SanDisk Earnings Call: Customers Underestimate NAND Demand, NBM Long-Term Orders Booked Through 2030

Driven by the explosive demand for AI inference, SanDisk's revenue for the first quarter of fiscal 2027 surged 378% year-over-year. Leveraging its deeply tied NBM long-term contracts with hyperscale cloud providers and its new HBF technology, NAND flash memory is advancing comprehensively into the core layer of AI infrastructure.

On October 30, after the U.S. market close, SanDisk delivered an exceptionally explosive first-quarter fiscal 2027 earnings report. The company achieved quarterly revenue of $11 billion, up 23% quarter-over-quarter and a staggering 378% year-over-year; non-GAAP gross margin reached 85.1%, generating $5.7 billion in adjusted free cash flow.

In the subsequent earnings call, management revealed that "approximately 60% of this quarter's revenue growth was attributable to volume growth and 40% to price increases." Based on strong order visibility, SanDisk expects second-quarter revenue to be between $12.2 billion and $12.6 billion. At the same time, the company committed to returning 100% of surplus cash to shareholders, having already spent $5 billion on share repurchases this quarter.

Data center business surges: "Customers underestimated the flash capacity needed to run infrastructure"

The data center segment was undoubtedly the brightest highlight of this quarter's earnings. In the first quarter, SanDisk's data center business revenue reached $4.8 billion, up 61% quarter-over-quarter.

With the evolution of AI large models, inference workloads are undergoing structural changes. Chairman and CEO David Goeckeler noted at the meeting: "Longer prompts, more complex reasoning, multimodal inputs and persistent user context all significantly increase storage demand. HBM and DRAM support the highest performance tiers, while flash provides the capacity needed to efficiently retain broader context and avoid costly recomputation."

This elevation in strategic status is directly reflected in customers' order anxiety. SanDisk's long-term cooperation model—the New Business Model (NBM)—is being eagerly sought after by hyperscale cloud providers. Goeckeler stated bluntly: "We hear from customers that they underestimated the NAND flash capacity needed to run infrastructure. As a result, we have many customers ordering again, while also working to expand to new customers."

The market landscape is undergoing an irreversible shift. Management estimates that "the data center business has grown from about 20% of total industry volume a decade ago to nearly half today." Looking ahead, the company expects data centers to account for more than 55% of the market by 2027, with bid demand growing about 40% year-over-year. Goeckeler even boldly predicted: "By the end of this decade, key-value cache (KV Cache) will account for about one-third of the AI data center flash market, and the AI data center market will be close to the size of today's NAND flash market."

NBM long-term contracts lock in future supply and demand

Behind the earnings explosion, SanDisk's promoted New Business Model (NBM) has become a key moat locking in long-term profits. This contract model based on long-term commitments smooths out the strong cyclical fluctuations of the traditional storage industry.

Chief Financial Officer Luis Felipe Visoso revealed that among current NBM partner companies, "95% are investment-grade companies, and they generate approximately $130 billion in operating cash flow." Currently, about 50% of SanDisk's supply for 2027 and about two-thirds of its supply for 2028 have been booked by NBM projects.

Speaking to the real demand from downstream customers, Goeckeler offered a thought-provoking remark during the Q&A session: "Overall, we hear from customers that they underestimated the NAND flash capacity needed to run infrastructure." He said that not only are many existing customers placing additional orders, but new customers are also continuously coming in. "We now have a strong supply pipeline extending through 2030 and even further."

Currently, about 50% of SanDisk's fiscal 2027 capacity has been locked by NBM projects, and this proportion will rise to about two-thirds in 2028. Management expects that the NBM project backlog will exceed the company's supply by the end of this decade.

First HBF chip taped out, SLC sparks new architectural thinking

If TLC and QLC form the current base, then the next-generation storage technology customized for AI opens up the market's imagination.

At this earnings call, SanDisk announced a major milestone: its first High Bandwidth Flash (HBF) memory chip has completed tape-out. Goeckeler explained: "HBF is a new memory architecture for inference, with read bandwidth comparable to HBM but with greater capacity... designed to complement HBM by bringing NAND-level capacity closer to the processor." The first samples are expected to be delivered to customers next year.

In addition, the market is closely watching the potential of extremely high-endurance SLC (single-level cell) flash in AI architectures. When analysts asked whether this would cannibalize TLC's share, Goeckeler firmly denied it: "This does not constitute cannibalization of other NAND products, but rather an expansion of the TAM (total addressable market). The question is: can running NAND in SLC mode achieve sufficient levels of endurance and performance to replace some DRAM use cases? After all, the cost gap between the two is quite stark, roughly 20x to 40x."

Edge and PC markets: short-term bottoming, per-device capacity expected to rise next year

Compared with the booming cloud segment, the edge computing business, including smartphones and PCs, performed relatively steadily, with first-quarter revenue of $5.8 billion, up 7% quarter-over-quarter. The consumer business, constrained by soaring prices, fell 29% quarter-over-quarter to $397 million.

Faced with the short-term weakness in the PC market, Goeckeler offered an objective forecast. He acknowledged that the current two quarters are a trough period for sales, and expects PC sales may still see a single-digit decline next year. However, he emphasized the logic behind the increase in device content: "We expect PC storage capacity (measured in EB) to grow next year. According to our internal models, all segments of the NAND flash market... will show positive growth next year." At the same time, automotive electronics is rising rapidly and is regarded by SanDisk as one of the largest segments in the future edge computing space.

The following is the full transcript of the conference call:

Operator: Good afternoon, and welcome to SanDisk's Fiscal Year 2027 First Quarter Earnings Conference Call. All participants will be in listen-only mode. (Operator Instructions) Please also note that this conference is being recorded. I will now turn the call over to Mr. Ivan Donaldson, Vice President of Investor Relations.

