Krish Sankar

Krish Sankar from TD Cowen asked if rising NAND prices are a catalyst for the hyperscaler business model shift, driven by hyperscalers' desire to avoid volatility, and whether a future crash in NAND prices would lead to another business model change.

Answer

CEO Charles Giancarlo indicated that rising NAND prices are not a primary catalyst for the hyperscaler business model shift, attributing it more to diverse purchasing preferences among hyperscalers. CFO Tarek Robbiati added that this necessitates a more nuanced model with varying cost of goods sold, acknowledging hyperscalers' desire for commodity price insulation, albeit with limitations.

Krish Sankar from TD Cowen inquired if rising NAND prices were a catalyst for the hyperscaler business model shift, suggesting hyperscalers might seek insulation from volatility, and whether the model would revert if NAND prices were to crash.

Answer

CEO Charles Giancarlo stated that rising NAND prices don't appear to be a major factor in the current business model discussions, as different hyperscalers have varying purchasing preferences. CFO Tarek Robbiati added that the shift involves a mix of options with different costs, and while hyperscalers desire insulation from commodity prices, there are limits to what Pure Storage can provide.

Krish Sankar of TD Cowen questioned the decline in the RPO growth rate from around 30% a year ago to 14% in the current quarter, asking about the underlying causes and implications for the core business.

Answer

CFO Kevan Krysler explained the RPO slowdown is a direct result of the year-over-year decline in Evergreen//One TCV sales for FY25. He highlighted the return to 20% growth for Evergreen//One in Q4 as a positive sign and noted that growth in this offering is factored into the FY26 revenue guide.

Krish Sankar with TD Cowen asked about the average size of the 200 AI deals closed, the current stage of enterprise AI adoption from pilot to production, and requested quantification of the growth rate for the first-party hyperscale services business.

Answer

CEO George Kurian stated that AI deal sizes vary widely, from smaller proof-of-concept to larger scale deployments, and did not provide an average. He noted that data lakes and data prep are typically pre-scale, while fine-tuning and inferencing are in various stages of proof-of-concept to production, with life sciences, financial services, manufacturing, and public sector showing broad adoption. He also confirmed that the first-party and marketplace cloud services business grew 32% year-on-year.

Krish Sankar asked about the average size of NetApp's 200 AI deals, the current stage of enterprise AI adoption (pilot-to-production journey), and the growth of first-party hyperscale services demand.

Answer

George Kurian (CEO) stated AI deal sizes vary widely, declining to provide an average, and described enterprise AI adoption stages from data prep to fine-tuning and inferencing, with life sciences and financial services showing advanced use cases. He also confirmed that the first-party and marketplace services business grew 32% year-on-year.

Krish Sankar from TD Cowen inquired about the deceleration in all-flash revenue growth to 5% and the decline in product gross margins, asking if pricing, demand, or NAND costs were the primary drivers. He also asked for a long-term view on the architecture of enterprise AI deployments.

Answer

CFO Wissam Jabre attributed the all-flash revenue slowdown to anticipated softness in the U.S. Public Sector and EMEA, which was offset by strength in the Americas. He explained that the year-over-year product gross margin decline was predominantly due to higher flash costs and projected that margins would improve going forward. CEO George Kurian detailed that recent AI wins fell into three categories: data lakes (20%), model training/fine-tuning (45%), and Retrieval-Augmented Generation (RAG) use cases, noting that the latter two typically utilize all-flash storage.

Krish Sankar from TD Cowen inquired about the deceleration in all-flash revenue growth to 5%, questioning if it was due to pricing or demand. He also asked about the decline in product gross margins and the outlook for enterprise AI adoption architectures.

Answer

CFO Wissam Jabre explained the all-flash revenue slowdown was anticipated due to softness in the U.S. Public Sector and EMEA, but he expects this trend to not continue. He attributed the year-over-year product gross margin decline primarily to higher NAND costs, stating Q1 was the low point and margins should improve. CEO George Kurian detailed that AI wins were split between data lakes (20%), model training (45%), and RAG/inferencing use cases, which typically utilize all-flash storage.

Krish Sankar from TD Cowen asked about the dynamics of the non-all-flash (hard drive) portion of the hybrid cloud business. He also inquired about the implied product gross margin within the FY26 guidance and requested an update on NetApp's public sector exposure.

