the AI, Nvidia every years new generation, new platform, you don't know what happen next year.

Phison's advantage is we are flexible, we are rapid, we are good in the engineering solid, and the things customization. You go to the every NAND company, ask them their control team resources.

Definitely, they're telling you they are very tight, demand's too strong, too much design, they cannot cover. Phison is a third party able to provide from IP, ASIC, design service, firmware, board level, system, software, fully customized.

This is why the reason we are able to win that many project from the hyperscales, our capability. The NAND company, no matter they ship the drive, they ship the server, they ship the flash, eventually they are selling the flash, right?

If Phison can help them to sell more flash into the systems, then why not? They're support us, they're more than happy.

Means Phison to create our own position and values.

Thanks. Just very, very last one.

Thanks Jeff and K.S. for your patience.

Your 1Q inventory kind of doubling from 4Q, right? I'm not sure how much of those is from bit shipment increase, how much of those are kind of inventory price increase, and how much of those kind go to your 1Q gross margin if there's any reevaluation.

By the way, K.S., thanks for being very reasonable to all our research. I appreciate it.

Thank you.

I don't think I'm going to dispute our inventory, but I'm just telling you, I have a confidence, this inventory, we still have a good enough of a cost structure. Second, we still get a good supply from our supplier far than enough, but we can use this component to create much higher value to the products.

You can see Phison our position, right? You go to this chart, Phison is not going to sell just only a SSD, okay?

We are going to sell a solutions, right? Nvidia is not selling you a GPU.

He not rather selling you a GPU board, he sell you the full servers with the software bundle, okay? This is a Phison's approach.

In this case, I wouldn't care much about my component cost. I care much about, can I get enough component to sell the full systems?

Because this full systems, we are able to gain more gross margin rate. If you're staying in my position, I still treat myself as a controllers, as a module players, yes, you are right.

We already transformed ourselves.

Great. Thank you.

Thanks K.S., thanks Jeff. Thank you.

Thanks Charlie. Simon, how about you?

Do you want to unmute yourself and ask your questions?

Yes, thank you very much, K.S. Congratulations.

Great, great results in 1Q. The number one, very quickly the Q1 revenue very strong.

You've got any rough idea ASP change contribution out of the 80% quarter-on-quarter revenue increase in Q1?

Definitely is ASP change. By module-wise, enterprise SSD volume increased a lot.

Of course, the ASP increased more than a lot, okay? We also win more unit in the PC OEM shipment.

We are now become the key player in the third party or we are become the biggest third party module supply to the PC OEM and ODMs. At this moment, we have no interest to win the unit share because it's nonsense, supply is constrained.

We try to use constrained, limited of flash resources create the much higher value products. The answer is, yes, ASP is improving, but ASP improving, we are not improve in the retail business.

This is different. What I mean?

Many module house, CQ1 good earning, but you can go to see their distribution still is a consumer module. No problem, this is from inventory, but consumer module is like 8%.

Means we allocate more resources into these three category: AI ecosystem module, embedded, and the industrial.

This is

future.

Yeah. Sure.

Great. Very quick follow-up question here.

The table showing enterprise module, which means ESSD, right? Is your own branded ESSD

or

just a box?

Yes. Every enterprise SSD from Phison, we name as a Pascari, okay?

We ship to CSP, hyperscale, near cloud, infra, networking, and also SI and retails, Pascari. One thing I want to share with you, Pascari already recognized, 100% recognized in industrial.

When they name Pascari, they think this is good in the industrial, it's a tier-one. We are going to enable the ODM business.

We are going to de-brand our drive to our close partner who can help us to gain the more share in the futures.

That's a great point. Pascari means ESSD that is already used by US big tech companies?

Yes, means already trust, approved by the big guy.

In that case, you have to compete with directly major NAND chip makers, which also supply ESSD.

Not really, not really. NAND supply constraint means NAND supplier not able to fulfill the customers, right?

Phison mainly take their customization drive, not the standard drive, okay? In US, they have a lot of AI-related company, they're looking for customization.

The NAND guys say, "Hey, only 5 petabyte, they no have R&D resource." Phison take it.

Yeah. Yeah.

Okay. Sorry.

Very quickly, your April revenue, 20 billion NT dollar, wow, already similar to the previous year's quarterly revenue or half of the Q1 revenue. If we multiply by three, second quarter revenue 60 billion.

You think 20 billion is kind of the inflection point to see the ongoing like a 20 billion monthly revenue at least going forward? Again,

this is a little bit complicated, okay? My inventory here, right?