Ivan Donaldson, Vice President of Investor Relations:

Before we begin, please note that today's discussion will contain forward-looking statements based on management's current assumptions and expectations, which are subject to various risks and uncertainties. These forward-looking statements include our expectations regarding the company's technology, energy and product portfolio, business plans and performance, capital allocation priorities, market trends and opportunities, and financial results. We undertake no obligation to update these statements. For more information on the risks and uncertainties that could cause actual results to differ materially from expectations, please refer to our Annual Report on Form 10-K and our other filings with the U.S. Securities and Exchange Commission.

We will also refer to non-GAAP financial measures today. Reconciliations of non-GAAP financial measures to comparable GAAP financial measures are included in the written materials posted in the "Investor Relations" section of our website.

Next, I will turn the call over to David.

David Goeckeler, Chairman and Chief Executive Officer:

Thank you, Ivan. Good afternoon, and thank you for joining SanDisk's Fiscal Year 2027 First Quarter Earnings Conference Call. We began fiscal year 2027 with another strong quarter, creating value for customers and shareholders and delivering results that exceeded expectations. Our strategy is working, and our execution is delivering strong results.

Revenue for the quarter was $11 billion, non-GAAP gross margin was 85.1%, and non-GAAP earnings per share was $49.51. We generated $5.7 billion in adjusted free cash flow and returned $5 billion to shareholders through share repurchases.

Our confidence in the company's long-term prospects stems from three core strengths: technology leadership, deep customer partnerships, and disciplined financial management. The rise of artificial intelligence and the signals customers are sending through our new business model—NBM partnerships—to extend business beyond fiscal year 2030 further reinforce the structural growth in NAND flash demand. Together, these strengths enable us to invest confidently in our technology roadmap, strengthen and expand customer relationships, generate sustained free cash flow, and deliver capital returns to shareholders.

Technology Leadership

It all starts with our technology leadership. BiCS 8 is at the core of our current portfolio, setting industry benchmarks for performance and density in both TLC and QLC flash. BiCS 8 introduces CMOS directly bonded to array technology, our CBA architecture, which delivers the performance AI data center customers require and lays the foundation for future generations. BiCS 9 combines the proven BiCS 8 array with next-generation CMOS technology, significantly improving interface performance while maintaining the proven memory architecture. BiCS 10 upgrades the node architecture across TLC and QLC flash, introducing a 1Tb TLC chip designed for high-performance workloads and a 2Tb QLC chip with industry-leading density and twice the read and write bandwidth of BiCS 8. The combination of these technologies will extend the BiCS 8 foundation to meet the growing performance and capacity demands of next-generation infrastructure.

High Bandwidth Flash (HBF) Progress

Beyond the core roadmap, High Bandwidth Flash (HBF) reached an important milestone in the first fiscal quarter—our first HBF memory chip has completed tape-out, and manufacturing is now underway. As we outlined at our Investor Day, HBF is a new memory architecture for inference that offers read bandwidth comparable to HBM but with greater capacity, enabling larger models, longer context lengths, and growing key-value caches. HBF is designed to complement HBM by bringing NAND-class capacity closer to the processor, and we remain on track to deliver the first HBF-based product samples next year.

HBF remains a future-facing innovation and is not included in the long-term financial model we presented at Investor Day. As customer adoption, product demand, and manufacturing readiness become clearer, we expect to provide more insight into the market opportunity.

Customers recognize our innovation strength and increasingly view SanDisk as an important partner for achieving their growth objectives. Through our NBM (New Business Model), we will continue to develop customer relationships into long-term partnerships built on innovation, collaboration, and shared commitment. Our recent customer engagements with the world's leading AI companies demonstrate that there is still room for further growth in market demand for our technology. Our market-leading technology, product portfolio, and customer partnerships are the cornerstone of our strong financial position, enabling us to reinvest to maintain technology leadership and expand supply through R&D-focused capital expenditures.

AI-driven demand growth

The AI boom further reinforces our strengths. As inference workloads grow, demand for scalable memory and storage continues to expand, making NAND flash play an increasingly important role in AI infrastructure. During inference, every generated token must reference its preceding context, creating a growing key-value cache that serves as the model's working memory. Longer prompts, more complex reasoning, multimodal inputs, and persistent user context all significantly increase these storage requirements. HBM and DRAM support the highest performance tiers, while flash provides the capacity needed to efficiently retain broader context and avoid costly recomputation. We believe this is a long-term trend that is expanding the role of NAND across the entire AI technology stack.

The memory hierarchy defines the role of each NAND architecture. TLC offers lower latency, higher speed performance, and greater endurance, meeting the needs of active datasets, checkpoints, and latency-sensitive key-value cache workloads. In the AI lab comparisons we presented at Investor Day, systems using SSDs achieved approximately three times higher token throughput than comparable systems relying solely on volatile memory, while reducing energy consumption by approximately 75%. QLC further increases storage density, making it suitable for capacity-intensive applications such as data lakes and the vector databases that power AI search and retrieval libraries. Across these tiers, increasingly persistent context ensures that data can be retained and recalled rather than regenerated.

As inference costs decline, we expect adoption and usage rates to continue to rise, confirming our view that by the end of this decade, key-value caches will account for approximately one-third of the AI data center flash market, while the AI data center market will approach the size of today's NAND flash market.

End-market overview

Next, let's look at the end markets. In the data center segment, our product portfolio can meet a wide range of AI infrastructure needs, from high-performance enterprise SSDs to high-capacity platforms. Market demand remains strong and continues to grow, and we estimate that the data center business has grown from approximately 20% of total industry volume a decade ago to nearly half today. This shift in customer mix is reshaping industry economics and is at the core of the transformation underway at SanDisk. In the quarter, the data center business again delivered strong growth and continued to outpace the overall market, driven by our increased supply to meet growing demand in this end market.

