Aaron Rakers

Aaron Rakers from Wells Fargo inquired about the significant 76% sequential increase in inventory, seeking clarification on whether it reflected strategic purchases, purchase obligations, or procurement for hyperscale customers, and its relation to the component pricing environment.

Answer

CEO Charles Giancarlo stated that Pure Storage leverages internal forecasts for market conditions. CFO Tarek Robbiati attributed the inventory increase to tariff mitigation purchases earlier in the year and strategic positions taken in key parts due to tightening hardware component markets, noting that overall inventory levels remain low relative to business size.

Aaron Rakers from Wells Fargo asked about the significant 76% sequential increase in inventory, seeking clarification on whether it was due to strategic purchases, purchase obligations, component pricing, or procurement for hyperscale customers.

Answer

CEO Charles Giancarlo highlighted Pure Storage's strong supply chain. CFO Tarek Robbiati attributed the 76% sequential inventory increase to tariff mitigation purchases and strategic component acquisitions to prevent supply chain disruptions, emphasizing that current inventory levels are low relative to the business size.

Aaron Rakers of Wells Fargo Securities asked for an update on the Meta relationship, seeking details on procurement cycle visibility, confidence in scaling to double-digit exabytes next year, and the expected margin profile of the business.

Answer

CEO Charles Giancarlo confirmed the relationship with Meta is progressing as planned and on schedule. He reiterated that the revenue model is effectively a royalty or software-based stream, resulting in a gross margin profile of over 90%, as Meta procures the physical hardware directly from the supply chain.

Aaron Rakers of Wells Fargo Securities, LLC asked about the progression of the hyperscale opportunity, how the business model might be evolving, and the status of discussions with other potential hyperscale customers.

Answer

CEO Charles Giancarlo and CTO Rob Lee confirmed the initial hyperscaler engagement is progressing as expected through advanced testing and implementation planning. They also noted that conversations with other hyperscalers have accelerated and are moving forward at a faster pace.

Aaron Rakers asked for clarification on the financial mechanics of the hyperscaler deal, specifically if it's a licensing model for Purity software and DirectFlash modules, and whether the customer is expected to transition fully to flash or maintain a hybrid environment.

Answer

CEO Charles Giancarlo confirmed it is a licensing model for both the software and the hardware design, with the customer purchasing hardware directly from their integrator. He stated that while Pure proved its value at the hard disk price-performance level, the expectation is that the customer will use this consistent architecture across the vast majority of their storage needs, including higher-performance tiers currently using SSDs.

On behalf of Aaron Rakers, Michael Senoff asked for more color on the hyperscaler opportunity, inquiring about any changes in the engagement and the biggest hurdles remaining to finalize a deal.

Answer

CEO Charles Giancarlo stated there is no single large hurdle, but rather many small ones related to aligning business models, logistics, and pricing. CTO Rob Lee added that the engagement has progressed well, moving from initial proof-of-concept to extensive testing of a co-engineered solution, alongside detailed contractual negotiations.

Aaron Rakers inquired about NetApp's component environment, including potential pricing changes, supply constraints, strategic purchases, and the visibility of pricing dynamics underpinning the gross margin outlook.

Answer

Wissam Jabre (CFO) confirmed visibility on component prices until the fiscal year-end, noting Q2 product margin exceeded the long-term model. He discussed managing volatility with a capable supply chain team and the favorable mix from high-growth cloud business. George Kurian (CEO) added that commodity price increases are typically passed through to customers.

Aaron Rakers with Wells Fargo asked about the component environment, specifically regarding potential pricing, supply constraints, strategic purchases, and visibility on pricing dynamics underpinning the gross margin outlook.

Answer

CFO Wissam Jabre explained that NetApp has secured component prices with visibility until the end of the fiscal year, noting that product gross margin was slightly better than the long-term model in Q2 and is expected to remain stable. He also mentioned that commodity prices, like NAND, could pose a headwind in fiscal 2027, but the company's mix shift towards high-margin cloud business and Keystone is favorable. He added that NetApp would reconsider pricing if faced with higher commodity costs.

Aaron Rakers asked for more details on the Q3 sales execution issues, the specific actions being taken to correct them, and the expected timeline for seeing results from these efforts.

Answer

CEO George Kurian explained that several large deals slipped in the final weeks of the quarter due to extended procurement cycles. He stated that NetApp has instituted a higher level of scrutiny on deal progression and closing plans, noting that many of the slipped deals have already closed in Q4. Kurian expects this enhanced discipline to become part of the normal course of business going forward.

Aaron Rakers sought clarification on the mix of first-party and marketplace services within the public cloud business and questioned why growth wouldn't accelerate more significantly.

Answer

CFO Mike Berry clarified that first-party and marketplace services were a little less than half of cloud revenue at the recent Analyst Day, and that a separate 80% figure referred to the broader category of consumption revenue. He reiterated confidence in the growth outlook as the business mix continues to shift toward these high-growth services.

Aaron Rakers inquired about the double-digit demand growth in the traditional server business, its relation to revenue, the status of the upgrade cycle for the aged installed base, and whether double-digit growth is a sustainable baseline for fiscal 2027.

Answer

COO Jeff Clarke confirmed double-digit demand growth, which built backlog, noting improvements in North America and continued double-digit international demand. He highlighted ongoing data center modernization, consolidation, and a significant opportunity with 70% of the installed base being older generation servers, expecting the cycle to continue into next year.

Aaron Rakers inquired about the discrepancy between double-digit demand growth and revenue for the traditional server business in Q3, the current stage of the upgrade cycle for the aged installed base, and whether double-digit growth is a sustainable baseline for FY27.

Answer

Jeff Clarke (COO, Dell Technologies) confirmed double-digit demand growth for traditional servers, noting that the P&L did not track it, resulting in backlog build. He highlighted improved demand in North America and double-digit international demand for two consecutive quarters. Clarke emphasized ongoing data center modernization and consolidation, reflected in increasing TRUs and content per server. He sees significant opportunity with roughly 70% of the installed base still being older generation servers, expecting the upgrade cycle to continue into next year.

Aaron Rakers from Wells Fargo inquired about Dell's capacity to exceed its newly raised $20 billion AI server shipment forecast for the year, considering the strong Q2 performance and the upcoming Grace Blackwell Ultra product cycle.

Answer

Jeff Clarke, Vice Chairman and COO, acknowledged the strong momentum, having shipped $10 billion in AI servers in the first half alone. He noted that while the demand pipeline is robust and growing, particularly with enterprise and sovereign opportunities, large-scale deployments face complexities like facility readiness and supply chain transitions. Clarke stated the $20 billion guide is their best current estimate but feels like a 'plus,' and Dell has adequate manufacturing capacity to meet further demand.

Aaron Rakers asked how the margin profile of Blackwell-based rack scale configurations compares to the Hopper product cycle and what levers Dell has to improve AI server margins through 2025.

Answer

COO and Vice Chairman Jeffrey Clarke confirmed that Blackwell margins are currently lower than Hopper's due to the large, competitive nature of early deals. He outlined margin improvement levers including value-add from system design and engineering, supply chain scale, attaching networking and storage, and providing services like installation and deployment. He reiterated that AI servers are margin rate-dilutive but dollar-accretive, and noted that enterprise AI margins are better than those for large clusters.

Aaron Rakers of Wells Fargo asked about the expected shipping cadence for Blackwell in volume and sought details on the potential product mix and margin profile for the GB200 systems.

Answer

Executive Jeffrey Clarke revealed a significant and rapid shift in Q3 orders toward the Blackwell GB200 design, which now represents a large part of the backlog. He confirmed that Dell was the first to ship a GB200 design, the product is currently in production, and it will continue to ramp.

Aaron Rakers asked for an assessment of the traditional server market, focusing on the pipeline of opportunity and the expected duration of the current demand recovery cycle.

Answer

COO Jeffrey Clarke described strong momentum, with five consecutive quarters of sequential growth. He identified three key drivers: 1) an aging installed base after a long digestion period; 2) customers consolidating older servers to create space and power for new AI infrastructure; and 3) workload repatriation from public cloud back to on-premise servers.

