Full text and analysis of the speech by the CEO of SanDisk at the 42nd Annual Strategic Decision Conference of Bernstein

Author: invest wallstreet

The core content of SanDisk CEO David Goeckeler's speech at the 42nd Annual Strategic Decision Conference by Bernstein on May 28 revolved around five key strategic themes: data centers will become the largest market for NAND for the first time, the reshaping of the new business model of LTA, disciplined supply restraint, continuous upgrades of the technology portfolio, and the construction of the industry narrative of NAND "de-cyclicalization."

I. Summary of Key Points

1. The Rise of the Data Center Market Driven by AI

Goeckeler clearly articulated the structural changes occurring in the NAND flash memory market during his speech. The core judgment is that data centers are expected to surpass the mobile market for the first time in 2026 (calendar year), becoming the largest application market for NAND flash memory. This marks the end of the storage cycle dominated by smartphones and PCs for over a decade, with AI infrastructure development becoming the new core growth engine. Customer demand forecasts have extended to 2028, and this unprecedented visibility of demand has prompted the company to engage in in-depth discussions with customers regarding supply agreements ranging from one to five years.

2. Transformation of Business Models through Long-Term Supply Agreements

The most strategic content of this speech is that SanDisk is fully transitioning from the traditional quarterly pricing negotiation model in the NAND industry to a new business model centered on long-term supply agreements (LTA). Goeckeler stated that the company is striving to establish multi-year contracts that "can balance predictable demand with attractive economic returns," transforming long-term agreements from price negotiation tools into key supply assurance mechanisms. He explicitly pointed out that the company's core goal is to "maximize two things—achieving highly attractive financial returns and making them sustainable."

CFO Luis Visoso further added that customers willing to make long-term commitments will receive priority allocation in a tight supply environment, while customers insisting on quarterly negotiations may face supply constraints. This strategy clearly demonstrates the company's strategic intent to leverage shortages as leverage to compel customers to accept the new business model. According to financial reports, SanDisk has signed five long-term supply agreements cumulatively for fiscal year 2026, three of which were signed in the third quarter, and two additional ones were added in early fourth quarter, with some agreements extending beyond 2030.

3. Disciplined Supply Strategy: Refusing Blind Expansion

In the context of extremely strong demand, Goeckeler exhibited a significant strategic restraint that differs from the behavior of industry giants in previous storage cycles. He emphasized that any substantial increase in capital expenditure requires a high level of confidence in multi-year demand at attractive price levels. The company adheres to its original capital expenditure plan of mid-to-high single-digit growth and has not expanded production due to short-term price surges, while also dismissing market speculation about prematurely deploying BiCS10 technology, stating that existing capacity plans are sufficient to meet demand. This disciplined supply management essentially draws lessons from the past storage industry's "blind expansion leading to price wars."

4. Technology and Product Portfolio Layout

In terms of products and technology, Goeckeler disclosed several key developments: the company expects to end fiscal year 2026 with BiCS8 as the dominant technology node; it is collaborating with hyperscale customers to certify new products, including a 128TB QLC enterprise SSD named "Stargate," which has not yet begun revenue shipments but has entered the certification phase; the company is also developing high-bandwidth flash memory technology in collaboration with SK Hynix for AI inference workloads, reportedly with a density ten times higher than alternatives, with prototype chips expected to be launched by the end of 2026 for customer testing about a year later. Additionally, SanDisk has extended its joint venture agreement with Kioxia for five years, ensuring manufacturing capacity continues until 2034, allowing the company to invest in R&D at a level comparable to the highest market share competitors while maintaining cost leadership.

5. Supply-Side Discipline Reshaping the Industry Landscape

Goeckeler explicitly stated in his speech that the current tightness in the NAND market is a structural change driven by real demand, rather than a temporary cyclical fluctuation. This judgment is the fundamental basis for a series of strategic adjustments by SanDisk. AI-driven storage demand is reshaping the NAND value chain, pushing demand significantly beyond supply, and this imbalance is expected to persist beyond the 2026 calendar year. Additionally, other key data supports this conclusion—there will be no new NAND wafer capacity increases between 2026 and 2027, while demand is expected to expand by over 20% in 2026. This ongoing supply-demand mismatch is driving NAND pricing to unprecedented levels.

Notably, the price of 128Gb MLC NAND surged from $2.18 at the beginning of 2025 to $17.73 in March 2026, with second-quarter NAND flash contract prices rising approximately 70% to 75%, surpassing DRAM's increase of 58% to 63% during the same period. Against this backdrop, Goeckeler emphasized in his speech that the company's strategic focus has shifted from pursuing shipment volume to creating sustainable profits, actively ceding the low-margin consumer market and fully focusing on high-end enterprise SSDs for data centers.

II. Interpretation of Deep Strategic Intent

1. Return to Strategic Positioning Post-Divestiture

To understand the strategic significance of this speech, it is essential to review SanDisk's recent history. SanDisk officially spun off from Western Digital on February 24, 2025, under pressure from activist investors like Elliott Management—the core logic was to eliminate the "conglomerate discount" that previously obscured the potential of the Flash business due to the HDD business, allowing both to become purer targets in the AI super cycle. Goeckeler himself was a key figure in leading the SanDisk spin-off and subsequently serving as its independent chairman and CEO, and his articulation of the company's future strategic direction naturally carries the important mission of laying the foundation for SanDisk's positioning post-independence and reshaping market expectations.

2. Construction of the Narrative of NAND "De-Cyclicalization"

The most controversial yet core strategic narrative in Goeckeler's speech is his judgment that NAND is shedding its traditional cyclical attributes, becoming an industry that is "more resilient, structurally more attractive, and with higher long-term average returns." Management also stated in the speech that NAND is being redefined as "a key component of AI infrastructure," rather than a traditional cyclical commodity.