Edge computing is a large and strategically important market for SanDisk, covering smartphones, PCs, automotive tablets, and an expanding range of physical AI applications. The automotive industry is increasingly becoming an important growth driver for our edge computing business and represents a significant long-term growth opportunity. Our collaboration with leading customers continues to deepen, reinforcing our confidence that the automotive industry will become one of the largest markets in edge computing in the future.

Our consumer business remains a differentiated strategic asset, combining direct brand engagement with hundreds of millions of end users and broad shelf coverage through leading channel partners worldwide. The consumer end market provides a solid foundation for our business, and its direct connection to the market also informs product development and gives us early insight into changes in end-user demand.

Taken together, these results further reinforce the robustness of the business model we presented at Investor Day. We have achieved strong profitability and cash flow, and our focus going forward is on ensuring consistent and stable execution across all of our businesses.

Next, I will turn the call over to Luis to provide details on the company's latest business developments and outlook.

Luis Felipe Visoso, Executive Vice President and Chief Financial Officer:

Thank you, David. We are executing the strategy and plan we laid out at Investor Day and delivering satisfactory financial results.

In the first quarter, we completed two NBM partnerships. Because both transactions were completed before the fourth-quarter earnings call, we announced these partnerships on that call. We are actively in discussions with existing NBM customers who are evaluating the possibility of extending existing agreements, and we are also actively in discussions with new customers who are evaluating new NBM partnerships. These long-term partnerships give us greater visibility into market demand and financial performance, enabling us to build a more robust financial model. We will remain patient and disciplined in selecting the right long-term partners and in determining the optimal percentage of bids allocated to NBM, which may be higher than our current level.

First-quarter financial results

First-quarter revenue reached $11 billion, up 23% quarter over quarter and up 378% year over year. This performance significantly exceeded our prior guidance of $10.3 billion to $10.8 billion. Both volume growth and price increases drove revenue growth, with approximately 60% of the growth attributable to volume growth and 40% to price increases.

Data center revenue reached $4.8 billion, up 61% quarter over quarter, exceeding the market's expected growth rate for the quarter. Our leading products and strong customer partnerships (including partnerships with three U.S. hyperscale data center operators) underpin our continued expansion in this fast-growing end market.

Edge computing revenue reached $5.8 billion, up 7% quarter over quarter, reflecting resilience in the high-end and mainstream markets, partially offset by contraction in the low-end market. In edge computing, we benefited from strong growth from automotive and IoT customers. Consumer revenue was $397 million, down 29% quarter over quarter, primarily because the market is adjusting to higher prices. We will continue to be committed to serving all three major end markets and to continuously optimizing performance through our diversified product portfolio.

First-quarter non-GAAP gross margin was 85.1%, up from 84.6% in the prior quarter, primarily driven by higher volume and prices. This performance was better than our prior guidance of 83% to 85%. Non-GAAP operating expenses were $513 million, or 4.7% of revenue. Non-GAAP operating margin was 80.4%, up from 79.2% in the prior quarter. Non-GAAP earnings per share were $49.51, up from $39.25 in the prior quarter and $1.22 in the year-ago period. This performance was better than our prior guidance range of $44 to $46. At the end of the quarter, the company had 154 million diluted shares outstanding.

Key adjustments between GAAP and non-GAAP financial measures include: $91 million in stock-based compensation expense, or 0.8% of revenue; and $513 million in gains on equity securities, primarily attributable to our investment in South Asia.

During the quarter, we repurchased approximately 3.2 million shares of common stock for a total of $5 billion. Over the past two quarters, we have repurchased a total of 6 million shares, representing approximately 4% of total shares outstanding. Our repurchases reflect our confidence in the company's long-term value and our disciplined capital allocation strategy. We will continue to make appropriate investments in the company and return 100% of excess cash to shareholders.

Free cash flow

Cash flow from operations was $6.4 billion. Total capital expenditures were $378 million, or 3.4% of revenue. We expect total capital expenditures to increase over the remainder of the year and to be in the single digits as a percentage of fiscal 2027 revenue. During the quarter, we generated $5.7 billion in adjusted free cash flow, a margin of approximately 52%. At the end of the quarter, we held $5.1 billion in cash and cash equivalents and had no debt.

Outlook

Our confidence in the long-term health of the market stems from the structural growth of AI infrastructure and the increasingly important role of NAND flash in inference workloads.

Before turning to second-quarter guidance, I am announcing that the SanDisk board of directors has approved a 41-for-1 stock split of the company's common stock. The additional shares are expected to be distributed on November 19, and the split-adjusted shares are expected to begin trading on November 20.

We expect fiscal 2027 second-quarter revenue to be between $12.2 billion and $12.6 billion. We expect non-GAAP gross margin to be between 84.5% and 85.5%, including modest same-store price increases. Before the split, we expect non-GAAP earnings per share to be between $55 and $57, assuming total fully diluted shares of 151 million. After adjusting for the 41-for-1 split, non-GAAP earnings per share would be between $6.88 and $7.13, assuming total fully diluted shares of approximately 1.208 billion.

For reference, we expect the market to maintain strong growth in 2027. Data centers are expected to once again be the growth engine, accounting for more than 55% of the market, with bid demand growing approximately 40% year over year, driven by increased AI-driven capital expenditures. Previously, data center bid demand had more than doubled in 2026. Our NBM pipeline will exceed supply by the end of this decade.

Next, I will hand the call back to David.