Aaron Rakers asked about Keysight's long-term operating margin target of 31%-32% by fiscal 2026, inquiring if it's still achievable given the recent downturn and new acquisitions, or if it's now a longer-term goal. He also asked for thematic thoughts on Silicon Photonics, when volume deployments are expected, and how Keysight sees this trend impacting scale-up/scale-across networking and AI architectures.

Answer

CFO Neil Dougherty clarified that the 31%-32% operating margin target for 2026 was taken off the table due to the business downturn, and it will take longer to climb back from current 26% levels, despite strong incrementals and acquisition value capture. President and CEO Satish Dhanasekaran reiterated that the fundamental tenets of Keysight's value creation remain intact. Kailash Narayanan, President of the Communication Solutions Group, noted strong demand for high-speed silicon optics driving R&D and validation activities, with concurrent design refreshes at 400 Gb, 800 Gb, and 1.6 Terabit. Jason Kary, President of the Electronic Industrial Solutions Group, added that AI-driven investments are benefiting semi businesses, with double-digit systems sold to foundry customers for Silicon Photonics, and continued growth expected as capacity expands from R&D to commercial production.

Aaron Rakers questioned Neil Dougherty about Keysight's target of attaining a 31-32% operating margin by fiscal 2026, which was set at the 2023 analyst day. He asked if this target was still achievable in FY2026, given the current operating incremental leverage and the layering in of acquisitions. He also asked about the company's thoughts on Silicon Photonics, specifically when volume deployments are expected from the market, appreciating the R&D side.

Answer

Neil Dougherty, CFO, clarified that the 31-32% operating margin target for FY2026 was taken 'off the table' due to the business downturn in 2023-2024, expecting it to take longer to climb back to those levels. Satish Dhanasekaran, President and CEO, added that the fundamental tenets of their value creation remain intact. Kailash Narayanan, President of the Communication Solutions Group, explained that Silicon Photonics demand is driving both R&D and validation activities. Jason Kary, President of the Electronic Industrial Solutions Group, confirmed AI-driven investments in semi businesses, with double-digit system sales to foundry customers on the Silicon Photonics side, expecting continued growth as capacity expands from R&D to commercial production.

Aaron Rakers of Wells Fargo asked if the long-term 5-7% revenue growth model is still appropriate for fiscal 2026 and questioned the mix and durability of AI-driven demand within the wireline business.

Answer

CFO Neil Dougherty expressed bullishness for 2026 but noted a significant caveat around tariffs and the macroeconomic environment. CEO Satish Dhanasekaran affirmed his conviction in the long-term AI opportunity, stating that wireline is the primary beneficiary and is expected to finish the year with strong double-digit growth, driven by an expanding ecosystem of customers.

Aaron Rakers asked about the company's incremental margin profile in light of the new guidance and how to think about the path back to a 30%+ operating margin. He also inquired about the durability of the improving wireless business.

Answer

CFO Neil Dougherty confirmed the company targets a 40% operating incremental on growth above 5%, but noted that new tariffs will impact the ability to deliver on that in the short run. CEO Satish Dhanasekaran characterized the wireless business as stable, with strength in network infrastructure (Open RAN, 6G research) offsetting softness in some smartphone-related segments.

Aaron Rakers asked about the company's incremental operating margin potential with revenue growth above 5% and the path back to a 30%+ margin. He also inquired about the durability of the improving wireless business.

Answer

CFO Neil Dougherty reaffirmed the target of a 40% incremental operating margin on revenue growth above 5% but cautioned that new tariff costs will be a short-term headwind. CEO Satish Dhanasekaran characterized the wireless business as stable, with strength in network infrastructure R&D (Open RAN, 6G) offsetting softness in some smartphone-related areas.

Aaron Rakers asked if the EISG segment's gross margin could return to the mid-60% range and questioned the feasibility of achieving the 31-32% long-term operating margin target by 2026 given the current growth profile.

Answer

CFO Neil Dougherty responded that EISG's gross margin recovery is mix-dependent and will take time. He acknowledged that achieving the 31-32% operating margin target by fiscal 2026 is unlikely with a 5% growth rate, but it remains the goal, with potential acceleration from a market bounce-back or acquisition synergies. CEO Satish Dhanasekaran noted that despite gross margin differences, both segments achieved 27% operating margins this quarter due to organizational synergies.

Aaron Rakers of Wells Fargo Securities inquired about the recovery in Keysight's wireless business and whether the company's growth outlook for the segment has evolved. He also asked about confidence in the Aerospace and Defense (ADG) growth profile amid potential federal spending uncertainties.

Answer

Executive Satish Dhanasekaran conveyed growing confidence in the wireless business following two quarters of stabilization, driven by infrastructure activity in Open RAN and 6G research, and projected continued stability for 2025. Regarding ADG, Dhanasekaran affirmed the long-term trajectory is predictable due to budget clarity and strong investment in defense modernization, expressing confidence despite potential short-term uncertainties from administration changes.

Aaron Rakers inquired if NVIDIA's view on the role of AI ASICs or dedicated XPUs in architecture buildouts has changed, especially given the Anthropic deal and the broad customer base.

Answer

Jensen Huang, President and CEO, NVIDIA, emphasized that modern AI systems require entire racks and multiple switch types, not just single chips, due to increased complexity and model diversity. He highlighted NVIDIA's five key advantages: accelerating every transition phase, excelling at every AI phase, running every frontier AI model, ubiquitous presence, and diverse offtake, favoring NVIDIA's versatile GPU architecture over dedicated ASICs.

Aaron Rakers asked if NVIDIA's view on the role of AI ASICs or dedicated XPUs in architecture buildouts has changed, particularly in light of deals like the one with Anthropic, or if the shift has further favored GPU architecture.

Answer

Jensen Huang, President and CEO, NVIDIA, emphasized that the complexity and diversity of AI models have exploded, making a versatile architecture crucial. He outlined five key advantages of NVIDIA's GPU architecture: accelerating every phase of AI transition, excelling at every phase of AI (pre-training, post-training, inference), running every frontier AI model, ubiquitous deployment (cloud to edge), and providing diverse offtake for customers, which ASICs cannot match.

Aaron Rakers of Wells Fargo asked about the new Spectrum XGS networking technology, inquiring about its market opportunity and how it fits into the broader Ethernet portfolio, which is now on a run rate exceeding $10 billion annually.

Answer

Founder, President, CEO & Director Jensen Huang detailed NVIDIA's three-tiered networking strategy: NVLink for 'scale-up' within a compute node, InfiniBand and Spectrum-X for 'scale-out' across a data center, and the new Spectrum XGS for 'scale-across' to connect multiple AI factories into a super factory. He emphasized that advanced networking is critical, as improving a multi-billion dollar AI factory's efficiency by even a small percentage yields a massive return, effectively making the networking investment free.

Aaron Rakers of Wells Fargo asked about the potential for an infrastructure replacement cycle for Hopper-era systems as newer platforms like Blackwell and Rubin are introduced.

Answer

President and CEO Jensen Huang downplayed the idea of a near-term replacement cycle, explaining that older architectures like Ampere and even Pascal remain highly utilized. Due to CUDA's programmability, the installed base can be repurposed for less compute-intensive but critical tasks like data processing and curation. This work creates the training data that is then fed to the newer, more powerful Hopper and Blackwell systems, ensuring the entire infrastructure stack remains productive and well-employed.

Aaron Rakers from Wells Fargo asked for an explanation of the sequential decline in the Networking business, given strong demand commentary, and about the progress of Spectrum-X adoption.

Answer

CFO Colette Kress attributed the sequential dip in Networking revenue to timing but emphasized the 'tremendous' year-over-year growth. She stated that demand is strong and growing, and the business will be 'right back up in terms of growing' in Q4. Kress highlighted that networking is a critical and growing component of their data center system sales, especially as they prepare for the Blackwell ramp.