This narrative has sparked a strong reaction in the capital markets. The market has elevated SanDisk's year-to-date return to over 356% based on the logic that "NAND has achieved de-cyclicalization through long-term contracts with hyperscale cloud providers," with the stock price soaring from approximately $33 per share at the time of the spin-off to $1406 (as of early May 2026). Evercore ISI initiated coverage of SanDisk with an outperform rating, setting a bull market target price of $2600; Bernstein itself provided an extreme target price of $3000, raising its price target from $1250 to $1700.

However, the authenticity of this "de-cyclicalization narrative" still faces considerable challenges. The market generally believes that SanDisk's non-GAAP P/E ratio has reached 40.36, significantly higher than Micron's 26.30 level, indicating a substantial valuation premium. More importantly, the core flaw of these long-term agreements is that a portion of the long-term contracts employs a floating pricing mechanism—this means that while approximately $42 billion in financial guarantees (RPO) can prevent customer defaults, it cannot guard against price declines. Once Samsung and SK Hynix's new capacities (such as 321-layer NAND) are released in 2026 to 2027, floating pricing will be re-priced along with the spot market, and the sustainability of long-term demand has yet to be ultimately verified.

III. Financial and Performance Background

The explosive growth of SanDisk's recent financial data is a key background for understanding the confidence and capital tone of the CEO's speech. The company's revenue for the third quarter of fiscal year 2026 (ending March 2026) reached $5.95 billion, a year-on-year increase of 252.1%, with non-GAAP earnings per share soaring to $23.41, significantly exceeding market expectations. However, what should be more focused on is that this performance exceeding expectations is primarily driven by pricing power and product portfolio optimization, rather than growth in shipment volume—bit shipment volume was actually flat year-on-year and down quarter-on-quarter.

Even more astonishing is the non-GAAP gross margin reaching 78.4%, far exceeding the historical average level of 30% to 40% in the industry, marking one of the steepest climbs in the history of the semiconductor industry. However, maintaining this unprecedented gross margin level almost entirely relies on the data center business's year-on-year surge of 645%, with management explicitly mentioning in the quarterly earnings call that "we cannot meet all of our customers' demands." The narrative of supply shortages conveyed in the speech is highly consistent with these astonishing financial figures.

IV. Far-Reaching Impact on the Storage Industry Landscape

1. Industry Model Shifts from "Price Wars" to "Supply Discipline"

SanDisk's supply discipline—refusing blind expansion and reconstructing customer relationships through LTA—is reshaping the competitive paradigm of the entire NAND industry. If this strategy is adopted and emulated by other major NAND manufacturers (including Samsung, Micron, SK Hynix, etc.), it could mean that NAND will embark on a new growth curve characterized by stable demand and controllable capacity, contrasting sharply with the severe cyclical fluctuations that have consistently marked NAND since its inception.

2. Industry Power Shifts to Suppliers

Against the backdrop of expanding AI infrastructure, NAND's strategic position as the foundation for AI storage has significantly elevated, and the bargaining power between suppliers and hyperscale customers has shifted. Customers are willing to provide demand forecasts extending to 2028 and supply commitments lasting up to five years, reflecting that this structural change is real. The most core industry signal conveyed by Goeckeler's speech is this: storage is no longer just a cost item but a key strategic component of AI infrastructure.

V. Outlook

From a positive perspective, Goeckeler has indeed captured a real, AI-driven structural demand shift and decisively transformed SanDisk from a diversified storage company into a pure-play company highly focused on AI storage needs. The spin-off itself, the signing of long-term contracts, the explosive growth of the data center business, and the clarity of the technology roadmap—these are all objective changes occurring in reality, rather than mere conceptual hype.

Overall, the core value of Goeckeler's speech lies in its provision of a highly transparent and logically clear narrative framework for corporate transformation. However, for investors and industry observers, distinguishing between the genuine business transformation narratives in the speech (the rise of data centers, supply discipline, LTA reshaping) and the yet-to-be-validated valuation assumptions ("de-cyclicalization," sustainability of 78% gross margin) remains a key prerequisite for making rational judgments.

Sandisk Corporation (SNDK) Bernstein 42nd Annual Strategic Decision Conference

Company Participants

David V. Goeckeler - Chairman and CEO

Conference Call Participants

Mark Newman - Bernstein Institutional Services LLC, Research Division

Presentation

Mark Newman, Bernstein Institutional Services LLC, Research Division

Hello everyone, good afternoon. I’m Mark Newman, the U.S. IT hardware analyst at Bernstein. I’m pleased to welcome David Goeckeler, the Chairman and CEO of SanDisk, who previously served as CEO of Western Digital and led the spin-off of SanDisk. Thank you for being here again today, David.

David V. Goeckeler Chairman and CEO

I’m glad to be here, Mark. Thank you for having us.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Thank you.

David V. Goeckeler Chairman and CEO

Can I start with a safe harbor? It seems I must.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Sure, go ahead.

David V. Goeckeler Chairman and CEO

Clearly, only I can do this. In today’s discussion, I will make forward-looking statements based on management's current assumptions and expectations, including about our technology and product portfolio, our business plans and performance, our capital allocation priorities, market trends and opportunities, and our future financial performance.

These forward-looking statements are subject to risks and uncertainties. We undertake no obligation to update these statements. For more information about risks and uncertainties that could cause actual results to differ materially from expectations, please refer to our annual report on Form 10-K, quarterly reports on Form 10-Q, and other documents filed with the SEC.

We will also mention non-GAAP financial data, and reconciliations between GAAP and non-GAAP performance can be found in the investor relations section of our website.