Yeah.

Some part I have to carry until CQ4 because I design win already and I worry CQ4 I cannot get supply, okay? Means I may not able to ship my inventory by CQ2, CQ3.

We need to make sure we are capable to fulfill the design win business. Means I'm no hurry, I'm no rush to make their revenue ramp up.

The answer, if asking about 20B is can be a bottom line, most likely, yes. Higher, definitely yes, but we have to manage the inventory carry linearly until end of the years.

Yeah. Your point is based on your great franchise with a NAND chip suppliers, meanwhile you have a 72 billion inventories.

In overall, the 20 billion NT dollar monthly revenue is not one-off. It could be sort of the starting point to see the monthly revenue at least.

True. True.

True. 你可以說 在哪一個辦公室啊？

在啊在啊 Sorry,

sorry, it's not me. I think

the voice 找個會議室喔

Unmute, wait.

201, 201 好了

Yes, okay, please. Yeah.

拜拜 Sorry. Then the very important thing is today you mentioned the KV cache memory, that is also emphasized by US big tech companies.

Would you spend maybe a few minutes what the Phison can do with your solution for the cache KV, key-value, memory area, how it works? Thank you, sir.

Simon, I can't disclose because this is covered by a lot of NDA. Phison's value, if I provide a simple KV cache 64 TB same as a NAND company, I cannot survive because I'm not going to compete with them, right?

The standard KV cache will coming from the NAND company, but Phison using our controller, firmware, system, we do a lot of customization to improve something, but unfortunately I cannot disclose that much.

Okay. Oh, one last thing, boot drive, your competitor also really emphasized with whether their Q1 result.

You think the boot drive is really necessary to turn on the AI server?

Simon, I think I set my principle. From now on, I'm not going to make comment on my competitors, but I just ask you a few questions, you can go to ask them.

Second half, no matter what, controller demand is declining. Revenue, if rely on controller, revenue is a big question.

To keep revenue, they better to build modules. Modules, why is the most easier to entry?

Boot drive. The same question just from Charlie to me, "You have a CQ1 good revenue because you have a last year CQ3 inventory, right?

Price go 5x." How about CQ3, you buy market price, you ship market price, you want to get your revenue, what is your gross margin rate?

Simple question. Mm.

Yeah.

Okay. NAND that you think the boot drive demand will be continuously stronger every quarter, that part?

Boot drive is a strong demand, but you have to make sure you can get the NAND supply, first. Second, you make sure the gross margin, again, what Charlie asking me, yes, we have a boot drive.

When price going up, they match the gross margin rate definitely declining, right? How Phison hedge?

Because we start the new business. If my competitors, they're willing to gain revenue, they cannot get flash, what's going to suffer?

Gross margin rate. Easier, right?

Yeah. Oh, one last thing, payout ratio still 60% something, even with your record high quarterly EPS number?

Would you recap your payout ratio?

No, no, this payout policy is end by CQ2 first half, right? We not decide yet.

Let's decide. You see, Phison, now we start to raise $800 million, right?

We need cash, but decision will made by end of June depends on the market situations, not decided yet. We still wish we can follow the principle 55%.

We wish, we hope.

How much? Sorry.

55%.

5 5?

55, yes.

Okay. All good.

Thank you so much, K.S., and again, congrats. Thank you.

Thank you.

Okay. Sebastian, do you want to unmute yourself and ask your question?

Yes, thanks Jeff. I will be quick.

I know time is running out of time. Hi, K.S.

Just a few questions from me. The first is, I know your wish, your hope is that the price does not go up too much from here, to stabilize.

I sense a contradiction in your message, because you're basically indicating that by Q4 or Q1 next year when Vera Rubin is up and running, the NAND flash industry going to experience a even more severe supply demand situation, which means the glut ratio are going to get larger. Economy 101 tells that when that happens, usually the pricing is hard to stabilize from there.

Just curious about trying to strike a balance between your wish versus the reality of the industry.

Yeah. In this world, every nation, they have a government.

If supply demand is big gap and impact to the national security, impact to the GDP development, then definitely government is going to do something, right? I'm not from the government officer, I'm not comment on this.

I wish the price keep flat or slightly increase, because the memory company already got 80% gross margin. We need to make sure the industrial can grow healthy, everybody can enjoy the AI, then government need to do something, this is my comment.

I also agree supply demand in the second half, especially end of the year, is a big gap. What to do?

My answer is, I don't know, but it's a wish. Please, don't increase that much anymore.