Chairman and Chief Executive Officer David Goeckeler:

Thank you, Luis. We will continue to execute the strategy we laid out at Investor Day, which has strengthened our confidence in SanDisk's long-term prospects. With technology leadership, deep customer partnerships and disciplined financial management, as well as the structural growth in AI-driven NAND flash demand, we are better positioned to serve customers, generate sustained free cash flow growth and deliver returns to shareholders.

Over the past year, we have significantly improved the company's profitability and cash flow and laid a solid foundation for continued future growth. We will continue to focus on disciplined execution, sustained value creation and achieving our long-term financial goals.

Now, we will begin the Q&A session. (Operator instructions.) Today's first question comes from Ben Reitzes of Melius.

Analyst Ben Reitzes:

Hi everyone, thanks very much. Dave, I have two questions. The first is about NBM (new business model). You signed two contracts before the last tender. I wanted to ask, will there be many more signed going forward? Because it looks like your capacity is already limited and you can't sign them all. Also, if there are any changes to the contracts, could you remind us of the specifics of the contracts for the coming years? I also have a follow-up question. Thank you.

Chairman and CEO David Goeckeler:

Thanks, Ben. Let me offer some comments on NBM, and then Luis will add.

We have a very strong NBM pipeline. With growing subscriptions and commitments in the portfolio, we have already filled half of this fiscal year and two-thirds of next year. Now is a good time to move forward with the next batch of projects, and our pipeline is very robust. We said last quarter that we would be patient and carefully consider the next batch of customers, and that is exactly what we have done. We expect more customers to join over the next few months and even quarters as we continue to allocate the remaining portfolio. Our pipeline is very robust and still growing, and this has become our primary way of serving customers. I expect the percentage of the portfolio that is committed to continue rising over the next several quarters.

Luis, please add.

EVP and CFO Luis Felipe Visoso:

David covered most of it, I'll just add a few points. We are very pleased with our progress on NBM so far. We have been very careful and highly strategic in choosing whom to work with. We reviewed the companies we are working with, and 95% of the NBM partner companies are investment-grade companies that generate about $130 billion in operating cash flow. We believe we made the right choices. We will continue to be patient and selective, and our pipeline is very robust.

As I mentioned in my prepared remarks, we expect these projects to continue growing. We have previously shared with everyone that about 50% of our projects in 2027 are NBM projects and about two-thirds in 2028 are NBM projects, and we expect that number to rise slightly from two-thirds.

Chairman and CEO David Goeckeler:

Ben, let me add one more point. We are indeed working to balance demand from customers who are reordering. Overall, we hear from customers that they underestimated the NAND flash capacity needed to run infrastructure. So we have many customers reordering while also working to bring on new customers. There is a lot to handle, and these are very complex, large agreements, but we are very confident in our current pipeline and portfolio, and the product response has been good. Over the coming quarters, we will continue to update on the latest progress of the NBM portfolio.

Analyst Ben Reitzes:

Okay, great. One follow-up question. You may have seen that Intel cut its guidance for next quarter due to a decline in the PC business, and the memory impact is bigger than expected.

I'm just curious whether you have noticed increased interest in NBM from the PC market. If so, do you think that for some device makers this market could develop into NBM, or actually become more attractive because they are willing to pay higher prices? Given that they seem to have a shortage of products and poor sales in this area, how should we view this situation?

EVP and CFO Luis Felipe Visoso:

We can't talk about specific customers. We have said in the past that we do have NBM in our data center business, which is data-center-centric, but we also have edge computing customers. We offer this model to any customer that needs it and start the relevant process where appropriate. We are not trying to limit any customer; we welcome collaboration.

Chairman and CEO David Goeckeler:

On that point, our own analysis confirms this. We know unit sales have declined this year and are expected to decline slightly again next year, in the low single digits. The current two quarters are the trough with the largest year-over-year declines, and in terms of unit sales, this is where we feel the largest year-over-year change in the market. Nevertheless, we expect PC storage capacity (in exabytes) to grow next year. According to our internal model, all segments of the NAND flash market, including PCs, smartphones and data centers, will show positive growth next year, as content per device continues to increase, partly because high-end products are becoming more widespread. The market has adjusted, and in terms of unit sales we are at the trough of the year-over-year decline, but we expect conditions to improve.

Great, thank you. Beyond the favorable market conditions, it seems the deeper reason lies in your company's positioning in the data center space. Your company's market share grew nearly 20x year over year and 61% quarter over quarter. Clearly, the market is strong, but your company's market share also seems to have risen sharply. Is this because of BiCS 8 certification? Or is it simply because the market environment has changed? What exactly is driving your company to achieve such high market penetration? If, in the future, data transmission in data centers accounts for half of total data transmission, can your company also achieve that goal within its own product portfolio?

Chairman and CEO David Goeckeler:

Joe, we are very pleased with where we are in the data center space. The growth in the performance figures we announce each year is the result of a combination of factors. First, it comes down to the underlying node of NAND flash. BiCS 8 is an excellent node, and the CBA process delivers very pure CMOS, enabling extremely high interface speeds. We built the right products—our TLC products have been on the market for several years; just a few quarters ago, we began selling QLC products and making a profit on them, and we expected volumes to grow rapidly, which is exactly what is happening. So we brought the right products to market at the right time, based on the right technology.

Over the past quarter or so, AI inference architectures have improved further. As we look at the market and segment it internally, customers have a more granular understanding of how to use NAND flash, TLC flash and QLC flash and other technologies, spanning everything from key-value caches to scratchpads to fast data lakes, as well as all the sub-segments of these markets, such as inference in agent orchestration, model pre-training and model post-training. Overall, we see NAND flash being used in more refined ways.