Answer

CEO George Kurian explained that the hard drive segment has declined as flash replaces performance drives and expects it to stabilize at a lower level for use cases like backup and archival. CFO Wissam Jabre reiterated that Q1 product gross margin will be flat to Q4 before improving quarterly, but did not provide a specific full-year figure. George Kurian stated that the US public sector represents a low-teens percentage of business, with federal comprising 75-80% of that.

Krish Sankar from TD Cowen inquired about the dynamics of the non-all-flash (hard drive) portion of the business, asking if it's in secular decline. He also asked for the implied product gross margin within the FY26 guidance and the company's current public sector exposure.

Answer

CEO George Kurian explained that the hard drive segment has declined as flash replaces high-performance drives and expects it to stabilize at a smaller number, primarily serving backup and archival use cases. He also noted that US public sector exposure is in the low teens percentage of revenue. CFO Wissam Jabre reiterated that Q1 product gross margin would be flattish with Q4, with gradual improvement expected thereafter, but did not provide a specific full-year target for the sub-segment.

Krish Sankar from TD Cowen inquired about the business dynamics of the non-all-flash (hard drive-based) segment. He also asked for the implied product gross margin within the full-year fiscal 2026 guidance and for an update on NetApp's public sector revenue exposure.

Answer

CEO George Kurian explained that the hard drive segment continues to decline as flash replaces performance tiers but will eventually stabilize for use cases like backup and archival. CFO Wissam Jabre reiterated that Q1 product gross margin will be flat to Q4 with gradual improvement thereafter. George Kurian specified that the US public sector represents a low-teens percentage of revenue, with federal business comprising 75-80% of that.

Krish Sankar inquired about the sequential growth expectations for Kulicke and Soffa's General Semi, Memory, and Auto-Industrial segments in the December and March quarters, including any anticipated seasonality.

Answer

Lester Wong, Interim CEO and CFO, confirmed expected sequential growth across all three segments for the December quarter, with General Semi and Memory showing strong utilization over 80%. He projected the March quarter to be flat to the December quarter, with no significant seasonality. Krish Sankar also asked about the status of K&S's FTC plasma solution for chip-to-wafer compared to competitors at a leading foundry. Lester Wong stated that K&S remains the sole provider for high-volume production at the foundry, offering both formic acid and plasma options, and believes their FTC solution is best in class.

Krish Sankar inquired about the sequential growth expectations for Kulicke & Soffa's three segments (general semi, memory, auto-industrial) into the March quarter, including any anticipated seasonality. He also asked for an update on Kulicke & Soffa's fluxless thermal compression (FTC) plasma solution for chip-to-wafer, specifically regarding a competitor's recent announcement and K&S's position at a leading foundry.

Answer

Lester Wong, Interim CEO and CFO, confirmed strong sequential growth for general semi and memory, with utilization rates over 80%. He noted auto-industrial is lagging but expects sequential growth into Q1, anticipating the March quarter to be flat with no significant seasonality. Regarding FTC, Mr. Wong stated Kulicke & Soffa remains the sole high-volume production provider at the foundry, highlighting their solution's formic acid and plasma capabilities, and expressing confidence in its best-in-class competitiveness.

Krish Sankar of TD Cowen asked about the revenue outlook for fiscal 2026, questioning if revenue would remain near the $170 million level. He also inquired about the impact of Intel's CapEx cuts on the fluxless TCB business and sought details on the planned TCB system shipment for HBM applications by the end of calendar year 2025.

Answer

President & CEO Fusen Chen projected that fiscal 2026 revenue would be flattish compared to 2025, supported by healthy industry utilization and new product growth in areas like clip attach, advanced dispense, and TCB. Regarding Intel, he confirmed that while engagement remains healthy, revenue from them would be down year-over-year. For HBM, Chen stated that a TCB system is expected to ship to one of two key customers by the end of calendar year 2025 to support the industry's transition to fluxless processes.

Krish Sankar of TD Cowen asked for clarity on the fiscal 2026 revenue outlook, the potential impact of Intel's capital expenditure cuts on the fluxless TCB business, and details about the upcoming TCB system shipment for HBM applications.