Based on the conversation you have with the six major flash suppliers right now, what's your sense on their attitude toward the Q3 pricing?

No. No guidance.

They want to ship you tomorrow, they decide by today.

Because if they decide earlier,

then

maybe a few weeks later, they will regret.

Yes. True, very true.

Okay. My second question is on your JV or investment in China.

Seems like the Hengxin has applied for IPO in Hong Kong Stock Exchange, I think the beginning of the year. Not sure about when will that happen, but seems like that's something in the pipeline.

You look at some of the other IPO like DapuStor or some of the NAND storage module suppliers, solution suppliers in China, whether they're listed in A-share or H-share, they're all on the lower profit or lower revenue scale, but they have a much higher market cap than Phison. What do you think about the subsidiary that we have going to be listed in China?

If they have a similar valuation versus the other peers in China, then how are they going to benefit us from investment gains or, etc.?

I'm not from Hengxin, right? I'm not get involved into their details about IPO, but for sure, they want to go IPO as soon as they can, and Phison do our best to help them to go IPO.

Hopefully by second half, hopefully. On the book, you can see Phison, we have a 22%+ share ownership.

Of course, when go IPO, after a year, Phison able to exercise some share to get the return, believe this is a very good number of returns. Second, I just read DapuStor today, their market cap is 2x of Phison, I'm asking, is this my problem or your problem?

If you give me the 50x, 80x of a PE, right, then we should be good, but unfortunately we are in the different market, no comment. Phison, we are doing ready every field of a storage.

Dapu is only in the enterprise. Dapu in the history, from the book, I didn't see they're making any earning, but Phison, every quarters, we deliver the earnings.

If you are to compare apple to apple, right, of course, it's different market, but I still believe the value of Phison is much below what the market given. That's why I keep working hard.

I want to deliver the much better business model to our shareholders, our market. We are not just in the single SSD provider, we are become the full systems solution suppliers.

Let's see. Recently I keep talking to all the fund manager, wait, wait, see.

等等看. Time will tell who can create the values.

Right. Basically, if you do nothing, just keep your shares there, and in the valuation, you back-calculate your holding, then that makes the valuation even more undervalued.

Or

if you take some profit and you can take some cash to buy more NAND

or

to do more R&D, yeah. No,

depends. After they go IPO, and no doubt, Phison's policy, we have to exercise gradually, right?

This is policy, okay? We need to get the return to invest back to Phison R&D, to make sure we are leading in the industry.

Right. That would probably help ease some of the cash burden or pressure that you have, be it on R&D or be on procurement.

Yes, very true.

Got it. Thanks.

K.S., maybe there's one last question before we finish. I know in the chat box, there's two questions.

I think Donnie's question on inventory and NAND supply, I think we've answered that, last question comes from Bruce. Says, "SanDisk shipped a lot of TLC to CSPs in Q1.

Why is Phison still shipping mostly QLC now?"

I don't know why you're asking me Phison ship only QLC. We ship also TLC.

We also shipping TLC, okay? We ship both TLC and QLC to the CSP to hyperscales, depends on what customer asking customization, okay?

Okay.

Then

last follow-up question from Bruce, "Phison's PCIe Gen 6 plan looks behind Micron's. Will this affect getting into Nvidia's CMX Gen 3.5 supplier list?

Also, Nvidia updated its DPU specs in March to Vera CPU to support Gen 6. Does this matter for Phison?"

No matter if Phison ahead the Gen 6 Micron, I am not the NAND manufacturers. How can I build the full 10 exabyte drive to Nvidia?

This is not my business, right? We know who we are.

Our controller is mainly doing customization. Of course, we engage with the NAND company, convince them to use our controller.

We just sell them the controllers. We use our silicon to build the modules, customize the products, to increase the value of the full product systems, then we can have a more premium.

This is our value, okay?

Great. I think that's all the questions we have K.S., over to you, any final comments or Oh, come on, they have

one question about Phison inventory strategy. Phison inventory strategy is simple.

If I get the design win in the business, I need to prepare the material, that's all. I wouldn't buy any component which I'm not going to use.

I'm not trading guy. I buy only what I need to use, and I have a 70+ billion NT revenue already, I run out my cash already.

I wouldn't spend my money to buy useless products. Every strategy is, I get design-in, I get design win, I need to keep inventories.

Okay. Great.

Should we pass over to you for any final comments?

Thank you for time, and CQ1, I know some of you may feel surprise, but I do all my best to surprise you again CQ2. Thank you.

Great, thank you very much. Have a good day.

Night, bye. Thank

you, bye.