In the past, QLC flash was a very good fit for certain use cases, and you could even see some QLC sub-customers using it for things like key-value caches where the technology allowed. On the other hand, we see SLC flash also being incorporated into architecture designs—it is not really being sold yet, and is more at the architecture discussion stage, and those discussions are not focused on performance but more on endurance under agentic workloads and on whether NAND flash can replace DRAM.

We believe this kind of architecture will become very popular over time. Once the transition is complete, given the price difference, users will stick with it. These NBMs allow us to have deep strategic conversations with customers about how they use our products, and also deep discussions with our engineers about how to solve specific problems for different use cases. All of this has greatly advanced our product portfolio. As I said, we have signed a second batch of NBMs with our largest customer and are in talks with others. The overall architecture is very strong, and we are pleased with it.

Analyst Joe Moore:

How do you think this blended style of product portfolio will evolve over the long term?

Chairman and CEO David Goeckeler:

We are working to catch up so that our portfolio mix aligns with the overall market. We will continue to push this forward throughout 2027. We will see whether we can truly achieve the goal by the end of this quarter, but if you look at fiscal 2026, our portfolio mix will be much higher than it is now. We will continue to make progress on this.

Operator: The next question comes from Mark Newman of Bernstein.

Analyst Mark Newman:

Thank you very much. First, I want to talk about your guidance on price and volume. In the first fiscal quarter results you just reported, you mentioned that about 60% of revenue growth came from volume and 40% from price. My question is, how do you expect that ratio to change in your guidance? Is there anything to add on pricing? I know these new NBMs account for a high share of volume, but I still want to understand the pricing side—I recall you mentioned a slight improvement in pricing in the second fiscal quarter. Could you elaborate on that?

Then I have a follow-up question about HBF and some new NAND technologies.

Chairman and CEO David Goeckeler:

Let me answer the first question. You are right that we reported that about 60% of first-quarter revenue growth came from volume and 40% from price. We expect both to contribute again next quarter. We typically do not break out that number, but as we said in our prepared remarks, we expect a slight increase in price. So we expect both volume and price to contribute this quarter, and we expect prices to continue to rise slightly throughout the year.

Analyst Mark Newman:

That is helpful and clearly explained. My next question is about HBF and other new NAND technologies in the data center. Someone mentioned earlier that there has been some progress on HBF, which is good, but I want to know if there is anything else on progress and discussions with partners and customers?

Dave, you just mentioned some other new NAND technologies in the data center, such as Storage Next and CMX context memory. What is your view on these emerging NAND technologies in the data center? Do you think they will all ultimately succeed? It would be great if you could share some thoughts on how these technologies interact, their respective roles, SanDisk's role in them, and the timing. Thank you.

Chairman and CEO David Goeckeler:

Of course. Mark, our conversations with potential customers about HBF are very in-depth and ongoing—we will continue talking with them until we put product samples in their hands so they can start using them and understand how the product fits into their systems. We will keep updating the relevant information, but this is an almost never-ending communication process.

On your second question, I think this is very good for the development of AI inference, because as it scales, architectures must continuously be optimized to improve economic efficiency. This leads NAND flash to be used more in various parts of the architecture, because NAND is the most scalable semiconductor technology. A few years ago, everything revolved around increasing density, from TLC to QLC, and we are still doing that now; that remains an important part of the strategy. But now some directions have even shifted: how can NAND flash be used to improve the endurance of specific parts of the architecture? Is NAND flash's endurance and performance sufficient to meet the needs of specific application scenarios in the architecture? In specific scenarios, does it offer advantages over competing technologies that are more expensive and harder to obtain?

I think this is a very active area of discussion, there are many ideas right now, and I think there should be many ideas. I believe all of these ideas will be adopted. The most important thing is to stay close to the people who are putting capital into expanding infrastructure, because ultimately they will make the decisions that best suit them based on the large-scale application scenarios they need to address and where their capital is being invested.

We are actively participating in these conversations, and I expect they will continue to deepen and that many different ideas will emerge about how to extend our technology in different directions to adapt to more and more application scenarios. In the product development process, the constant principle is: deliver the right technology at the right time, at the right intensity, and at the right cost—no more, no less, just right. That is the core of product development, because only by delivering products at the right price can you earn the right return. In the field of AI inference architecture, we work closely with customers to advance this process across the board.

Analyst Mark Newman:

Great, thank you for the information.

Operator: The next question comes from David Gibson of MST Financial.

Analyst David Gibson:

Thank you all, this was really a great quarter! I have a brief question. I want to ask whether you are considering accelerating the transition of BiCS from 8 to 9, 10, and so on, in order to accelerate bit growth in the short term and meet some very strong customer demand?

Chairman and CEO David Goeckeler:

No, we remain committed to achieving the capacity growth targets discussed previously, which is in the teens to low twenties percentage range. These key node transitions require a great deal of planning, equipment procurement, and production line setup. We have already set factory plans to address the capacity growth we have been discussing for a long time. There is no change to the plan at present.

Analyst David Gibson:

Okay, thank you. One follow-up question. You mentioned earlier that more resources would be allocated to the data center in the first quarter. So, are there any constraints on allocating more resources to the data center?

Chairman and CEO David Goeckeler:

In some respects, yes. After coming out of the fab, you get a certain percentage of TLC, QLC, specific chip mixes, and so on, so we need to continuously adjust to get the right products to the right customers. In addition, there is a lot of operations research work within Luis's organization; he works closely with the sales team to understand existing inventory, fab shipments, product mix, and ultimately where sales are headed.

Luis, would you like to add a few words?

Executive Vice President and Chief Financial Officer Luis Visoso:

I think the only thing to add is that this obviously varies depending on the time horizon we are considering. In the short term, it is almost impossible to do anything—producing a wafer takes about 120 days. But if you look further out, there is much more room to maneuver.