Question-and-Answer Session

Mark Newman, Bernstein Institutional Services LLC, Research Division

Thank you very much. Now that we’ve covered that, I’ll start the Q&A session. I have a bunch of questions to ask. Just a reminder, you should all have a Pigeonhole link where you can submit your own questions at any time. I have an iPad here, and I’ll check the incoming questions. If possible, after I’ve asked some of my own questions, I’ll try to pick a few audience questions to ask.

I want to start with demand. If you could first talk about the demand situation, then we can discuss other matters. First, regarding demand, given the current development of AI and other areas like mobile and consumer, can you describe the demand environment you are seeing now? What changes have occurred compared to when we last discussed demand, especially in terms of AI?

David V. Goeckeler Chairman and CEO

First, I want to say that this is one of the reasons I really love this market and this business: there are many demand drivers. I mean, NAND is used in almost all interesting technologies in the world. Traditional smart markets, like smartphones, personal computers, data centers (which are obviously growing significantly), and it also extends to IoT devices, automotive, and robotics.

It just continues like this. It’s a very, very diverse market with many demand drivers. These drivers change at different speeds. I think that really makes this a very, very interesting place for a company like ours to grow its business. So what’s happening now? I think it’s no secret. Data centers are indeed growing very, very rapidly.

If we look back to about three forecast cycles ago, we initially thought that growth in data centers this year would be in the mid-20% range. Then we raised the forecast to the mid-40% range, then to the mid-60% range, and now we’ve slightly raised it again. If we look at how much data centers will grow in the 2026 calendar year.

This is indeed happening. Clearly, it is a major driver of the market, and a lot is happening. But other markets remain—demand across various markets is still strong. Whether it’s PCs or smartphones, we are still maintaining good communication with these customers, including those in the automotive and IoT sectors. I think the overall demand environment is very robust.

Mark Newman, Bernstein Institutional Services LLC, Research Division

I mean, given that AI is so strong, with data center growth exceeding 60%, other parts of the market are being squeezed out a bit. That might be part of what we’re seeing. How would you articulate that?

David V. Goeckeler Chairman and CEO

You see, I mean, it’s a market, right? I think markets always rationalize supply and demand. They always stay in balance to some extent. Clearly, some will achieve clearance through pricing. And in this market, a lot of very, very attractive demand environments are forming. This is very, very exciting.

This will always affect other parts of the market that may not be as attractive from an economic perspective. This situation occurs in any market at any time. We just happen to be in—this is a big market. It’s a very liquid market. We always know the prices.

In fact, this market has been accustomed to constantly engaging in pricing transactions, even in the contracted portion of the market, where prices are traditionally set every quarter, right? This indeed brings a lot of volatility. This is one of the things we frankly want to get away from. I think in any market that continually creates a lot of new, attractive addressable markets (TAM), you will see this dynamic.

You see, we talked about this a bit when we came in. Just a little while ago, I was in New York, and when we completed the spin-off and launched this company, I said we would invest to achieve mid-to-high teens growth.

We believed our view was—that was early 2025. We expected that by the end of 2025, market pricing would show an upward inflection. And until last summer, the mainstream view was that this was the wrong view.

I saw reports later (say, late summer) saying, oh, SanDisk would not meet its performance in December because prices would fall. Things did not develop entirely as expected. I think no one could have anticipated the demand from data centers would be so fierce.

But we believe this has always been a very good market. We have been investing for the growth of this market. We have to make investment decisions many, many years in advance of actual supply appearing. We are investing heavily: capital expenditures in the billions, investing hundreds of millions to enhance R&D efficiency. By the way, regarding R&D efficiency and NAND, that in itself is a complete topic.

It’s spectacular. With each node delivered, we can achieve significant growth through additional capacity. But we have long confidently believed we can expand the market size, and we are committed to letting the market grow at a mid-to-high teens rate.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Back to demand. Besides the major numbers announced, what leading indicators are you looking at, such as order book depth, customer forecast revisions, qualification activities? What are you looking at to most convincingly assure you that this current demand cycle has durability? Because clearly, given the current pricing trends, demand far exceeds supply.

David V. Goeckeler Chairman and CEO

Yes. What we want to see is what that environment will look like in the coming period—my point is, for example, a few months ago we invested $1 billion to extend our agreement with Kioxia from 2030 to 2034; that’s a fantastic agreement. So clearly, we start thinking about how demand will look far into the future.

To answer your question, we have many different methods. First, we do a lot of bottom-up research, communicating with customers to understand what they are doing. For example, in various markets like smartphones, personal computers, etc. We have close relationships with customers and know what devices they want to launch in the future.

We have our judgments about how the composition will look. We have conducted various bottom-up analyses of the major markets to assess what will drive bit growth. We are clearly watching capital expenditures. I mean, capital expenditures are rising every earnings cycle, which is precisely why data center-related numbers are increasing.

We know we have a fairly clear understanding of how this relates to our growth in technology. So these are the medium- to long-term issues we are focusing on on the demand side. And then we are out in the market every day. I mean, we are in dialogue with customers. They call us to talk about what they need now and in the future—clearly, we are also constantly discussing pricing.

So all of this convergence allows us to gain insight into where the market is heading.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Specifically regarding AI, we have gone through different phases of AI training: the early impacts driven by chatbots, followed by impacts from more advanced stages, and now we are entering the agent era. How do you think these changes will affect NAND demand over time as we move from early to later stages?

David V. Goeckeler Chairman and CEO

We have always believed that inference is where NAND truly shines. So we have to achieve that. I want to say that in the first two or three years of AI, I was often asked this question: Will the development and deployment of AI affect the NAND business? The answer has always been yes, it will, but we have to achieve that first, right?

You have to build the models, deploy them, promote them, and get users to start using the technology. There must be valuable use cases that drive usage. I think we have passed those stages. Now we are rapidly advancing all of this, and over the past year, you have begun to see its impact.