Aaron Rakers asked for details on the Blackwell product cycle, specifically the expected mix of full rack-scale systems like the GB200 NVL72 versus disaggregated components.

Answer

President and CEO Jensen Huang clarified that while NVIDIA architects the entire rack, it sells disaggregated system components. He explained that partners (ODMs/OEMs) perform the final integration because each data center has unique power, cooling, and configuration requirements. This model leverages the partners' value-add in design-in, installation, and logistics, allowing NVIDIA to focus on being a technology provider rather than a system integrator.

Aaron Rakers inquired about the evolution of Cisco's AI orders, particularly within the WebScale vertical, asking if the diversity of opportunities has expanded to include super spine or scale-across engagements. He also asked for an update on the enterprise AI pipeline, specifically how the $2 billion pipeline is expected to progress through the fiscal year, given the $200 million in orders this past quarter.

Answer

Chairman and CEO Chuck Robbins clarified that the $3 billion AI number for fiscal year 2026 is a revenue projection, not orders, and that Q1 saw $1.3 billion in new orders from Hyperscalers, expecting at least double the orders from this segment compared to fiscal year 2025. He noted expanding use cases, success with pluggable optics, and emerging inferencing wins. For the enterprise, Neocloud, and Sovereign Cloud pipeline, he confirmed it exceeds $2 billion for the remaining three quarters of the fiscal year, with $200 million booked in Q1.

Aaron Rakers asked for deeper insight into AI orders, specifically the diversity and evolution of Cisco's WebScale opportunity, including engagement in super spine and scale-across. He also inquired about the progression of the $2 billion enterprise pipeline for the fiscal year, following $200 million in Q1 orders.

Answer

Chuck Robbins, Chairman and CEO, clarified that the $3 billion AI figure for fiscal 2026 is revenue, with $1.3 billion in new Q1 orders from Hyperscale AI infrastructure. He expects at least double the orders from Hyperscale customers in FY26 compared to FY25, driven by expanding use cases, the P200-based router for scale-across, and success with pluggable optics. For the enterprise, NeoCloud, and Sovereign Cloud pipeline, he confirmed over $2 billion for the remaining three quarters, with $200 million booked in Q1.

Aaron Rakers of Wells Fargo & Company questioned Cisco's FY2026 guidance, asking if the implied growth deceleration after Q1 reflects conservatism or a changing demand outlook, given the AI opportunity.

Answer

CEO Chuck Robbins clarified that the FY2026 growth trajectory is shaped by tougher year-over-year comps later in the fiscal year, not a change in demand. CFO Mark Patterson added that prior quarters had higher growth rates due to the Splunk acquisition not being fully lapped until Q4 FY2025.

Aaron Rakers of Wells Fargo asked if the recently announced large sovereign cloud opportunity was included in the current AI order book and inquired about the performance of the G200 silicon in data center switching.

Answer

CEO Chuck Robbins clarified that the sovereign opportunity with Saudi Arabia's 'Human' is not in the current order numbers and represents a significant future opportunity. He confirmed the G200 silicon is central to the systems orders, comprising two-thirds of the $600M+ in AI orders, and stated that customer demand currently exceeds Cisco's supply capacity.

Aaron Rakers of Wells Fargo asked if the recently announced sovereign cloud opportunity in Saudi Arabia was included in the current quarter's AI order book and inquired about the role of the G200 silicon as a catalyst for data center switching.

Answer

CEO Chuck Robbins clarified that the Saudi opportunity with the company 'Human' is not included in the $600 million of Q3 AI orders and is just beginning. Regarding the G200 silicon, he confirmed it is central to the AI systems orders and that customer demand currently exceeds Cisco's supply capacity.

Aaron Rakers of Wells Fargo asked how Cisco sees its AI back-end networking success translating into front-end opportunities with hyperscalers. He also inquired about Cisco's view on NVIDIA's competitive position in the Ethernet back-end networking market.

Answer

CEO Chuck Robbins explained that as enterprises pervasively deploy AI applications, it will drive demand for front-end cloud capacity, creating the link between the back-end and front-end opportunities. Regarding competition with NVIDIA, Robbins acknowledged the "co-opetition" but expressed confidence in Cisco's position, citing 40 years of Ethernet experience, customer demand for vendor diversity, and the competitive advantage of its own Silicon One intellectual property.

Aaron Rakers asked about SanDisk's assessment of the enterprise SSD market opportunity relative to hard disk drives (HDDs), and whether HDD shortages have influenced customer engagements. He also sought clarification on expected seasonality into the March quarter and how startup costs would trend going forward.

Answer

Chairman and CEO David Goeckeler views enterprise SSDs and HDDs as complementary technologies, with enterprise SSDs growing faster due to AI making more data 'warmer.' He noted that the data center market is becoming the largest segment in NAND, driving better visibility and changing market dynamics. EVP and CFO Luis Visoso clarified that startup costs are expected to decrease significantly, from $60 million last quarter to $30 million this quarter, and then to near zero. He advised caution regarding Q3 seasonality, historically seeing bids down 12-14% sequentially, though data center strength might alter the mix.

Aaron Rakers asked about Sandisk's assessment of the enterprise SSD market opportunity relative to hard disk drives, and whether HDD supply constraints have influenced engagements. He also sought clarification on Q3 seasonality, expected sequential declines, pricing trends, and the future trajectory of startup costs.

Answer

Chairman and CEO David Goeckeler views enterprise SSDs and hard disk drives as complementary technologies, with enterprise SSDs growing faster due to AI making more data 'warmer.' He noted the data center market is becoming the largest NAND market in 2026, changing dynamics for customer conversations and pricing. EVP and CFO Luis Visoso advised caution for Q3 models due to historical sequential bid declines of 12-14%, particularly impacting the consumer business, but noted that startup costs are expected to decrease from $30 million in Q2 to near zero going forward.

Aaron Rakers of Wells Fargo analyzed the implied cost per bit trend from the guidance, suggesting a potential improvement, and asked about the expected cost reduction trajectory as the BiCS 8 technology matures through the second half of the year and into 2026.

Answer

CEO David V. Goeckeler explained the company is moving away from providing explicit cost-per-bit guidance. He stated the focus is shifting from cost reduction to technology differentiation (performance, power, density) to drive profitability for capital investment, noting that cost-downs are less consistent and can lead to oversupply. However, he acknowledged that BiCS 8 will naturally deliver some cost reductions.

Aaron Rakers of Wells Fargo requested a glide path for the declining startup costs, clarification on when BIX-eight cost-downs will benefit margins, and the specific bit growth rate SanDisk assumes for the broader market.

Answer

EVP & CFO Luis Visoso projected startup costs would step down significantly in Q2 and again in Q3. He noted BIX-eight cost benefits will ramp through fiscal 2026 as its mix grows from 7% to 40-50%. CEO David Goeckeler added that they see market demand growth in the low-double-digits for 2025, rising to mid-to-high double digits in 2026, with supply expected to lag demand through 2026.

Aaron Rakers questioned the cost-per-bit trend, noting it appeared flat sequentially after adjustments, and asked about the expected cost reduction trajectory as the BiCS 8 technology node matures through the second half of the year and into 2026.

Answer

Executive David V. Goeckeler stated that the company is strategically moving away from providing explicit cost-per-bit guidance. He argued the focus is shifting from cost reduction to technology differentiation—performance, power, and density—to drive profitability needed for capital investment. While acknowledging BiCS 8 will deliver cost downs, he emphasized they are becoming less consistent and harder to achieve, and the company will not forecast them quarterly.

Aaron Rakers asked for insights into the server strength, specifically how unit growth versus ASP expansion is expected to trend through the Turin product cycle. He also inquired about AMD's updated views on the total AI silicon TAM opportunity, considering the large megawatt deployments, and how it compares to the previously stated $500 billion.

Answer

Lisa Su, Chair and CEO of AMD, stated that Turin, being more content-rich, drives ASP growth as it ramps, while also noting continued strong demand for Genoa. She described the CPU demand as broad-based, driven by AI workloads, and sees strong early pull for Venice. Regarding the AI silicon TAM, Lisa Su indicated that AMD sees the AI compute TAM 'just going up' and will provide updated, larger numbers next week at the Financial Analyst Day.