Analyst David Gibson:

Okay, thank you very much.

Operator: The next question comes from CJ Muse of Cantor Fitzgerald.

Analyst CJ Muse:

Good afternoon, and thank you for taking the question. I wanted to ask about your comment in the slide about NBM pipeline supply exceeding supply by the end of the decade. Could you elaborate on that? How does that change how you think about the industry's bit supply over the next few years? Are you adjusting your capex plans, and by how much? I know you mentioned mid-single-digit increases, which implies total capex of roughly $2.5 billion to $3 billion, which is quite low. I'd love to hear how that visibility shapes your view on bit growth and supply?

Chairman and CEO David Goeckeler:

The first thing I'd say about that, CJ, is that this process is working. Just two or three quarters ago, we were discussing whether to establish this business model, we had some like-minded customers, and we made very good progress. Now, a few quarters later, we discussed this after Investor Day, customers are coming to us seeking NBM services, and this is no longer our only business model. We have many customers with different volumes and products, and we want to be able to provide services based on customer needs, there's no doubt about that.

What's encouraging is that we now have a strong supply pipeline extending through 2030 and even beyond—we have many customers willing to give us their demand signals and financial commitments. As for the question raised earlier, if we ultimately choose to change the supply-demand relationship and so on, whether these factors can make a difference for us, but we're not at that point yet. We still need to negotiate all these issues, secure commitments, and get everything aligned.

The process is going well right now, we're very pleased with where we are, and it gives us a clearer view of our future planning. The process is still ongoing, and over the next few quarters we'll continue to update everyone as the work progresses.

Analyst CJ Muse:

Luis, I have one follow-up question. Regarding these contracts, given that some have deposits, can those deposits be used for buybacks? Or is there a restriction on how that money can be used?

Executive Vice President and Chief Financial Officer Luis Visoso:

In theory we could, CJ. But as we explained at Analyst Day, the vast majority are financial guarantees held by third-party financial institutions, so they are not on my balance sheet. The only thing included on the balance sheet is prepayments, and that amount is minimal. But there are no legal restrictions on that.

Analyst CJ Muse:

Great, thank you very much.

Operator: The next question comes from Karl Ackerman of BNP Paribas.

Analyst Karl Ackerman:

Thank you. It's encouraging to see NAND flash capacity per XPU and AI server increasing to meet key-value cache requirements. I think this contrasts sharply with declining memory capacity and the reduction in some server SKUs. David, could you talk about how you see enterprise SSD capacity per server trending in 2027 in the context of AI adoption? Because AI is undoubtedly becoming an increasingly important part of your portfolio.

Chairman and CEO David Goeckeler:

Karl, that's really a big question. What we're doing right now, without a doubt, in agent-orchestrated inference, our products are closely tied to the CPU requirements of these types of applications. We have a very clear idea of the expected demand per CPU, which is roughly 60 to 90 GB per CPU for these types of applications. I haven't calculated the elasticity relationship between the server TAM and these two, but we can follow up.

As I said earlier, we're seeing more and more NAND flash being deployed in various different ways across parts of the inference architecture. TLC, QLC, and as I mentioned earlier, those now seeking high-endurance solutions want solutions that can meet the needs of certain parts of the workload—where they don't need to use DRAM and can use this type of NAND flash instead. So it's all very encouraging, and we'll continue to watch it closely, and we'll also start analyzing the market size of these two types of NAND flash and the trade-offs between them.

Analyst Karl Ackerman:

If I could, I'd like to ask one more follow-up question. Do you have any update or outlook on your enterprise SSD market share target? Given the quite good results you've achieved so far, can you reach your 10% to 15% market share target in 2027? Thank you.

Chairman and CEO David Goeckeler:

Many years ago, we set some share targets, but I've always tried to avoid setting share targets because it can sometimes lead to unreasonable choices in the portfolio. We want to maximize portfolio value across all time horizons, including short, medium, and long term, so we want to keep the portfolio balanced.

Obviously, we will increase the proportion of data center configurations, as you have seen. We discussed this before as well, and I think we achieved 17x year-over-year growth—in a business of this scale, you rarely see 17x growth. I expect this number to continue climbing through 2027. To some extent, the balance point will depend on what we observe each quarter and the NBM agreements we sign, which will give us a clearer understanding of how the market landscape will evolve over the next few years. So this number will certainly continue to climb, but I won't set a specific target.

Analyst Karl Ackerman:

Thank you.

Operator: The next question comes from Vijay Rakesh of Mizuho Bank.

Analyst Vijay Rakesh:

Hi, David. This really is a great quarterly report and guidance. On data centers, you mentioned that demand will grow 40% in 2027, which is very impressive. Do you think the transition to BiCS 9 and BiCS 10 will help drive some of that demand growth? I'd like to ask about the supply situation in this regard. I also have a follow-up question.

Chairman and CEO David Goeckeler:

BiCS 8 and BiCS 9 are an important part of our data center strategy. They allow us to achieve faster interface speeds within the same memory cell, and CBA is exactly what gives us that flexibility, which we also discussed at Investor Day. In that sense, yes, they are an important part of the strategy—there is no doubt about that. The key is delivering the right products to the right customers. Our TLC products have consistently performed well, and now QLC is gradually ramping, and BiCS 9 will also gradually ramp, but it will be some time before real adoption. We have released BiCS 10, but we are still gradually rolling it out, and we still have a long way to go before real adoption in future BiCS nodes.

Analyst Vijay Rakesh:

Understood. On endurance, as you mentioned, some peers are discussing SLC NAND flash. How do you think the market will develop next year? Beyond the overall demand growth you are observing, could the emergence of SLC NAND flash also intensify competition in the NAND market? Thank you.

