Sanjay Mehrotra
Chief Executive Officer, President & Director at Micron Technology
Thank you, Farhan. Good afternoon, everyone. Micron delivered fiscal third quarter revenue within our guidance range with gross margin and EPS above the range. The ongoing improvement of customer inventories and memory content growth are driving higher industry demand, while production cuts across the industry continue to help reduce excess supply. As a result, pricing trends are improving, and we have increased confidence that the industry has passed the bottom for both quarterly revenue and year-on-year revenue growth.
Our technology leadership and strengthening product portfolio, positions us well across diverse growth markets, including AI and memory centric computing. Beyond this downturn, we expect to see record TAM in calendar 2025, along with a return to more normalized levels of profitability. The impact of the May 21st, decision by the Cyberspace Administration of China on Microns business remains uncertain and fluid. Several Micron customers including mobile OEMs had been contacted by 13 critical information infrastructure operators or representatives of the government in China, concerning the future use of Micron products.
As discussed before microns revenue with companies, headquartered in mainland China and Hong-Kong, including direct sales as well as indirect sales-through distributors accounts for approximately a quarter of Micron's worldwide revenue and remains the principal exposure. We currently estimate that approximately half of that China headquartered customer revenue, which equates to a lower double-digit percentage of Micron's worldwide revenue is at-risk of being impacted. The significant headwind is impacting our outlook and slowing our recovery. Micron is working to mitigate this impact over-time and expect increased quarter-to-quarter revenue variability. Micron's long-term goal is to retain its worldwide DRAM and NAND share.
Turning to technology. Micron continues to lead the industry in both DRAM and NAND technology. We are investing prudently to maintain our technology competitiveness while managing capex, node ramps and wafer start reductions to reduce our bit supply and align it with demand. Our industry-leading one beta DRAM and 232-layer NAND nodes are achieving world class yields and the yields ramp for these nodes has been faster than any of our prior nodes. These leadership nodes, provide a strong cost capability along with best-in class powered and performance specifications that will be leveraged across the portfolio of DRAM and NAND products.
In fiscal Q3, we achieved several important product qualifications on these advanced nodes and are well-positioned to ramp them in fiscal 2024. We are also making good progress towards the introduction of our EUV-based 1-gamma node in 2025. This node will be manufactured first in our Taiwan site, where we already have EUV capability installed and operating in preparation for this ramp. We also recently announced plans to bring EUV technology to our fab in Hiroshima Japan with support from the Japanese government. Micron will be the first company to bring EUV technology to Japan for production.
We are also advancing our global assembly and test network in order to support our product portfolio and extend our ability to deliver on global customer demand in the future. In China, we announced an investment of approximately $600 million over the next several years in our operations in Xi'an. This builds on our long history of significant investment in our Xi'an assembly, packaging and test operations. As a part of this investment, we have decided to purchase the assembly equipment of our partner Powertech Semiconductors Xi'an, who has been operating inside our Xi'an facility for the last eight years. We also intend to construct a new building at our Xi'an site to provide space to add more product capabilities. This will allow us over-time to serve more of the demand from our customers in China from the Xi'an side.
In India, with the strong support of the Indian government, we will build a new assembly and test facility in Gujarat to address demand for the latter half of this decade. We are also increasing our investments in assembly and test capacity in Taiwan for high-bandwidth memory products as we gear up for stronger demand in this segment driven by the AI wave.
Now turning to our end-markets. Customers continue to make progress in reducing that excess inventory in fiscal Q3. Most customer inventories in the PC and smartphone segments are close to normal levels now, consistent with our forecast six months ago. Some of these inventory levels can get distorted by customer attempts to leverage current prices, which are deemed to be transient and unsustainable at these levels to purchase additional volumes before prices rise significantly. Datacenter customer inventory is also improving and will likely normalize around the end of this calendar year or somewhat thereafter, depending on the growth in traditional data center spending.
In data center, we saw strong sequential revenue growth in both cloud and enterprise in fiscal Q3 driven by some recovery from depressed sales levels in fiscal Q2. The recent acceleration in the adoption of generative AI is driving higher-than-expected industry demand for memory and storage for AI servers, while traditional server demand for mainstream datacenter applications continues to be lackluster. Micron's product portfolio and roadmap of innovative products positions us to capture growth opportunities from AI and data-centric computing architectures for both training and inferencing. Increasingly large AI models with an exponentially growing number of parameters are driving demand for dramatically higher memory content.
As we have said before, AI servers have six to eight times the DRAM content of a regular server and three times the NAND content. In fact, some customers are deploying AI compute capability with substantially higher memory content. A striking example is NVIDIA's DGX GH200 supercluster, which shows just how memory intensive AI workloads can be. It provides developers, the ability to support giant models with a massive shared memory space of 144 terabyte.
A significant majority of that memory footprint is enabled by a joint development project between our two companies that extends Micron's low-power DRAM leadership to server class applications. We are proud to pioneer this differentiated LP DRAM innovation to deliver a significant reduction in data center power consumption compared to DDR based solutions, helping to support our customers' green initiatives.
High-bandwidth memory used in high-performance computing is seeing very strong demand this year, driven by demand for generated AI. We are working closely with our customers and have begun sampling our industry-leading HBM3 product offering. The customer response has been strong and we believe our HBM3 product delivers significantly higher bandwidth than competing solutions and establishes the new benchmark in performance and power consumption, supported by our one beta technology, TSV and other innovations, enabling a differentiated advanced packaging solution. We expect to begin mass production ramp for this exciting HBM3 product in early calendar 2024 and to achieve meaningful revenues in fiscal 2024.