Aaron Rakers inquired about the breakdown of server strength between unit growth and ASP expansion through the Turin product cycle and beyond. He also asked for updated views on the total AI silicon TAM opportunity, considering recent large megawatt deployments, potentially exceeding the previously stated $500 billion figure.

Answer

Dr. Lisa Su, Chair and CEO, explained that Turin brings more content, leading to ASP growth, while also noting strong demand for Genoa. She highlighted broad-based CPU demand driven by AI workloads. Dr. Su indicated that the AI compute TAM is increasing, and updated numbers would be provided at the upcoming analyst day, suggesting a larger opportunity than previously estimated.

Aaron Rakers of Wells Fargo asked for more detail on the server CPU outlook within the Q3 data center guidance and inquired about AMD's current market share in enterprise servers. He also followed up on the China MI308 situation, asking if the $800 million inventory write-down could be reversed upon license approval.

Answer

CFO Jean Hu confirmed that both server CPU and MI GPU businesses are expected to grow sequentially in Q3, with the MI ramp being the most significant driver. CEO Lisa Su added that they are bullish on server CPU demand. Regarding the inventory, Jean Hu clarified that the write-down was for work-in-process materials, not finished goods, so there is no immediate inventory to ship or a simple reversal of the charge upon license approval.

Aaron Rakers asked about the server CPU growth expectations within the Q3 data center guide and inquired if the $800 million inventory write-down for MI308 could be reversed upon license approval for China.

Answer

EVP, CFO & Treasurer Jean Hu confirmed that both server CPU and MI GPU businesses are expected to grow sequentially in Q3, with MI being the larger driver. She also clarified that the $800 million charge was primarily for work-in-process inventory, not finished goods, so a reversal is not straightforward and shipments would take time to ramp even with a license.

Aaron Rakers sought to reconcile previous guidance for a flat first-half Data Center GPU performance with the new impact from China export controls. He also asked for the gross margin outlook for the second half of 2025, following the inventory charge in Q2.

Answer

CFO Jean Hu confirmed the $1.5 billion China impact is concentrated in Q2 and Q3, affecting the first-half comparison. However, she reiterated expectations for significant year-over-year growth in the Data Center business for the full year, weighted to the second half with the MI350 launch. Regarding gross margin, Hu stated that excluding the Q2 charge, the margin is around 54%. She expects a slight improvement in the second half, driven by a strong Data Center mix, which will be partially offset by continued strength in the lower-margin Client and Gaming segments.

Aaron Rakers asked for the 2025 outlook for the Instinct GPU business, including the progression of revenue in the first half versus the second half, and sought clarification on the Q1 guidance, specifically if the Data Center segment was expected to be down sequentially.

Answer

CEO Lisa Su stated that the Data Center GPU business will grow strong double digits in 2025, with the second half being stronger than the first, catalyzed by the MI350 series production ramp being pulled forward to mid-year. For Q1, Su confirmed the Data Center segment would be down sequentially, roughly in line with the corporate average of -7%, while Client and Embedded would be down more and Gaming down less.

Aaron Rakers inquired about the evolution of AMD's supply chain to support its growing Data Center GPU business into 2025. He also asked for a breakdown of the drivers behind the sequential revenue growth guided for Q4, specifically between the Data Center and Client segments.

Answer

CEO Dr. Lisa Su stated she was very pleased with the supply chain's execution, noting that while the environment remains tight, AMD has secured capacity and planned for significant growth in 2025. She clarified that the largest contributor to the Q4 sequential revenue growth is the Data Center business, followed by the Client segment, with more modest growth expected from Gaming and Embedded.

Aaron Rakers asked for clarification on Arista's gross margin guidance of 62%-63% for the quarter, specifically inquiring about the drivers behind the implied product gross margin falling below 60% and the impact of tariffs.

Answer

Jayshree Ullal, CEO and Chairperson, explained that the gross margin is influenced by a product mix, with lower margins from cloud and AI titans due to volume, balanced by higher margins from enterprise customers. Chantelle Breithaupt, CFO and SVP, emphasized this is a normal part of their mix and doesn't indicate a new model for future years.

Aaron Rakers inquired about the gross margin guidance for the current quarter (62%-63%), which implies a product gross margin below 60%, asking for an explanation of the drivers, the impact of tariffs, and if this changes future expectations.

Answer

Jayshree Ullal, Arista Networks' CEO and Chairperson, explained that the gross margin is influenced by a mix of product margins, with cloud and AI titans driving volume at lower margins and enterprise customers at higher margins. Chantelle Breithaupt, Arista Networks' CFO and SVP, reiterated that this is a normal part of their mixed conversation and does not suggest a new model for the next year.

Aaron Rakers from Wells Fargo inquired about the sovereign AI opportunity, whether it is currently contributing to results, and its potential to become a material driver in 2026 and beyond.

Answer

Chairperson & CEO Jayshree Ullal stated they are "once bitten twice shy" and are not factoring sovereign AI into 2025 numbers after a key customer in that segment fell through. However, she expressed cautious optimism for the future, noting it could be an important segment in the coming years and that they are currently engaged with a sovereign AI neo-cloud customer for potential 2026 deployment.

Aaron Rakers asked about the significant sequential increase in product deferred revenue and whether it is connected to the early deployment stages of new Etherlink platforms.

Answer

CEO Jayshree Ullal confirmed the increase is related to new AI platform deployments, comparing it to the 100-gig cycle in 2016. She explained that the complexity of integrating entire AI ecosystems, from switches to NICs and optics, extends the trial-to-production timeline to 12-18 months, which in turn impacts deferred revenue recognition.

Aaron Rakers asked for Arista's perspective on the opportunity presented by large-scale AI projects like Stargate and the company's view on the emerging sovereign AI market.

Answer

CEO Jayshree Ullal described projects like Stargate as a future trend (2026-2027) toward vertically integrated racks combining processors and networking for single-point control. She noted this will drive demand for higher network speeds (1.6T) and new technologies like liquid cooling, with Arista's team, including Andy Bechtolsheim, actively involved in these next-generation designs.

Aaron Rakers inquired about the underlying reasons for the projected gross margin decline in Arista's 2025 outlook and mid-term model, asking if it was due to customer mix or other factors.

Answer

CFO Chantelle Breithaupt confirmed that the expected margin pressure is primarily due to customer mix. She noted that they anticipate continued supply chain discipline, making mix the key variable. CEO Jayshree Ullal added that while Meta and Microsoft will likely remain 10%+ customers in 2025, it becomes harder to have customers at that concentration level as overall revenue grows, and they don't anticipate others reaching that threshold.

Aaron Rakers asked about any discernible changes in the mix within the iPhone 17 categories (Pro vs. Pro Max) relative to prior cycles. He also sought updated thoughts on the buildout of Apple's Private Cloud Compute and its future outlook.

Answer

CEO Tim Cook stated it's too early to call the iPhone 17 mix due to competitive reasons and current supply constraints on both high-end and entry models. He confirmed that Private Cloud Compute (PCC) is used for Siri queries and will continue to be built out, with a new manufacturing plant in Houston for servers. CFO Kevan Parekh added that CapEx costs in 2025 were associated with building out the PCC environment in first-party data centers.

Aaron Rakers inquired about any discernible changes in the mix within iPhone 17 categories (Pro vs. Pro Max) compared to prior cycles, given the strong demand. He also sought updated thoughts on the buildout of Apple's Private Compute Cloud (PCC) and its future role.

Answer

CEO Tim Cook stated it's too early to call the iPhone 17 mix due to competitive reasons and current supply constraints affecting both high-end and entry models. He confirmed that Apple is actively building out its Private Cloud Compute, using it for Siri queries, and noted that the manufacturing plant for PCC servers recently started production in Houston, with a planned ramp for data center use. CFO Kevan Parekh added that CapEx costs in fiscal year 2025 were associated with building out the PCC environment in first-party data centers.