Answer

President & CEO Fusen Chen projected that fiscal 2026 revenue would be 'flattish' compared to 2025, supported by high industry utilization and new product growth, despite typical seasonality. Regarding Intel, he acknowledged that revenue would be down from the prior year but noted that customer engagement remains healthy. For HBM, Chen confirmed they are working with two customers and plan to ship an initial system to one of them by the end of calendar year 2025 for the upcoming fluxless transition.

Krish Sankar of TD Cowen inquired about the fiscal 2026 revenue outlook, the potential impact of Intel's capital expenditure cuts on the fluxless bonding business, and specifics regarding the upcoming Thermal Compression Bonding (TCB) system shipment for High Bandwidth Memory (HBM).

Answer

President and CEO Fusen Chen projected that fiscal 2026 revenue would be flattish compared to 2025, supported by healthy industry utilization and new product growth. While acknowledging a revenue decline related to Intel, he noted that customer engagement remains healthy. For HBM, Chen confirmed a TCB system shipment is planned for the end of calendar year 2025 to one of two key customers for the next-generation, fluxless HBM.

Krish Sankar asked about the implications for blended ASPs, both overall and excluding mobile, given that wafer shipments were up 4% sequentially despite flat revenues, and full-year shipments are tracking around 13% on flattish revenues. He also inquired whether the expanded partnership with Silicon Labs represents a share gain from another foundry or new chip designs.

Answer

CEO Tim Breen attributed the ASP dynamic primarily to proactive price changes with specific dual-source mobile customers to gain share and increase overall profit dollars, noting this reset is complete and pricing is stable elsewhere. Interim CFO Sam Franklin emphasized focusing on margin structure, as GF grew margins despite these ASP trends. Tim Breen confirmed the Silicon Labs partnership is "absolutely a share gain," driven by their desire for a strong U.S. sourcing footprint and GF's technology partnership.

Krish Sankar from TD Cowen asked about the implications for blended ASPs and ASP ex-mobile, given that wafer shipments were up 4% sequentially despite flat revenues in Q3, and full-year shipments tracking around 13% on flattish revenues. He also inquired whether the expanded partnership with Silicon Labs represents a share gain from another foundry or new chip designs.

Answer

CEO Tim Breen explained that the ASP dynamic was primarily due to proactive price changes with specific dual-source customers in mobile to gain share and increase overall profit dollars, a reset that is now complete. He noted that outside these specific mobile customers, ASPs are stable, especially in sole-source businesses and differentiated technologies. Interim CFO Sam Franklin emphasized focusing on the margin structure, as gross margin grew year-over-year and quarter-over-quarter despite ASP trends, dispelling a pure ASP focus. Tim Breen confirmed that the Silicon Labs partnership is "absolutely a share gain," driven by Silicon Labs' desire for a strong US sourcing footprint and GlobalFoundries' technology platforms.

Krish Sankar of TD Cowen asked about the MIPS acquisition, specifically regarding RISC-V demand trends between Asia and Western companies. He also inquired about the 'China for China' strategy, asking for the partner's identity, margin profile, and details on IP protection.

Answer

CEO Tim Breen described RISC-V demand as a global phenomenon with strong interest across Asia, Europe, and the U.S. Regarding the China strategy, he stated they are not disclosing the partner, referring to the initiative as 'GF China.' He assured that the margin profile is in line with corporate averages and that robust IP protection controls, vetted by automotive-grade customers, are in place.

Krish Sankar asked for a breakdown of Sandisk's data center revenue growth, specifically how much is attributable to AI data centers and Neocloud versus traditional cloud, and if AI is the majority driver. He also sought confirmation on data center exabyte growth projections for 2026 (high 300s) and 2027 (400-600+).

Answer

Chairman and CEO David Goeckeler confirmed that the majority of the data center revenue growth is AI-driven. He projected data center exabytes to be in the high 300s for 2026, representing healthy growth over 2025, which followed 130% year-over-year growth in 2024 (around 230 exabytes) and high teens growth in 2025. Goeckeler noted that these customer demands are fundamentally changing the supply-demand dynamics.

Krish Sankar asked for a breakdown of SanDisk's data center revenue, distinguishing between AI data center/NeoCloud and traditional cloud, and inquired about the projected data center exabyte market growth for 2026 and 2027.