Micron also has a strong position in the industry transition to D5, which is the latest generation of DDR memory. Our D5 percentage of DRAMs shipments has more than doubled from fiscal Q2 to Q3. And we expect Micron D5 volume to crossover D4, at the end of first calendar quarter of 2024 versus mid-calendar 2024 for the industry. Micron's 1-Aplha D5 modules are qualified and shipping to datacenter customers. We are also making good progress on our 1-Beta based high-density, 128 gigabyte D5 modules using 32 gigabit die that optimize cost and performance to provide customers with a lower-cost of ownership solution for memory intensive workloads like AI.
We expect these high-density modules to ramp-in calendar Q2 of 2024, with significant cost improvements over today's expensive TSV-based solutions in the industry. Our 96 gigabyte D5 high-density module built on 1-Alpha technology using 24 gigabit die is already shipping in volume and delivers equivalent performance for the majority of workloads versus the more expensive TSV dual die package based 128 gigabyte modules.
In data center SSDs, Micron's entire portfolio is now on 176 layer or 232-layer NAND demonstrating our product and technology leadership. We are in a strong position to serve AI demand for fast storage as these data-intensive applications proliferate. In fiscal Q3, we launched the world's first 200 plus layer NAND datacenter SSD and qualification is in progress at multiple key customers to support AI cluster installations. In fact, we have already passed qualification of the product at a critical server OEM partner. We also launched our extreme endurance, data center SSD, which offers superior scalability and affordability versus hard drives.
In PCs, we now forecast calendar 2023 PC unit volume to decline by a low-double-digit percentage year-over-year with PC units expected to be below the pre-COVID levels last seen in 2019. We are excited about the ongoing industry transition to D5 and are well-positioned for it with our strong D5 product lineup. Industry client's D5 mix is expected to crossover from D4 in early calendar 2024.
In fiscal Q3, we achieved record quarterly client SSD bit shipments driven by share growth in client SSDs as customers adopted our industry-leading solutions. Our SSD QLC bit shipment mix reached a new record for the third consecutive quarter with growth in both client and consumer. Last month, we launched crucial T700, the world's fastest Gen5 PCIe consumer SSD, built with our 232-layer NAND.
In graphics industry analysts continue to expect graphics TAM growth CAGR to outpace the broader market supported by applications across client and datacenter. We expect customer inventories to normalize in calendar Q3. We plan to introduce our next-generation G7 product on our industry-leading one beta node in the first-half of calendar year 2024.
In mobile, we now expect calendar 2023 smartphone unit volumes to be down by a mid-single-digit percentage year-over-year. While units are weaker, we are seeing stronger memory content growth driven by a mix-shift towards premium phones and elasticity. We expect sequential growth in fiscal Q4 as customers prepare for upcoming product launches in the back-half of calendar 2023.
In fiscal Q3, we achieved key mobile customer qualifications on our 1-beta base LP 5X and started high-volume revenue shipments to tier-1 OEMs. In addition, we achieved significant milestones in UFS with the qualification and ramp of a high-capacity UMC P5 featuring 16 gigabyte of DRAM and 512 gigabyte of NAND. We have also started to sample a new UFS4 product. Based on our latest two 232-layer NAND technology, which enables industry-leading performance for flagship handsets.
Last, I'll cover the auto and industrial end-markets, which represent over 20% of our revenue and contribute more stable revenue and profitability. Micron continues to lead-in automotive which is a key market and growth driver for us. In fiscal Q3 auto revenue reached another quarterly record and grew by a high-single-digit percentage year-over-year. We continue to expect growth in auto memory demand for the second half of calendar 2023, driven by easing non-memory semiconductor supply, normalizing customer inventory levels and increasing memory content per vehicle.
The industrial markets saw early signs of recovery in fiscal Q3. Inventory levels are stabilizing at distribution partners, and at the majority of our customers. As a result, we expect an improvement in-demand in the second half of calendar 2023. We are excited about our growth prospects in this market as industrial customers continue to adopt and implement IoT, AI and machine-learning in the factory.
Now turning to industry outlook. Our expectations for calendar 2023 industry bit demand growth have been further reduced to low-to mid-single-digits in DRAM and to-high single-digits in NAND, which are well below the expected long-term CAGR of mid-teens percentage range in DRAM and low 20s percentage range in NAND. While the AI driven demand has been stronger than our expectation, three months ago, the PC, smartphone and traditional server demand forecasts are now lower. We continue to expect stronger industry bit shipments for DRAM and NAND in the second half of the calendar year, driven by secular content growth and continued improvement in customer inventories.
While the industry demand forecasts for calendar 2023 is now lower, the significant supply reductions across the industry have started to stabilize the market. We see both DRAM and NAND year-over-year supply growth to be negative for the industry in calendar 2023 as utilization and capex cuts across the industry, impact supply growth. While supply-demand balance is improving, due to the excess inventory, profitability and cash-flow will remain extremely challenged for some time. Market recovery can accelerate, if there is further reduction in industry production and these cuts are sustained well into calendar 2024.
In response to the industry environment Micron has taken decisive actions to bring our supply back-in balance with demand. We expect Micron's year-on-year bit supply growth to be meaningfully negative for DRAM. We also expect to produce fewer NAND bits in calendar 2023, then in calendar 2022. Our fiscal 2023 capex plan of $7 billion is down more than 40% from last year with WFE down more than 50%. We continue to expect fiscal '24 WFE to be down year-on-year.
Recently, we have further reduced wafer starts to approach 30% in both DRAM and NAND. We currently expect to reduce wafer starts will continue well into calendar 2024 as we remain focused on managing down our inventories and controlling our supply.
I will now turn it over to Mark.