That has indeed been the case. You are starting to see the impact of NAND on inference architectures. I think as our customers begin to build these architectures, you will be thinking about how to scale inference globally, right?

In training, you don’t need to achieve global scaling. You are training a lot of very smart people and a vast amount of infrastructure supporting the training. But in inference, you want to scale it to the masses, if you will. Billions of people will use inference to some extent.

So when you go through that process and scale something like that globally, you must have economics. Right? In any early stage of technology, you naturally want to—when you as a technologist first do these things or build a market, you often use all the resources you might need, almost filling your architecture to the brim.

Give me all the compute power, give me all the memory, give me all the power, give me all the network resources, give me everything I need, and then I want to build a system. But when you want to scale that system, you need to really dig deep into what exactly you need to scale, and that must be economically viable.

Because if it’s not economically viable, it will clearly be too expensive. That means you have to raise prices. That means you will give others the opportunity to come in and do it more economically, thereby pushing you out of the market. So these decisions are very, very significant and very, very hard to make.

I think what has happened over the past year is that the people responsible for doing this—those amazing tech companies—have the rich expertise to scale technology globally. I think that has indeed been the story of the past 20 years.

I mean, the distribution of technology has become almost completely frictionless, right? You just point to a URL on a device and get the most amazing technology in the world. It wasn’t like that ten or twenty years ago when we had to ship you things, or you had to upgrade software, and there were all sorts of frictions in the system.

All those frictions have been eliminated, which means we can deploy technology at scale very quickly, which is astonishing, isn’t it? We are witnessing all of this happening firsthand. But the people responsible for this work face a very, very daunting task because the costs are extremely high, and you must do it in the most economical way.

So those people have been going through a process: How do I build that architecture? And this is where NAND gradually begins to come into view. Why? NAND is highly scalable. It is the most scalable semiconductor technology in the world. We can provide supply, right? Therefore, as models get larger, context lengths get longer, all these factors are driving you—if you want to do this economically, you must use more scalable technology. Or even if you just want to do it, there isn’t enough of anything else, right? DRAM is excellent technology. HBM is excellent technology. It has incredible properties.

Its scale is still insufficient to solve global inference problems. So I think companies have been figuring out what that architecture is and starting to scale it and study how we will scale it. That is the reason behind the increasing demand for NAND.

I find that when I study that architecture and precisely tune the goals I want to build, my demand for NAND increases or decreases, and the conclusion is that we need more. So this is driving the market and prompting those customers to come to us and say, "Hey, look, we are planning for the next few years—this is our business."

We want to understand your plans for supplying us with this critical technology years down the line. We don’t want to come every quarter to talk about prices and figure out if there is enough supply. We need to know now: Can you supply me in 2028? Can you supply me in 2029? That is what is driving us through this comprehensive transformation.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Thank you. We are seeing incredible demand right now. Jensen Huang articulated this KV cache vision at CES earlier this year, which is roughly equivalent to adding about 17 TB per GPU. Have you seen this? Is it reflected in your demand data? Do you think this will have a significant additional impact on NAND demand?

David V. Goeckeler Chairman and CEO

Yes. I mean, that’s exactly the process I just described—people are designing systems and configuring systems. KV cache is migrating to NAND because it has to scale. I need scalable storage technology, and that is NAND. Depending on the use case you are building, like I know people want a very clear "if I do this much of this, I will get that much of that" measurement, but it’s not that simple.

For example, you need to figure out what use case you want to build for and what scale to expand to; once you figure that out, you can design the architecture to achieve it. As you go through this process, there will be many variables. How big is the model you are using, how many tokens are there, what is the size of the KV cache? Or do you have a cache somewhere, and what is the hit rate of that cache?

You have to go through this very complex calculation, and we have done some work on this and shared it. Ultimately, you will arrive at how much NAND you will need, and then you will come to us or our peers and ask how to obtain that much NAND in the coming years. So we are very confident in this vision.

I think this is far more than just a vision. This is what is happening in reality, and it has been going on for quite some time because companies need to bring this amazing AI technology to the ground and scale it so that all of us can use it.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Yes. I think you mean this is about density, right? NAND flash and DRAM density. In terms of how many GB you can get per dollar and how many GB can fit per unit area, NAND is far superior.

David V. Goeckeler Chairman and CEO

Yes. Just—we can provide more—we can provide more capacity, yes, higher density. Absolutely correct. It’s a different technology. It addresses different use cases. It’s not a substitute, nor does it mean one is better than the other. That’s not the point at all. You need both.

And you have to incorporate this highly scalable storage technology as part of that architecture, which is also why the number of data centers is continuously rising: as people repeatedly explore how to define that architecture and how to scale it, the scale requirements keep increasing, and the numbers keep climbing, thus pushing up demand.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Now slightly shifting to pricing, focusing on ASPs rather than LTAs, just looking at the industry pricing environment. For those not closely following SanDisk, last quarter, SanDisk's average selling price per GB increased by about 140% quarter-on-quarter.

But that’s my estimate. I don’t think you’ve really given that exact number, but it’s roughly around there, which is astonishing. My question is, how would you describe the current pricing environment? I mean, clearly, it’s not sustainable to achieve 100% quarter-on-quarter growth; that’s unsustainable. But do you still see strength continuing in different segments? Are prices still on an upward trend? Or how do you view it?

David V. Goeckeler Chairman and CEO

Look, I mean, we have already made predictions about our own forecasts. I’m not going to discuss what future pricing will be. Look, I mean, what we want to do is—what’s most important in our business is to create very valuable technology. It all starts with technology; it’s always about technology.