Answer

Chairman and CEO David Goeckeler indicated that the majority of the data center growth is AI-driven. He provided exabyte market projections, noting high 300s for 2026, building on 130% year-over-year growth in 2024 and high teens in 2025, followed by a mid-40s percentage increase in 2026. He highlighted that these healthy growth rates are causing customers to fundamentally change how they approach supply-demand dynamics.

Representing Krish Sankar of TD Cowen, an analyst asked about SanDisk's exposure to the China mobile market in light of a competitor's reported exit and questioned which end market drove the upside in Q4 bit shipments.

Answer

CEO David Goeckeler affirmed that China remains an attractive and important market for SanDisk. CFO Luis Visoso added that the Q4 bit shipment outperformance was broad-based across the entire product portfolio, not concentrated in a single end market.

Krish Sankar asked about the future implications for Arm regarding chip demand and token generation, particularly as the world shifts towards an inference-heavy environment with a greater role for edge devices.

Answer

CEO Rene Haas discussed the anticipated shift from training to inference, driving demand for edge inference solutions. He highlighted Lumix CPUs with scalable matrix extensions for AI workloads and the Meta partnership, which aims to provide scalable software solutions for efficient model portability between cloud and edge, positioning Arm uniquely to address this trend.

Krish Sankar asked for Arm's perspective on future chip demand and token generation over the next few years, particularly the implications for Arm as the industry shifts towards an inference-heavy world where edge devices are expected to play a larger role.

Answer

Rene Haas, Arm's Chief Executive Officer, discussed the anticipated shift from training to inference, driving significant demand for inference on edge devices. He highlighted Arm's unique position due to its power efficiency and scalable software solutions, exemplified by the Meta partnership, which enables efficient model portability between cloud and edge. He concluded that increased token generation translates to more compute, especially at the edge, which is highly beneficial for Arm.

Krish Sankar (question asked by Eddie) asked about the drivers behind the approximately 50% year-over-year revenue growth when excluding the largest customer, inquiring whether it was primarily due to units or content, and how to project this segment's future growth. He also asked about Skyworks' comfort level with its China market exposure, which is below 10%, and any plans to reduce it.

Answer

Philip Carter, Senior Vice President and Chief Financial Officer, stated that the growth outside the largest customer is primarily driven by strength across the broad markets business, including automotive, IoT, and edge. Phil Brace, CEO and President, added that Skyworks' focus in China remains on the premium segment and other business areas like automotive, choosing not to participate in economically unattractive, higher-volume, lower-ASP handset segments.

Krish Sankar from TD Cowen asked about the current size and growth potential of the automotive business, the relative visibility into its largest customer versus Android, and the outlook for operating expenses into next year.

Answer

VP of Investor Relations Rajvindra Gill reported that the automotive business is now tracking at approximately $60 million per quarter with significant year-over-year growth. CEO Philip Brace stated that visibility is currently strong across the board with no material difference between customer segments. He also reiterated a disciplined approach to OpEx, expecting only nominal growth targeted at core R&D.

An associate on behalf of Kris Sankar asked for a quantification of the performance gap between Skyworks' product and a competitor's product that led to the dual-sourcing decision, and how that compares to the previous year.

Answer

CFO Kris Sennesael clarified that it was not a case of a competitor significantly improving performance, but rather a competitor competing for this specific socket for the first time. He declined to provide specific technical performance metrics, stating that their part was as good as the competitor's, but not substantially better.

Krish Sankar asked if Apple could quantify the business lost due to iPhone constraints and whether different manufacturing regions contributed to these constraints. He also inquired about the potential impact of chatbots and AI-infused web services on consumer behavior within mobile app ecosystems and the App Store.

Answer

CEO Tim Cook clarified that iPhone constraints were due to underestimating demand, not manufacturing capacity, and while Apple could have sold more, they do not publicly quantify the extent. He expressed optimism about opportunities for the App Store with artificial intelligence, noting that developers are beginning to adopt Apple's on-device models, which could benefit both developers and Apple through new app features.

Krish Sankar asked if Apple could quantify the business lost due to iPhone constraints and whether different iPhone manufacturing regions contributed to these constraints. He also inquired about the potential impact of chatbots and AI-infused web services on mobile app ecosystems and the App Store.