Aaron Rakers asked for clarification on the impact of currency on the quarter's results and the September guidance. He also inquired if the capital expenditure run-rate should be expected to increase appreciably from its current levels as the company leans more into AI.

Answer

SVP & CFO Kevan Parekh stated that foreign exchange had virtually no impact on year-over-year results in Q3 and would be a very small tailwind in Q4. Regarding CapEx, he confirmed that investments in AI are increasing significantly and that CapEx will continue to grow substantially, though not exponentially, as part of a hybrid model that also leverages third-party infrastructure.

Aaron Rakers asked about the importance of Apple developing its own foundational AI models and how its data center spending strategy compares to competitors. He also inquired about the strategy for the new internal C1 modem and the broader opportunity in custom silicon.

Answer

CEO Tim Cook described Apple's data center approach as a hybrid strategy, utilizing third parties while also making its own investments, and confirmed they are not gating necessary spending. He stated Apple wants to have its own foundation models but will also partner where it makes sense. Regarding the C1 modem, Cook expressed excitement about shipping the first one, highlighting the benefits of vertical integration for improving product features like battery life, and described it as the start of a journey.

Aaron Rakers inquired about the systems business within the data center segment, its historical levels and future outlook, and whether Seagate's mid-20% nearline demand CAGR outlook has changed given current growth.

Answer

Seagate's Chair and CEO, Dave Mosley, indicated that the systems business is steady with a limited customer base, and the mid-20% CAGR for nearline demand is still being evaluated, with focus on bigger drives for exabyte supply. CFO Gianluca Romano confirmed systems are reported as part of data center revenue.

Aaron Rakers asked for an update on the systems business within the data center segment, specifically its historical levels and future outlook, and whether the mid-20% nearline demand CAGR outlined at the analyst day has structurally changed to a higher rate.

Answer

Dave Mosley, Chair and CEO, described the systems business as steady, not expecting significant changes in customer behavior. He noted that the mid-20% nearline CAGR is still being evaluated, with the primary focus on increasing exabyte supply through higher-capacity drives. Gianluca Romano, CFO, added that systems revenue is reported as part of the data center segment, aligning with industry practice.

Aaron Rakers of Wells Fargo asked about the reasons for the guided increase in CapEx for fiscal '26 and sought clarity on the company's plans for share repurchases and operational cash levels.

Answer

CEO Dave Mosley attributed the higher CapEx to replacing older equipment and investing in future 4TB and 5TB per platter platforms. CFO Gianluca Romano confirmed that with debt at its target level and strong free cash flow, the company will resume share buybacks, with the majority of free cash flow being returned to shareholders via dividends and repurchases.

Aaron Rakers of Wells Fargo asked if HAMR drive gross margins are accretive to the current HDD-only gross margin (estimated at ~37%) and if there are any structural barriers to reaching a 40%+ gross margin.

Answer

CFO Gianluca Romano confirmed that increased HAMR volume is a key reason for confidence in March's gross margin improvement, as it is a good contributor to profitability. CEO Dave Mosley added that he is happy with the current margin structure and that the ultimate potential depends on the demand picture, but the strategy of transitioning to higher capacities is working.

Aaron Rakers from Wells Fargo asked about the future trajectory of operating expenses beyond the December quarter's guidance and inquired about the timing for potentially re-entering share repurchase programs after addressing the upcoming debt maturity.

Answer

CFO Gianluca Romano explained that the recent OpEx increase was due to higher variable compensation, not headcount, and expects OpEx to remain relatively flat around the $280M-$285M range for the rest of the fiscal year. Regarding capital allocation, he confirmed the upcoming debt maturity will be paid with cash on hand. The priority is the recently increased dividend, followed by further debt reduction, with share buybacks being considered 'a little bit further in time.'

Aaron Rakers with Wells Fargo inquired about the shift in term duration, its impact on implied fiscal Q4 growth deceleration, and Commvault's ability to grow above the 12% TAM CAGR by taking market share in the competitive landscape.

Answer

Jen DiRico, CFO, explained that Q2 saw strong term software ARR and net new customer additions, with shorter term durations reflecting customer desire for cloud transition flexibility. She noted that deals over $100,000 increased 17% and emphasized ARR as the key growth indicator, with raised subscription and total ARR guidance. Sanjay Mirchandani, CEO, added that Commvault is taking market share by outpacing SaaS and software TAM growth, driven by customer consolidation and innovation in cyber resilience. Both executives expressed confidence in growing faster than the overall market.

Aaron Rakers asked about the shift in term duration, particularly its impact on the implied deceleration of growth into fiscal Q4, and sought clarification on the competitive landscape and Commvault's ability to grow above the 12% TAM CAGR.

Answer

Jen DiRico (CFO, Commvault) explained that Q2 saw strong term software ARR volume but a shift in term duration, with customers seeking flexibility for cloud transitions, normalizing to levels seen a few quarters ago. She emphasized ARR as the best growth indicator, noting raised subscription and total ARR guidance. Sanjay Mirchandani (CEO, Commvault) added that Commvault is taking market share, with SaaS growing significantly faster than the market and software growing double digits in a low-single-digit market. Jen DiRico reiterated the company's opportunity to outpace the 12% TAM growth.

Aaron Rakers inquired about the potential revenue contribution from the Satori Cyber acquisition and the progress of Commvault's cross-sell and upsell initiatives within its SaaS customer base.

Answer

CFO Jen DiRico stated that Satori's revenue contribution is immaterial and not factored into the guidance uplift. She noted that cross-selling is in its "early innings" but showing strong progress, with a 45% increase in customers using two or more products. DiRico highlighted that the cross-sell portion of SaaS net dollar retention increased from one-third to 40% this quarter, and security SKUs contributed 20% of net new ARR.

Aaron Rakers asked for an update on the revenue contribution and momentum from the Clumio acquisition. He also inquired about the strategy to increase the percentage of SaaS customers using multiple products, which has remained at 30%, and how the company defines success for this metric.

Answer

CFO Jennifer DiRico stated that while specific Clumio revenue will not be broken out going forward due to its integration, it accounts for about 100 basis points of the total revenue guidance for FY26. CEO Sanjay Mirchandani highlighted Clumio's unique capabilities for large AWS datasets. Regarding multi-product adoption, DiRico noted that while the 30% figure is consistent, the absolute number of customers is materially up, and the goal is to eventually have customers adopt 2, 3, or 4 products, similar to industry peers.

Aaron Rakers asked about the drivers behind the sharp acceleration in SaaS customer additions, the contribution from the Clumio acquisition, and the extent of multi-product adoption among SaaS customers. He also inquired about the ramp-up time for new partner ecosystem opportunities like AWS and Microsoft Active Directory, and the current competitive landscape.

Answer

CFO Jennifer DiRico clarified that of the 1,000 new subscription customers, approximately 200 were from the Clumio acquisition, with the rest being organic growth. She noted that about 30% of customers use two or more SaaS products. CEO Sanjay Mirchandani explained that new partnership revenue takes time to ramp as it aligns with customer buying cycles, though some offerings see quicker adoption. He emphasized that Commvault's key differentiator is its ability to provide comprehensive resilience at scale for hybrid multi-cloud enterprises, a capability he believes competitors lack.

Aaron Rakers asked for an update on Commvault's up-sell and cross-sell motion within the SaaS business and inquired about the strategic alignment with Dell's Data Domain platform.

Answer

CFO Jen DiRico explained that the strong SaaS momentum is driven by a 127% net dollar retention rate, composed of one-third cross-sell and two-thirds upsell, with particular traction in Cyber Resilience products. CEO Sanjay Mirchandani added that the Dell partnership is a work-in-progress and a long-term strategy for displacing incumbents, but declined to quantify its near-term contribution.

Aaron Rakers asked about Rambus's opportunity set in SOCAM2 modules, including expected timelines and potential dollar content. He also inquired about the PMICs' representation within the product chipset business and its ramp speed. In a follow-up, he asked how the channel discussion works with MRDIMMs.