If you find yourself creating excellent products that can solve real needs, then we are on this journey of exploring the value of these technologies. And our job is to do that, and we will continue to do so.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Got it. Well, the surprise is a bit strong. Well, got it. Historically, NAND pricing has been quite cyclical, and we will talk a bit about LTA later. Besides LTA, what gives you confidence in the sustainability of the current price levels, as we will talk about LTA next?

David V. Goeckeler Chairman and CEO

I mean, I think this really is... I mean, this is largely how I view my work: making everything sustainable. I feel this cyclicality is— I mean, I have used this term many times; it’s incredibly corrosive. We are either in a situation that looks like we are desperately trying to survive on the supply side.

I was in that situation in '23 as well. A year ago when we founded the company, the valuations people were giving us—I felt they were very low, and later it proved to be true. Now we are in the situation we just discussed, where you asked me about that situation where everyone is not getting what they want.

To me, that indicates our incentives are misaligned. Our business model is also misaligned. So I think what I am doing, what our team is doing, if there are about three big things we need to do, it’s in this technology business—actually in any technology business.

This is what I am constantly trying to balance. First, you must achieve a fair return for what you create, right? We are very proud of our technology, but it’s hard to do. We are not only investing a lot of intellectual property (IP) in building NAND, but also a lot of IP in building systems. We don’t just have one R&D team; we have two: one for NAND and another for SSD and all product development and manufacturing—these are things we are doing as well.

We have to invest all the capital expenditures to manufacture. The backend? Yes, we also do the backend. We actually have a backend, so we do the whole process. Clearly, we have many excellent suppliers providing us with many important technologies that enable us to do this. But most importantly, we need to achieve a fair return on the investments we have made for a long, long time.

So that’s the first thing. I would say we are doing okay in that regard now. Frankly, for a long time, we were not doing well in that regard. I mean, just go back a year ago; people were basically telling us we were not doing well enough in that regard because they didn’t want to invest in your company.

So that’s the first thing you have to do. The second thing I am really concerned about is that we need to do something about the cyclicality, right? It’s too harmful because everyone is waiting for when the recession will come. You have a good quarter, oh, you are just one quarter closer to a bad quarter; this kind of crazy psychology. People either don’t get what they need or have too much. To me, that’s not helpful at all.

So we want to do this, and we are achieving it through business practices. That’s why we call these new business models. How do we change the way we interact with customers? Then, in any tech company, the third thing you need to do is achieve growth, right?

You understand the right economics. You strip away the cyclical factors or deal with them differently, and then you must grow. And in every tech business I have managed in my career, the third one is always the hardest. Growth is hard, right? I mean—especially in those large, profitable businesses, growth is hard.

But we have handled that one, right? We said our growth would be in the mid-to-high teens, and people would say, can’t you grow faster? I would say, let me solve the first two goals first, and then we can start talking about this. So balancing this equation is very difficult.

If you start changing—you can talk about one of the three separately, but you must discuss all three together because if you start adjusting one, the other will change in the opposite direction. So we have been working hard to balance this entire equation and focus on all three.

I would say the hardest to solve is the growth issue, which is a huge advantage for us, right? People always love to debate: should you be growing faster? Well, maybe we could grow faster, but that would come at the expense of economic efficiency, which is not a good trade-off from a valuation perspective.

Should we focus more on economics, pursue that more, and accept greater cyclicality? That doesn’t seem like a good trade-off. So you have to do all three. We have been—this is how we think. At least that’s my thought. This is also what we are trying to balance. It’s really interesting.

I think in this environment, our business is undergoing very significant changes that can truly address those first two issues.

Mark Newman, Bernstein Institutional Services LLC, Research Division

This is very helpful. Then further delving into the second point, which is long-term agreements—the new business model you mentioned. Can you now talk about what these agreements look like in terms of sustainability, quantity commitments, and pricing structures?

If you could explain what you can do, like how you view those agreements and how your current progress is.

I know you mentioned in the last conference call that over a third of the volume in these long-term agreements you refer to comes from this. What percentage do you expect that to reach? It would be helpful if you could elaborate on that.

David V. Goeckeler Chairman and CEO

Yes. So let’s talk about—you just mentioned a bit, and I think everyone understands. I mean, this has traditionally been a very volatile business, right? In fact, prices change almost every quarter. It’s a business that’s hard to plan for and hard to predict.

And traditionally, there have been some situations regarding agreements—by the way, we have very remarkable customers. I mean—this is again one of the reasons this business is so attractive. Our customers are some of the most enviable companies in the world. Really, they do incredibly well, whether in personal computers, smartphones, data centers, or any other field; overall, what our customers do is incredible.

But the traditional view of long-term agreements is that I will first commit to a supply volume, and then we will discuss pricing. That is, it’s better than nothing, right? At least we understand that if a price agreement is reached, we know how much volume to allocate to everyone.

But we want to get away from that volatility. So how do we think about this differently? That’s why we—the idea of "long-term"—the industry has thrown out many terms: long-term agreements, NCNRs, take-or-pay, and various other things.

And in my experience over the past two years, as soon as you mention any of those terms, the person sitting across from you will immediately start telling you all the reasons why they don’t work. So we have done in-depth research on this.

We said, listen, what we want to do is align our business model with our customers’ business models, right? And more and more customers are coming to us, especially after we have certified these data center qualifications. Think about our data center business that we have been developing; we are doing enterprise-grade SSDs. This takes several years; it’s a very arduous process. Then you start engaging with customers, which can take two years, to help them understand what you are doing, give them samples, put thousands of devices in labs, and let them run for a year to complete certification; it’s a very, very difficult process.

So at the end of that process, there will come a stage where customers will say, well, you have made a great product, right? We have invested a lot in this. We have made a great product. I want to buy it. Great, right? And I want to buy long-term. I want to keep buying for the next five years because my—back to what we said at the beginning. I am doing all this work, building this new technology.