Chairman and CEO David Goeckeler:

It is too early to form a very definitive view now. We are in communication with some customers, who have revealed some demand signals for SLC. Obviously, once you start mixing in SLC, there is a clear conversion ratio, whether from TLC or QLC, of roughly three-to-one or four-to-one. Therefore, the amount of SLC used and actually deployed will affect overall market supply. But it is too early to make specific forecasts now.

I think this is an architectural discussion. Customers want a better understanding of this technology and how it applies to the various use cases in which they build inference architectures. We will continue to update on progress over the next few quarters and see how things develop, but it is definitely worth watching.

Good afternoon, and thank you for taking my question. I'd like to get a more specific strategic perspective. Your company's consumer business clearly has seasonal fluctuations in the first quarter, and so does the PC business, but the data center business is currently in a long-term growth phase. Louis, based on the information you have now, I'd like you to give a rough forecast of whether you think third fiscal quarter results will grow, decline, or be flat sequentially.

Executive Vice President and Chief Financial Officer Luis Visoso:

You mean the third fiscal quarter? Historically, the third quarter is usually our weakest quarter. For example, last quarter our bid declined 19% sequentially. Although we do not provide guidance for third quarter results, for modeling purposes, I would take a conservative view on sequential growth from the third quarter to the second quarter.

Analyst Jim Schneider:

Thank you. Also, on share repurchases, it's good to see the program continuing. I'm a bit curious: as the authorization period approaches, do you expect the board to renew the authorization quarterly? Or, given that the current stock price is well below expectations, can you elaborate further on how you plan to execute the actual buybacks? Thank you.

Chairman and CEO David Goeckeler:

We have a very clear strategy: 100% of surplus cash will be returned to shareholders. We will execute this strategy actively and effectively, and we know how to communicate with the board. We meet every quarter, and I don't think we will encounter any resistance in execution. I expect we will remain committed to this strategy of returning 100% of surplus cash to shareholders. Over the past few quarters, we have gotten off to a good start—we have repurchased shares equivalent to 4% of our initial market capitalization or shares outstanding, and we will continue to execute this program properly.

Analyst Jim Schneider:

Thank you.

Operator: The next question comes from Krish Shankar of TD Cowen.

Analyst Steven (on behalf of Krish):

Hello everyone, thank you very much for taking my question. This is Steven calling on behalf of Krish. My first question is for David. I'd like to understand your CSP customer base—what is the current demand situation for TLC and QLC flash chips? As NBM becomes more widespread, will the mix of TLC and QLC flash chips change significantly later this decade?

Chairman and CEO David Goeckeler:

The TLC-to-QLC mix is still holding at around 70:30. TLC has consistently been the main driver of the product mix. We are also increasingly hearing from customers that they want to reaccelerate QLC and explore different use cases. We will watch how this develops toward the end of the decade. The current 70:30 mix is quite good; there is some fluctuation each quarter, but the mix is basically holding at this level, and we expect that to remain the case at least for the rest of this fiscal year. We set that time frame, and then we will continue discussing with customers and see how they view QLC.

I should also mention that the Stargate product—our QLC product—is ramping very well. Just a few quarters ago we were still discussing product qualification, and now qualification is fully complete and customers are gradually increasing orders. This product has been very well received and is a very key part of our portfolio.

Analyst Steven:

Understood. One more question on the mobile edge computing market—given that this market also saw a double-digit unit decline this year, although the decline was mainly concentrated in entry-level and mid-range. I'm a bit curious: based on the demand signals some of your customers in that market are providing, can your mobile business still grow overall next year?

Chairman and CEO David Goeckeler:

Let's talk about the market first. The low end is indeed under pressure, which will lead to lower units but higher content per unit. Overall, we expect data volume in that market to grow in exabytes next year—even though units will decline slightly. As for our portfolio, this is a bit like the data center; I don't want to make predictions for individual markets. We will manage the portfolio with flexibility and agility. We have been very clear that we want a balanced portfolio, but we will optimize financial results over the short, medium and long term and integrate these factors to achieve the best outcome. We will keep you updated on the composition of the mobile business.

Analyst Steven:

Great, thank you very much!

Operator: The next question comes from Wamsi Mohan of Bank of America.

Analyst Wamsi Mohan:

Thank you very much. You mentioned on the slide that leading AI companies may signal future demand growth. I'd like to know whether these companies refer to AI labs? Also, is the current credit issue with Neo Clouds having any impact on you? I have a follow-up question as well.

Chairman and CEO David Goeckeler:

We have not seen any such situation. On the contrary, we are seeing very strong customer demand. If we look at our largest customers, they are basically all investment-grade companies with very ample cash flow. So we have not seen any impact from credit issues, even if such issues do exist in the market, and demand from our customers is strengthening compared with last quarter.

Analyst Wamsi Mohan:

Executive Vice President and Chief Financial Officer Luis Visoso:

We sell products across all business segments—we sell to hyperscalers, we sell to model builders, we sell to everyone.

Analyst Wamsi Mohan:

Okay, thank you Luis. For this quarter, accounts receivable seem to have picked up slightly. Is that because payment terms have improved as more bids adopt NBM payment methods, and should we expect that to continue? Or were there other factors this quarter?

Executive Vice President and Chief Financial Officer Luis Visoso:

We basically saw some changes in customer mix, which had some impact on our accounts receivable—just a little, but mainly a change in customer mix.

Analyst Wamsi Mohan:

Okay, thank you very much.

Operator: The next question comes from Amit Daryanani of Evercore.