Answer

Luc Seraphin (CEO) expressed excitement for new architectures like SOCAM2, noting JEDEC standardization is positive for Rambus. He sees opportunities for SPD Hub chips and voltage regulators (PMICs) but expects lower volumes due to tight system integration, with content opportunity still early to quantify. For PMICs, Luc Seraphin (CEO) detailed their contribution to product revenue: low single-digit in Q2, mid single-digit in Q3, and projected mid to high single-digit in Q4, noting strong momentum, especially for high-end PMICs. He expects continued quarter-over-quarter growth. Luc Seraphin (CEO) clarified that MRDIMMs will intercept next-generation AMD and Intel platforms, which are anticipated to be 16-channel solutions.

Aaron Rakers asked about Rambus's opportunity in SOCAM2 modules, including expected timing and potential dollar content. He also inquired about the current representation and ramp speed of PMICs within Rambus's product chipset business. In a follow-up, he questioned the evolution of channels per socket/CPU, the outlook for 12 to 16 channels by 2026, and how this relates to MRDIMMs.

Answer

Luc Seraphin (CEO, Rambus Inc.) welcomed new architectures like SOCAM2, noting they align with Rambus's signal and power integrity strengths. He sees opportunities for SPD Hub chips and voltage regulators, expecting JEDEC standardization to resolve issues, though volumes might not be very high. For PMICs, Mr. Seraphin detailed their ramp from low single-digit contribution in Q2 to mid-to-high single-digits in Q4, driven by a suite of new products and strong momentum, particularly with high-end PMICs for next-gen AMD/Intel platforms. Regarding channels, Mr. Seraphin confirmed AI workloads drive demand for more memory and bandwidth, making the industry's move to 12 channels positive, with 16-channel solutions from CPU vendors acting as a tailwind. He clarified that MRDIMM will intercept these next-generation 16-channel platforms from AMD and Intel.

Aaron Rakers from Wells Fargo inquired about the drivers of the 43% year-over-year product revenue growth, specifically Rambus's DDR5 RCD market share, the ramp of PMIC products, and the outlook for memory channel expansion in future server CPU platforms.

Answer

CEO Luc Seraphin confirmed that RCDs remain strong, with market share slightly above 40% and expected to grow. He noted that new chips, including PMICs, made a low single-digit contribution to Q2 product revenue, projected to grow to mid-to-upper single digits in Q3. Seraphin also affirmed that the transition from 12 to 16 memory channels on upcoming platforms from Intel, AMD, and ARM will drive further demand for DDR5 products.

Aaron Rakers inquired about customer scenarios regarding potential tariffs, signs of inventory builds, the product mix progression for chipsets (RCDs vs. PMICs), and the expected ramp timeline for MRDIMMs.

Answer

CEO Luc Seraphin stated that while there is no direct tariff impact, the company is monitoring indirect effects but has not seen customer inventory pull-ins. He confirmed that product revenue is still dominated by DDR5 RCDs, with companion chips like PMICs contributing in the low-single digits but expected to grow. He projected that MRDIMM revenue would begin to ramp in the second half of 2026, aligned with future computing platforms.

In a follow-up, Aaron Rakers asked about the impact of the companion chip ramp on product gross margins and requested an updated view on the market timing for MRDIMMs.

Answer

CFO Desmond Lynch affirmed that the company expects to maintain its long-term product gross margin target of 60% to 65%, even with the new product mix. CEO Luc Seraphin stated that interest in MRDIMM remains very high for AI servers, with the market ramp still timed for the 'Diamond Rapids' platform in the second half of 2026.

Aaron Rakers from Wells Fargo Securities asked if the Nvidia relationship, initially focused on NVLink integration with x86, could expand to include other areas like Gaudi for inference workloads. He also requested an update on the non-controlling interest (NCI) expense outlook for 2026.

Answer

CEO Pat Gelsinger clarified that NVLink serves as the hub for x86 and GPU connection, and Intel's AI strategy includes defining Crescent Island and new product lines for agentic/physical AI and inference, with further updates to come. CFO David Zinsner provided an estimate for 2026 non-controlling interest expense in the range of $1.2 billion to $1.4 billion, with a focus on minimization.

Aaron Rakers from Wells Fargo Securities asked if the Nvidia relationship, initially focused on NVLink fusion with x86, is a starting point for broader integration, potentially including Gaudi for dedicated inference workloads. He also requested an update on the non-controlling interest (NCI) expense for the current year and projections for 2026.

Answer

CEO Pat Gelsinger stated that NVLink is a 'hub' connecting x86 and GPU, and Intel is defining new product lines like Crescent Island to address agentic and physical AI, particularly on the inference side, with more updates to come. CFO David Zinsner provided an estimate for 2026 NCI expense, projecting it to be in the range of $1.2 billion to $1.4 billion, with efforts to minimize it.

Aaron Rakers asked if recapturing server market share is dependent on the Diamond Rapids product and its timing. He also requested an update on the expected financial headwind from new fab start-up costs.

Answer

CEO Lip-Bu Tan indicated that Diamond Rapids is a key part of the roadmap around 2026, which is being fine-tuned with customer validation to regain competitiveness. CFO David Zinsner confirmed that prior guidance on fab start-up costs, including a headwind of $1.3-$1.5 billion next year, remains the correct forecast.

Aaron Rakers of Wells Fargo asked if recapturing server market share is dependent on Diamond Rapids and its timing. He also requested an update on the expected financial headwind from new fab start-up costs.

Answer

CEO Lip-Bu Tan confirmed the importance of the server roadmap, including Clearwater Forest and Diamond Rapids, with a timeframe of plus or minus six months, and noted he is making leadership changes to make the data center business competitive again. CFO David Zinsner confirmed that the previously communicated financial impact of fab start-up costs (around $1.3-1.5 billion next year) remains roughly correct and will be higher in 2027-2028 as fabs hit normalized run rates.

Aaron Rakers of Wells Fargo sought clarification on whether the updated OpEx guidance includes the pending Altera divestiture and asked about Intel's strategy for rack-scale networking in the AI data center.

Answer

CFO David Zinsner clarified that the $16B OpEx target for 2026 currently includes Altera's spending; the target will be reduced by Altera's OpEx amount upon deconsolidation. CEO Lip-Bu Tan and Executive Michelle C. Holthaus outlined a rack-scale strategy leveraging XPUs, with Holthaus highlighting strong growth in IPU revenue and the unique advantage of Intel's optics-based foundry capabilities.

Aaron Rakers of Wells Fargo asked for a framework to measure success in EUV wafer mix through 2025 and for the margin difference of an EUV wafer. He also sought to confirm that the non-controlling interest (NCI) is a deduction in the non-GAAP EPS calculation.

Answer

CFO and Interim Co-CEO David Zinsner explained that the move to 18A EUV wafers provides a dramatic margin improvement, with the price increasing 3x relative to the cost increase versus pre-EUV nodes. He expects a meaningful jump in EUV mix by the end of 2025. He also confirmed that the non-controlling interest (NCI) is deducted to calculate non-GAAP EPS.

Aaron Rakers asked about the size of the external revenue within the foundry business and its expected growth trajectory. He also inquired about Intel's strategy to stabilize and recapture market share in the server CPU market.

Answer

CEO Pat Gelsinger stated that external foundry revenue is currently a modest portion of the total but is growing, with the advanced packaging business turning profitable in Q3. The long-term goal is over $15 billion by the end of the decade. Regarding server share, Gelsinger said the strategy is to first stabilize and then regain share by leveraging the strength of Xeon in AI head nodes, delivering a competitive roadmap with Granite Rapids and Clearwater Forest, and capitalizing on the market's shift toward more CPU-centric inference workloads.

Aaron Rakers inquired about a recently announced Western OEM design win for Surround ADAS, asking if it was additive to previous engagements and for details on timing of volume contributions. He also asked about the gross margin impact of increasing EyeQ5 volumes and the expected inflection as EyeQ6 ramps up in 2026.