I have a huge demand for your products. I don’t just want to buy a bit this quarter. I want you to tell me that you can supply me for the next five years. So, okay, then tell me what your demand is, and then we can start talking. This is actually a bit new, right? Because it’s usually like this: I will tell you my demand for the next 12 months, and then we will talk about prices four times a year.

Now it’s like this, no, no, no, no. I don’t want to just talk about these 12 months. I need to know if I can get what I need from you two, three, four years down the line? Because when I am building, what they are building is spectacular; it’s truly incredible technology. So we will start discussing how we align the business models, right? You want to consume NAND, and I want to produce NAND.

Now the way I produce NAND happens to be a business model that may be very different from yours, right? I have to invest ten years in advance. I want to build a massive fab that can be seen from space. It’s enormous. I also have to plan my capacity several years in advance.

The good news is I have already done that. We have fabs. We have R&D. We know what the technology roadmap looks like for the next few years. But now I have completed all that, and the fab is operational, right? Now I am investing for growth, right? I want to achieve mid-to-high teens growth. So my fab is running, and the number of wafers tomorrow will be more than yesterday. Every day is like this; wafers come out of the fab.

And I have to sell them. I can’t put them in inventory, and I can’t let them fall to the ground; someone has to take them. For ordinary consumers, this business model is a bit unnatural. They also have large businesses that are growing, but guys, do I have to buy something every month? Do I have to buy more than last month? The answer is yes.

So how do we reach an agreement—you will need this supply, and I will provide this supply. How do we adjust our business models so that I am confident you will be my strategic partner while you are also confident that I will deliver to you? How do we write this into the contract? This is why we came up with these new business models. So how did we think about this? First, we need partners that will consume a lot of products, right? Because this will be a significant contractual arrangement. Second, we need your demand growth rate to be at least as fast as our supply, preferably faster than our supply growth.

So if I am investing for mid-to-high teens growth, and you come to me saying you want the same amount for four consecutive years, that doesn’t really help me, right? The consumption rate you need must be faster than what I (and that strategic partner who is very important to me) can provide.

Next, you need to have predictable consumption. Remember, the fab is running every day, and wafers are constantly coming out. If you are my strategic partner, you need to maintain predictability in demand weekly, um—let’s say monthly, quarterly.

The more you can let me understand that demand, the better it is for us. What will your product mix look like? How much will this product and that product each account for? We need to clarify all of this. Then we also need to set up incentive mechanisms because you are a public company, and I am a public company. Some things may happen that force you to exit this contract.

I understand, right? Things will always happen. When black swan events occur, the entire economy will fluctuate. For example, we encounter a global pandemic. Suppose an event occurs that could affect the entire world. So at that moment, I need an incentive mechanism that gives you the motivation to continue fulfilling the contract.

If you don’t fulfill the contract, I will be compensated, okay? So I want you to set aside some money in advance; we will have a third party hold it for us, right? We won’t—I won’t—we won’t argue. I won’t sue you. That absolutely won’t happen, right?

So—you wouldn’t sue your customers; we are partners, right? If something happens, you have to exit the contract. So beforehand, let a third party hold some of the money you have—if you want to use that term. In short, some third party will hold some funds. But the simplest way is for you to give me all the money upfront.

That’s a bit unrealistic, right? I mean, this is a five-year relationship; you won’t— that’s a big check for anyone. For various reasons, that’s not realistic. So we have to think of other ways. Let the third party hold that money; they will hold the contract and can determine whether you are in breach or not. If a breach occurs, the third party will give the money to me. That will appear on my balance sheet, and we can part ways amicably.

At that moment, the contract ends. From then on, I keep everything, and you keep everything you paid, and then we each go our separate ways, right? So we think this will align our incentive mechanisms. Now you have the motivation to continue fulfilling the contract. You might think, oh my gosh, I want to exit this contract. Do you really want to exit? You will have to give up some money, possibly billions. So you better think it through. If you really have to exit, at least I can have a buffer. I can get some cash, which is very helpful in situations like black swan events. For example, I want to eliminate cyclical fluctuations; what do I need to get through a significant downturn? I need cash.

This way, we protect ourselves, and both of us—can continue moving forward, all is well, and we can do business again sometime in the future. So this is the general idea behind the contract structure we have set.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Is that a $12 billion funding commitment?

David V. Goeckeler Chairman and CEO

Yes. So let’s break down those numbers. We have discussed this. We now have RPOs. This is something you would consider. I have managed many software companies, right? So that’s an indicator that comes from there. But that is indeed an accounting metric, right? We didn’t suddenly wake up and say, "Oh, we need to use this metric," but when there are contracts in the industry and future obligations exist, people will do this.

We signed three contracts at the end of the quarter. The approximately $40 billion, about $42 billion figure we mentioned refers to the minimum purchase obligations of these three contracts over their entire term.

Then we also talked about another number, which is a bit different, so it’s a bit complex. We signed two contracts after the end of this quarter, so they are not included in our data. But among the five contracts we signed, the amount reserved in case someone exits totals $11 billion. Roughly, this is how these two numbers relate to each other.

But to be honest, we don’t expect to see that money at all. I don’t want to see that money. I think we have great partners. I believe these contracts will be fulfilled. I think our interests are aligned, and everything will be fine.

But we live in the real world. There must be some incentive mechanism, and I think we have aligned those incentives. I also believe we have willing partners who want to walk this path with us because they value supply commitments.

Mark Newman, Bernstein Institutional Services LLC, Research Division

You mentioned there are five customers; are those all hyperscale cloud service providers, or not?