Analyst Amit Daryanani:

Thank you very much. David, I'd like to start with capital allocation. As you transition to a more predictable NBM-driven business model, does that change how you think about your capital structure? Historically, memory chips have been widely viewed as more cyclical, so balance sheet flexibility is crucial. But as your business becomes less cyclical, does that make you willing to add leverage and thereby accelerate capital returns? Or are you still comfortable with the $5 billion quarterly share repurchase program? I'm just wondering when you think we could add leverage to take advantage of the lower cyclicality of the business?

Chairman and CEO David Goeckeler:

We have already published a financial model, at our analyst day, we spent three and a half hours discussing this issue, and committed to executing the financial model through the end of 2030. We plan to execute this model, and we are very confident in our strategy of maintaining zero debt, having a very strong balance sheet, and then investing the proceeds in our own business. Our business is growing, and there are many opportunities for internal innovation, and we will continue to invest in projects such as HBF and 3D matrix memory, in addition to the BiCS roadmap and all the excellent products we are developing. After completing these, we will return 100% of cash flow to shareholders, that is our plan, and we will continue to execute it for the foreseeable future until a change of plan is announced otherwise. I do not expect any changes in the near term.

Analyst Amit Daryanani:

Okay. So, regarding HBF, rather than asking you exactly how large the total addressable market (TAM) is, I'd like you to talk about what exactly the scaling variable is here as inference models and context windows expand? Should we think that HBF capacity will grow in sync with GPU compute power, or faster? Also, you said you will begin initial sample testing in 2027, so from then to achieving meaningful revenue growth, what are the key milestones? Thank you.

Chairman and CEO David Goeckeler:

On your first question, it is still too early to assess. On your second question: after we get the chip and complete the tape-out of the base die, we will integrate the controller with the base die, and then deliver it to customers so they can begin evaluating how the product actually performs in their infrastructure.

We have done a lot of modeling work previously, but after actually delivering physical products to customers and carrying out integration work, customer feedback will help us gradually answer the questions everyone cares about, such as: how big the market opportunity is, how fast the scaling process will be, what the TAM is, and in which scenarios the architecture will be deployed. We will advance in the direction of customer demand.

The key next step is to truly put the product in customers' hands. This will open a series of new conversations—we will no longer rely on simulations, but directly observe the product's real performance in their infrastructure and whether it meets customers' expected results. This will also help continuously expand the product's application scenarios.

At the current stage of the R&D process, a certain degree of restraint is needed to avoid overwhelming the engineering team with a large number of application ideas from customers. When we truly have a deliverable physical product, we can open the window for dialogue, which will bring more application scenarios. However, these remain to be determined.

Overall, when the key milestone of "having a real product" is reached, many things will move forward—customers can begin actual use and evaluate the value the product contributes to their goals. Their goal is to build world-class inference architectures at optimal cost, and we believe HBF is an excellent technology for achieving this goal. But the specifics remain to be verified at that time.

Amit Daryanani (Analyst): Thank you very much.

David Goeckeler (Chairman and CEO): Thank you.

Operator: The next question comes from Mehdi Hosseini of SIG, please go ahead.

Mehdi Hosseini (Analyst):

David, I want to go back to a point you mentioned earlier—you said AI inference is continuously being optimized and that this has affected SLC demand, requiring more SLC. My question is, has this adversely affected TLC adoption or growth? Or is SLC demand additional to market demand for TLC and QLC?

David Goeckeler (Chairman and CEO):

I think this does not constitute cannibalization of other NAND products, but rather an expansion of the TAM—the market is beginning to need memory technologies with different characteristics, including higher endurance and stronger performance. Our products have enough flexibility to run wafers in SLC mode. Customers are gradually discovering and understanding how to flexibly use this technology according to the characteristics of the problem they need to solve. This is just a more refined understanding of the architecture, not cannibalization of NAND products at various levels.

Going further—of course it is still early, and judgment should be cautious—the question is: can running NAND in SLC mode reach a sufficient level of endurance and performance to replace some DRAM use cases? After all, the cost gap between the two is quite stark, about 20x to 40x.

Wamsi Mohan (Analyst):

Mehdi, I think this is a very positive signal in many respects. From the actual situation of architecture scaling, when you scale the architecture—

David Goeckeler (Chairman and CEO): You have to do it in the most cost-effective way. NAND is the most scalable semiconductor technology, so it is not surprising that people keep finding more use cases. In a sense, scalable technologies tend to attract and expand the TAM, and that is exactly what is happening now—as these architectures become more refined and more mature.

Mehdi Hosseini (Analyst): Okay. Let me ask from another angle: given market demand for higher-endurance SLC, and the period in which customers are waiting for HBF commercialization to land, how much flexibility does the joint venture (JV) have to increase SLC capacity to meet short-term demand?

David Goeckeler (Chairman and CEO):

I would not think about this from the perspective of "SLC capacity." Our memory cells are flexible enough to run in any mode—SLC, MLC, or TLC—as needed. Therefore, we can use existing wafers to supply what the market needs, and that is our basic approach.

I understand you want to go directly to the supply-level question, but it is really too early. At this stage, it is mainly still in the conversation stage where customers are exploring how to apply the technology and forming preliminary demand expectations. This is a direction worth continuing to watch, and it will undoubtedly have an impact on the market—after all, a small amount of SLC corresponds to a large amount of QLC. But as for more details, Mehdi, I will share them with everyone later.

Mehdi Hosseini (Analyst): Okay, understood. Thank you for the detailed information.

Operator: Everyone, that concludes today's Q&A session. Now let us invite management to make closing remarks.

David Goeckeler (Chairman and CEO): Thank you all for participating in this conference call, much appreciated. Today's discussion was very thorough, and I look forward to meeting everyone again this quarter. Thank you.

Luis Visoso (EVP and CFO): Thank you. Risk Disclosure and Disclaimer

Markets carry risk, and investment requires caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment based on this is at the user's own risk.