Answer

Nimrod Nehushtan, EVP of Business Development and Strategy, clarified that the design win is for a second Surround ADAS program with a new leading Western OEM, expected to contribute significant volumes across multiple vehicle models, with more details to follow. Dan Galves, Chief Communications Officer, confirmed it was an OEM previously listed for Surround ADAS. Moran Shemesh, CFO, explained that EyeQ5 volumes, while increasing to 15% next year, have a slightly lower gross margin than EyeQ4 but are not dramatically different from EyeQ6. She noted that EyeQ6 Lite's ramp-up in 2027 would provide a tailwind to margins.

Aaron Rakers inquired about Mobileye's recent Western OEM design win for surround ADAS, seeking clarification on whether it's additive to previous engagements and its expected volume contribution timing. He also asked about gross margin trends, specifically the impact of IQ5 volumes peaking and the anticipated inflection as IQ6 volumes ramp in 2026.

Answer

Nimrod Nehushtan, EVP of Business Development and Strategy, confirmed the design win is for a second surround ADAS program with a leading Western OEM, expected to be a significant portion of their future vehicle lineup, with more details to follow. Dan Galves, Chief Communications Officer, clarified it was an OEM previously listed for surround ADAS. Moran Shemesh, CFO, explained that IQ5's lower profitability is a minor headwind, peaking at 15% of volume next year, while IQ6 Light's ramp in 2027 will provide a tailwind, with IQ6 gross margins similar to IQ4.

Aaron Rakers sought to confirm the status of the seven SuperVision and Chauffeur opportunities mentioned at the Analyst Day, and asked how the robotaxi one-time payment relates to the consumer vehicle ASP projections.

Answer

CEO Amnon Shashua confirmed the opportunities are largely still active, with some expanding in scope and a growing interest in Level 3. He clarified that the consumer vehicle ASP chart (projected to grow from $55 to $200+) is entirely separate from the robotaxi business, which represents pure upside with a five-figure upfront cost and a different revenue model.

Aaron Rakers sought clarification on the cost down targets for fiscal 2025 for DRAM (excluding HBM), DRAM (with HBM), and NAND. He also asked about the trending of average capacity for SSDs in AI servers and the expected inflection point for high-capacity drives in fiscal 2026.

Answer

Mark Murphy, EVP & CFO, clarified the fiscal 2025 cost down targets: slightly better than high single digits for DRAM front-end (excluding HBM), low single-digit percentage for DRAM (with HBM), and low teens for NAND. Sumit Sadana, EVP and Chief Business Officer, stated that average SSD capacities in AI servers are escalating rapidly, with meaningful uptake expected for high-capacity drives (e.g., 122TB, 245TB). He also noted that a shortage of hard drive storage for hyperscalers would drive more NAND usage, providing an additional tailwind.

Aaron Rakers asked for clarification on Micron's cost down targets for fiscal 2025 across DRAM (ex-HBM, with HBM) and NAND, and how these cost curves are expected to materialize with new process node transitions in fiscal 2026. He also inquired about the structural trends in the NAND market, specifically the average capacity of SSDs in AI servers and the expected inflection point for high-capacity drives.

Answer

Mark Murphy, CFO, clarified that fiscal 2025 cost downs were slightly better than high single digits for DRAM front-end ex-HBM, mid-teens for NAND, and low single-digit percentage for DRAM with HBM. Sumit Sadana, Chief Business Officer, confirmed that average SSD capacities in AI servers are escalating rapidly, driven by demand for high-capacity drives (e.g., 122 TB, 245 TB) and a projected shortage of HDD storage for hyperscalers, which will further drive NAND usage.

Aaron Rakers asked for clarification on the cost down curve for DRAM (excluding HBM and with HBM) and NAND for fiscal 2025, and how these trends might extend into fiscal 2026 with new process node transitions. He also inquired about the average capacity trending on AI servers for SSDs and the expected material increase through fiscal 2026 with the adoption of high-capacity drives.

Answer

Mark Murphy, CFO, clarified the fiscal 2025 cost downs: slightly better than high single digits for DRAM front-end (excluding HBM), mid-teens for NAND, low single-digit percentage for DRAM with HBM, and low teens for NAND cost reductions. He stated that only 2025 targets were provided. Sumit Sadana, EVP and Chief Business Officer, expects average SSD capacities in AI servers to escalate rapidly, citing the demand for 60TB, 102TB, 122TB, and 245TB drives. He anticipates meaningful uptake of these high-capacity drives, driving higher average capacities, and sees an additional tailwind from hard drive storage shortages for hyperscalers.

Aaron Rakers from Wells Fargo inquired about the future impact of startup costs and underutilization on NAND gross margins. He also asked for an update on the LPDDR business in the data center, specifically its growth trajectory and customer base expansion.

Answer

EVP & CFO Mark Murphy explained that underutilization costs are now absorbed into inventory due to structurally lower NAND capacity, while fab startup costs will increase modestly through 2026 but are offset by a higher revenue outlook. EVP and Chief Business Officer Sumit Sadana highlighted Micron's sole-source position in data center LPDDR and expects broader adoption as customers focus on power efficiency.

Aaron Rakers inquired about the future impact of startup costs and NAND underutilization on gross margins over the next few quarters. He also asked for an update on the LPDDR business in the data center, specifically regarding customer base expansion and market broadening.

Answer

Executive VP & CFO Mark Murphy explained that underutilization costs are now absorbed into inventory and that while startup costs for new fabs will increase into 2026, the impact is lessened by a stronger revenue outlook. EVP and Chief Business Officer Sumit Sadana added that interest in LPDRAM in the data center remains high due to power consumption concerns, and Micron expects penetration to grow as it engages with multiple customers.

Aaron Rakers sought clarification on the projection for flat all-in DRAM costs, specifically asking if HBM was included. He also inquired about Micron's ability to maintain its LPDDR5X share in AI servers with the transition to the SOCAMM architecture and the potential for an ASP uplift.

Answer

CFO Mark Murphy confirmed that the projection for flat all-in DRAM costs for the year does include HBM. Chief Business Officer Sumit Sadana stated that while he couldn't comment on specific ASPs, Micron has a significant lead in LPDRAM for the data center and will be the first to have volume production of SOCAMM, developed in partnership with NVIDIA. He highlighted that high-capacity DIMMs plus LPDRAM crossed the $1 billion quarterly revenue milestone.

Aaron Rakers from Wells Fargo inquired about the customer concentration within the custom XPU business and the timeline for new design wins to begin contributing to revenue. He also asked about Marvell's perspective on the 'scale across' DCI opportunity.

Answer

Chairman & CEO Matt Murphy stated that the 18 design wins discussed at their AI Day will ramp over the next 18-24 months, diversifying the revenue base beyond the initial ramping sockets. Sandeep Bharathi, President of the Data Center Group, added that Marvell is heavily investing in scale-up switches for Ethernet and UA-Link fabrics, leveraging key IP, and expects to see momentum in the next couple of years.

Aaron Rakers of Wells Fargo asked for clarification on Marvell's fourth hyperscaler engagement, questioning if it was a compute opportunity or the previously mentioned custom NIC. He also inquired about the opportunity and potential timing for custom logic related to HBM4.

Answer

Matthew Murphy, Chairman and CEO, confirmed the engagement with the fourth hyperscaler is the publicly announced custom NIC. Regarding HBM, he described it as a critical IP for next-generation accelerators, highlighting benefits such as die area reduction and improved throughput. He noted strong engagement with both memory partners and hyperscale customers on this front.

Aaron Rakers inquired about the data center switching opportunity, asking how Marvell expects the 51.2T silicon cycle to ramp and what the opportunity is to participate in AI fabric build-outs.

Answer

CEO Matt Murphy stated that Marvell is well-positioned for the 51.2T cycle and is starting production with a lead customer this year. He noted strong interest in their current products and future roadmap. While acknowledging Marvell is a smaller player today, he sees the switching platform as highly strategic and complementary to their leadership in interconnect and custom silicon, but declined to size the opportunity yet.

Aaron Rakers inquired about the future pace of share repurchases given the strong free cash flow and reduced debt. He also asked for the ASP per terabyte metric, which was absent from the earnings release.