David V. Goeckeler Chairman and CEO

No, we didn’t say that. We wouldn’t say that. What we want is—just like you did smoothly earlier—I mean, you see, we want customer diversification, right? What we want and how we treat... I have mentioned the portfolio several times; it’s the same.

We want a diversified product portfolio with a lot of optionality in the products we sell. That’s where I started. Why do I like the NAND market? I like the NAND market for many reasons—it's a very diverse market. There are many excellent customers, and you can sell your products in many places. But you must have the corresponding technology to do that. You can’t just sell raw wafers; you have to turn them into products. If you want to enter the consumer market, you need a team to build those products.

You need a backend to create all those things. So you want to make the portfolio as diverse as possible to have as much optionality as possible. The same goes for these new business models. We want arrangements of various durations, right?

You don’t want them all to expire on the same day. So you want some to be one-year terms, some to be three-year terms, some to be five-year terms, and then you also need diversified customers, preferably covering as much of your portfolio as possible because that keeps the portfolio vibrant and maintains that optionality. So we have taken the first step, right?

That’s what we announced in the earnings call. Again, back to the three things I mentioned: achieving fair returns, addressing cyclicality, and achieving growth. Fair returns. I think we can still improve, but overall, it’s good. Now we have five items in the middle column starting to address this issue. Over a third of the portfolio has visibility, rather than visibility that lasts only three months or possibly just twelve months.

Now we are talking about visibility—spans like 2 years, 3 years, 5 years; that makes a big difference. As for growth, remember, growth has always been there. That item has always been checked, right? That has always been the most important reason this market is so excellent: it will grow.

So we have checked that growth item. The first item is achieved, and we are in a favorable position. Now what we need to do is maintain that, which is the second part. That’s why we call it a new business model because it’s a different business model to achieve that.

Mark Newman, Bernstein Institutional Services LLC, Research Division

So that 33%, do you hope to raise it to 50%, 60%, 70%, or do you think that’s unrealistic?

David V. Goeckeler Chairman and CEO

No, that’s not unrealistic. It’s still pending, right? Again, this is something we haven’t completed. We have taken the first step—perhaps I should say we have taken the first five steps; maybe that’s more appropriate. But we are still discussing, and it depends on the portfolio thing I mentioned earlier.

Look, as I said, we have many—we have truly remarkable customers, really remarkable customers. They are great companies, great people. They develop incredible technology. Some of them like our previous business model. They like to do it quarterly, hey, we just—let’s do it quarterly. Okay, fine. We are very accepting of that. We have no problem with that. We know how to do it. If that’s what they want, we will do our best. So we look forward to...

Mark Newman, Bernstein Institutional Services LLC, Research Division

But if those customers don’t sign up... will they get enough supply?

David V. Goeckeler Chairman and CEO

I can’t run their business for them. They have to run it themselves, right? I’m not the only supplier in the market. But what I want to do is get a set of agreements that bring me the kind of diversity I mentioned, covering a significant portion of my portfolio. It doesn’t need to cover everything. It will never reach 100% because there are many customers outside who are too small, though they are good customers and valuable businesses, but not enough to participate in this collaboration.

We look forward to it. I feel there are still some uncertainties, which may be a bit unsatisfactory for you right now. The final landing point is still to be determined. But I think if there’s an opportunity, we will continue to push it higher.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Great. I mean, this is very clear, much clearer than what we have gotten from your competitors regarding long-term agreements so far. So I really appreciate that.

David V. Goeckeler Chairman and CEO

I love mine. Those companies are great. Indeed.

Mark Newman, Bernstein Institutional Services LLC, Research Division

By the way, talking about supply and capacity issues, given that prices and demand are very strong right now. Many of your competitors, like Samsung, Hynix, Micron, have no room to increase capacity because they are using all their fab capacity for DRAM, and DRAM capacity is also tight.

But Sandisk and Kioxia together, you are one of the few companies that actually still have room to increase capacity. I’m not encouraging you to do that. I’m just asking. I was just asking.

David V. Goeckeler Chairman and CEO

Are you asking for a friend? Is that how it is?

Mark Newman, Bernstein Institutional Services LLC, Research Division

How do you view this issue? What I want to ask is, considering you might increase capacity while almost no one else can do that (possibly only China’s YMTC), can you increase more capacity? Or are you just looking to optimize pricing right now?

David V. Goeckeler Chairman and CEO

Well, it might be a bit more complex than you think. First, we have been increasing capacity. I think we should start from that point. We have been expanding—remember, we are growing, and the growth rate is in the mid-to-high teens. This market is very large, and the transaction volume is growing at a mid-to-high teens rate. First, that’s truly remarkable, right?

The second point. So under normal circumstances, we would increase capacity. That’s it. That’s the business we decided to enter. We have to make decisions a long time in advance. For example, the demand for next quarter does not affect my capacity decisions.

I have to make that decision three years ago. Our fab planning is aimed at many years in the future. Moving equipment between different nodes is very complex—inside a fab, you are not just running one node; you are running many, many different nodes at the same time, and the transition is very complex.

So you have to make those decisions a long time in advance. Also, I don’t want to dwell too much on this, but think back to about 12 months ago; everyone was telling me we were investing too much, right? And we said, no, no, no. We believe mid-to-high teens growth is the correct number, but some people said, well, based on pricing, that’s not the correct number.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Less than 12 months ago.

David V. Goeckeler Chairman and CEO

Yes, less than 12 months ago, right? So we can’t afford such large back-and-forth fluctuations. How do I think about it? My thought is that we are investing at a mid-to-high teens growth rate, and we have a great partner in Kioxia, and the relationship is very good.

This has been going on for a long time, and there are good reasons for it because it’s very productive and very valuable. We are good at planning. We are good at planning and ensuring we have everything needed to continue growing the business at the right time.