Answer

CFO Kris Sennesael highlighted the strong free cash flow ($675M) and confirmed the company can increase the pace of its share repurchase program over time. CEO Irving Tan addressed the second question, stating that ASP per terabyte was down low-single digits in the quarter, primarily due to product mix.

Aaron Rakers of Wells Fargo asked for details on the anticipated impact of tariffs mentioned in the guidance and questioned if the strong gross margin performance could structurally trend into the mid-40% range, surpassing the long-term model.

Answer

CEO Irving Tan clarified that while no direct tariff impact is expected in Q4, the guidance accounts for potential demand uncertainty in the enterprise and retail segments. Interim CFO Don Bennett added that mitigation teams are in place. Regarding margins, Tan stated the 38% long-term model is a floor, and the current outperformance is driven by the high value of their technology, strong operational execution, and pricing discipline.

Aaron Rakers inquired about the scale of NAND capacity reduction, asking if the $20-$30 million in underutilization charges implies a cut similar to competitors' reported 10% reductions, and asked about the potential impact of these charges in the fiscal fourth quarter and the risk of inventory write-downs.

Answer

CEO David Goeckeler declined to quantify the exact capacity reduction but detailed the drivers of the NAND gross margin decline, including a low-single-digit cost-per-bit increase, underutilization charges, and moderating ASP headwinds. He projected Q4 underutilization charges would be slightly higher. CFO Wissam Jabre added that while inventory is currently valued appropriately, any future risk depends on pricing dynamics.

Aaron Rakers asked for a quantification of the dissynergies from the business separation included in the December quarter guidance. He also followed up on HDD capacity, asking how quickly it can be expanded and if the company anticipates being constrained.

Answer

CFO Wissam Jabre quantified the dissynergies in the Q2 guidance at $25 million to $35 million, included within the operating expense forecast, and noted the steady-state dissynergy run-rate is expected to be around $40 million per quarter, split between the two businesses. CEO David Goeckeler added that HDD capacity decisions depend on gaining long-term demand conviction from customers over the next year, as it takes a year to build a drive from a wafer start.

Aaron Rakers from Wells Fargo asked for an update on Broadcom's capital allocation strategy, particularly the balance between shareholder returns and potential M&A, given the strong cash flow and progress with the VMware integration.

Answer

CEO Hock Tan described the policy as returning approximately 50% of the prior year's free cash flow as dividends. He stated the top priority for remaining cash is paying down debt to reach a leverage ratio below 2x Debt-to-EBITDA, though opportunistic share buybacks may occur. He noted that reducing debt also strategically preserves and expands borrowing capacity for potential future large-scale M&A.

Aaron Rakers of Wells Fargo asked for an update on Broadcom's capital allocation strategy, specifically the priorities between capital returns like dividends and buybacks versus potential M&A, given the strong cash flow and progress on the VMware integration.

Answer

CEO Hock Tan outlined a clear capital allocation framework. The first priority is the dividend, set at 50% of the prior year's free cash flow. The next priority is using remaining cash flow to pay down debt towards a target of under 2x debt-to-EBITDA. He noted that share buybacks, like the $4.2B in Q2, are done opportunistically. Deleveraging also preserves borrowing capacity for potential future large-scale M&A.

Aaron Rakers asked about the potential for Broadcom to deepen its participation in NVIDIA's upcoming Blackwell product cycle, specifically within optical components and DSPs.

Answer

President and CEO Hock Tan stated that Broadcom is not directly participating in NVIDIA's product roadmap. However, he acknowledged that Broadcom's base technologies, such as optical components and DSPs, are foundational and could be used to enable the Blackwell ecosystem, and they are happy to be part of that ecosystem.

Aaron Rakers asked about the competitive landscape for AI servers, highlighting the difference between HPE's $1.1 billion in new orders and a competitor's much larger figure, and questioned HPE's go-to-market focus.

Answer

CEO Antonio Neri explained that HPE is focused on all market segments but participates with discipline, prioritizing gross margin accretion and working capital management. He suggested that HPE chose not to participate in a couple of very large, low-margin opportunities. He reiterated that HPE is "all in" on the enterprise and sovereign AI markets and will be selective in the service provider space to sustain its target 10% operating margin.

Aaron Rakers inquired about the gross margin profile of AI server revenue, asking if transitory inventory charges were a factor and how margins might evolve with the transition from Hopper to Blackwell GPUs.

Answer

CEO Antonio Neri explained that the AI market consists of three segments with different margin profiles: time-to-market-focused service providers, sovereign nations, and enterprise customers using older tech. He noted that higher-than-normal AI inventory of prior-generation GPUs will take longer to sell to enterprise clients, creating a working capital pressure that compounded the Q1 margin issue, rather than it being an issue with AI deal pricing itself.

Aaron Rakers asked for an update on HPE's positioning for the upcoming NVIDIA Blackwell platform cycle, specifically regarding the GB200 NVL rack scale system, and when HPE anticipates shipping these systems in volume.

Answer

President and CEO Antonio Neri stated that service providers and model builders are leading the adoption of Blackwell technology. He confirmed HPE will be shipping these systems in volume in 2025, with growth expected to begin in the first half of the calendar year. He differentiated this from the enterprise segment, which he believes will adopt N-1 or N-2 generation technology like the H100, prioritizing simplicity and time-to-value over having the absolute latest GPU.

Aaron Rakers sought clarification on the shipping status of the Blackwell GB200 NVL72, asking if it is shipping today and if it's a significant factor in the current quarter's guidance. He also asked for the specific variables that drove the sequential change in gross margin and the potential margin impact from higher factory utilization rates.

Answer

CEO Charles Liang clarified that for the GB200 NVL72, Super Micro's position is similar to competitors: the solution is ready, and they can ship once they receive support and allocation from NVIDIA. He noted strong demand for their B200 platforms, especially the 4U DLC model, which they are ready to ship now. CFO David Weigand explained the Q2 margin decline was due to customer/product mix with more end-of-life products and higher R&D spending on advanced chips. Liang estimated that higher utilization could impact margins by '20, 30 point' (likely basis points), while Weigand reiterated a prior estimate of 1-2 points of savings from manufacturing in Asia.

Aaron Rakers sought to bridge the gap between the initial $6-7 billion guidance and the actual result, asking if the shortfall was solely due to the new product transition and whether a previously mentioned $800 million sales push-out had closed. He also asked about the inventory outlook and the timeline for appointing a new auditor.

Answer

CEO Charles Liang attributed the revenue gap primarily to customers awaiting new chips, with a possible minor impact from the 10-K filing delay. He expects the current $5 billion inventory level to be maintained to support future growth. CFO David Weigand stated there was no new update on the timeline for appointing an auditor but affirmed they are working diligently on it.

Aaron Rakers of Wells Fargo asked if HP has data to size the aged Windows installed base for the upcoming refresh cycle. He also questioned why the PC growth forecast doesn't seem to explicitly layer an ASP benefit from AI PCs on top of unit growth.

Answer

CEO Enrique Lores confirmed that HP uses extensive analytics, shared with Microsoft, to assess the refresh opportunity. CFO Karen Parkhill and CEO Enrique Lores clarified that an ASP benefit is indeed expected in FY25, driven by a richer mix of premium and AI PCs, and that this is factored into the overall revenue guidance where revenue growth is expected to outpace unit growth.

On behalf of Aaron Rakers from Wells Fargo, a question was asked about the market positioning of GPT-in-a-Box 2.0 relative to full-stack solutions with NVIDIA's AI Enterprise software. A second question asked if win rates have improved with the new Dell and Cisco partnerships.

Answer

CEO Rajiv Ramaswami positioned GPT-in-a-Box as focused on cost-effective AI inferencing on smaller enterprise clusters, distinguishing it from massive training clusters, and noted they partner with NVIDIA. Regarding partnerships, he said Cisco has been a good contributor to new logos and it's still early for the Dell relationship. He emphasized that storage refresh cycles, more so than server refreshes, create key insertion opportunities.