Now there is one thing I want to say that is very important and must be understood. We can achieve growth through process node transitions. That is what I call R&D productivity. The number of bits per wafer continues to rise at a compound growth rate faster than the mid-to-high teens rate I mentioned.

So if we just advance through each process node with the same number of wafers, it would create a market oversupply. So we are constantly adjusting this equation. Remember, each process node requires more cleanroom space. Each node is more complex, with more steps, more equipment, and requires more cleanroom space.

But this dynamic is extremely important, right—if you look at our capital expenditures (CapEx) as a percentage of revenue, it continues to decline as revenue rises because our R&D productivity remains very high. So for the investors present, this is very, very important.

It means I can achieve growth without investing a lot of new capital expenditures. Back to my earlier model—yes, economically eliminate cyclicality and expand those things you will eventually discover. Ultimately, why do we do this business? It’s to generate free cash flow.

You will find that this business is very good at this because our investments in capital expenditures are very efficient and can yield more incremental output.

Mark Newman, Bernstein Institutional Services LLC, Research Division

I received some questions from the audience. Before I go to the audience questions, can I ask one more? HBF, high-bandwidth flash. Can you talk about what the latest developments are regarding it?

David V. Goeckeler Chairman and CEO

Since we founded the company and announced it last February, we have been very excited about this technology. We have long believed that once we enter the inference stage, NAND will become a very important technology.

You don’t need to convince us that memory architecture needs to change to make inference scalable. That is essentially part of HBF. It doesn’t mean HBF will replace enterprise-grade SSDs, nor does it mean HBF will become a substitute for DRAM, and so on.

This indicates that as inference scales, there are huge opportunities for innovation. For those with new ideas, like I see AI and the large-scale expansion happening now, I see this as a huge green light for innovation. If you have new ideas, bring them out, right?

Because the whole world is figuring out how to scale this amazing technology. And the faster we can do this—as I said earlier—the more frictionless we can scale this technology.

It’s astonishing how quickly technology can be made available to everyone if it’s economically viable. HBF is a strategy aimed at providing high density for inference, which is primarily a read-dominant and deterministic read activity. So we are very excited about this technology. It’s brand new. We are currently manufacturing NAND chips, and we expect to have them by the end of this year. At some point next year, we will have a complete system. We are building controllers on it, and there is still a lot of work to do. We are working with customers to study how they will integrate it into their architectures, right? Because it’s not plug-and-play.

It’s not that we just plug in components, and you pull out other things; it’s a systemic solution. So you have to get your customers to adopt it into what they are building, and we are going through that process. We will keep updating as we progress.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Okay, sounds great. Now for the audience questions, the first question is: Will the shift to edge or device-side computing be a boon or a risk for Sandisk's growth forecasts?

David V. Goeckeler Chairman and CEO

No, I think—I mean, anywhere you have a shift—you say AI...

Mark Newman, Bernstein Institutional Services LLC, Research Division

So, AI basically, AI and edge means on your devices, like your smartphones and personal computers.

David V. Goeckeler Chairman and CEO

Well, I mean, this is just the same theme: NAND is everywhere. As you start to scale technology, you will find you need more capacity, and we have scalable technology.

We see this as very important—that’s why even last year we maintained that mid-to-high teens growth rate because this market has a perennial characteristic.

Just like the world is constantly innovating, coming up with new ways to use our technology, that’s a wonderful thing.

Mark Newman, Bernstein Institutional Services LLC, Research Division

I have a good question from the audience: Can I ask the audience present to raise their hands if you hold Sandisk stock? Can I ask you to raise your hands if you hold Sandisk stock?

Good. Okay.

David V. Goeckeler Chairman and CEO

Thank you. We are working very hard for you.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Next question from the audience: Can you elaborate on the lessons learned from past booms and busts? Have management incentives across the industry changed to avoid repeating the mistakes, similar to the oil and gas industry?

David V. Goeckeler Chairman and CEO

I don’t know much about oil and gas. But I mean—listen, I have been in the tech industry for possibly longer than I should admit, probably about 40 years, building global technologies. I started my career long ago at Bell Labs. And I have managed many different technology lines, many different technology lines covering hardware, software, SaaS, with very large scales.

When I really entered this industry as a CEO, I was somewhat surprised by how it operates. Just like you said, this whole "boom and bust" cycle—I have said this many times—I think it’s harmful; there’s always someone who feels they are not getting what they want: either suppliers are like I was in '23, desperately trying to survive; or we are in a situation where people say, "I can’t get everything I need." I think this is due to how we do things. As I said earlier, there are some reasons for this, right?

We need to extend the investment cycle; the increase in supply is more like a step function, while demand is more like a curve. So aligning these factors is not easy. I don’t think this is something we should give up on. I also don’t think it’s inevitable. I feel that just because it has been this way for a long time doesn’t mean it must be this way in the future. Maybe I’m a bit arrogant, but this is our approach. We are innovators. We mean we invent new things, and this can also apply to business models.

This is basically what we do with the money; I think the world is doing well in this regard. I think if we think about this business model, yes, what have we learned from booms and busts? Don’t do that anymore. Really, it would be great if we didn’t do that anymore.

So how can we avoid repeating the mistakes? How can we ensure that the technology we develop gets a fair return? That’s really hard. Technology is very complex and not easy. This is 3D semiconductor technology; people have dedicated their lives to it.

The costs of doing this are very high. It requires massive capital expenditures, and fabs are very difficult to build and operate. Let’s get a fair return for this and establish a business model that can make all of this smoother. I think that’s entirely possible. I also think we have taken a few steps down that path and will continue to move forward.

Mark Newman, Bernstein Institutional Services LLC, Research Division

Great. Well, time is up. Thank you very much, David.

David V. Goeckeler Chairman and CEO

Thank you, everyone.