Answer

CEO Tim Cook clarified that iPhone constraints were due to under-calling demand for iPhone 16, not manufacturing capacity, and declined to quantify lost business, noting strong demand and backorders for iPhone 17. He expressed optimism about opportunities for the App Store with artificial intelligence, as developers adopt Apple's on-device models to add new features to their apps.

Krish Sankar asked if there was a pull-in of iPhone demand during the June quarter and inquired about channel inventory levels for June and September relative to seasonal norms. He also posed a long-term question about the future of AI on edge devices, asking if LLMs could become a commodity and if the smartphone will remain the dominant device.

Answer

SVP & CFO Kevan Parekh confirmed an estimated one-point demand pull-ahead at the total company level due to tariff discussions in April. CEO Tim Cook added that iPhone channel inventory was reduced during the quarter and ended at the low end of the target range. Regarding AI, Cook described it as one of the most profound technologies of our lifetime that will affect all devices, but declined to comment on strategy specifics like commoditization.

Krish Sankar asked if Seagate is rethinking its HAMR pricing strategy to increase prices given the AI tailwind, and sought clarification on whether the tighter revenue guidance range reflects better visibility from build-to-order contracts.

Answer

Dave Mosley, Chair and CEO, noted that market pricing will adjust slowly over time, driven by the TCO proposition of higher-capacity drives. Gianluca Romano, CFO, clarified that an initial agreement to offer one customer a slightly lower HAMR price for post-qualification support is phasing out. He confirmed that the tighter revenue guidance range is indeed due to improved visibility from build-to-order contracts.

Krish Sankar asked if Seagate is rethinking its HAMR pricing strategy, moving from sharing cost benefits to increasing prices, and sought clarification on why the revenue guidance range has tightened from $150 million to $100 million.

Answer

Seagate's Chair and CEO, Dave Mosley, explained that predictable customer contracts and market adjustments will influence pricing over time. CFO Gianluca Romano clarified that a specific customer's lower HAMR pricing is transitioning away, and the tighter guidance range reflects better visibility from build-to-order contracts.

Krish Sankar asked about the implications of NAND utilization rates nearing 100%, questioning whether this might lead to a temporary drop-off in NAND orders/shipments or if a smooth transition is expected. He also posed a hypothetical question about Lam Research's EPS growth in 2026 relative to revenue growth, considering factors like China headwinds, a higher tax rate, and fixed cost leverage.

Answer

Doug Bettinger, Executive Vice President and CFO, noted that nothing goes up and to the right every period, and NAND revenue can have quarterly variability, but the overall trend is towards the $40 billion TAM. Tim Archer, President and CEO, added that Lam's equipment intensity rises above 200 layers, driving both systems and upgrades. Doug Bettinger declined to give specific EPS guidance for 2026 but confirmed a slight tax rate increase and customer mix headwind for gross margin, with fixed cost leverage being beneficial.

Krish Sankar inquired whether limited clean room space and high NAND utilization rates might lead to a temporary drop in Lam Research's NAND orders or shipments, or if a smooth transition is expected. He also posed a hypothetical question about Lam Research's EPS growth in 2026, considering expected WFE growth alongside potential gross margin headwinds from reduced China mix and a slightly higher tax rate.

Answer

EVP and CFO Doug Bettinger noted that revenue trends are not always linear, and NAND revenue experienced a slight dip in the most recent quarter, indicating variability. President and CEO Tim Archer added that Lam Research's business drivers for NAND extend beyond bit demand, as equipment intensity increases for higher layer counts, contributing to both systems and upgrades. Doug Bettinger declined to provide a specific EPS forecast for 2026 but confirmed a slight increase in the tax rate (approaching mid-teens) and a gross margin headwind from customer mix, suggesting anchoring models around the December quarter's guided range.

Krish Sankar asked for clarification on the significant growth from multinational customers in China and whether tool decisions for 2-nanometer gate-all-around (GAA) are already finalized at leading foundries.

Answer

EVP and CFO Douglas Bittinger confirmed a greater than 90% sequential spending increase from global multinationals in China but did not specify a reason. President and CEO Timothy Archer stated that while not commenting on specific customer decisions, Lam is already deeply engaged in inflections beyond 2nm, such as Moly adoption and dry resist, positioning the company for future nodes.

Krish Sankar of TD Cowen sought clarification on the growth of multinational customers in China and asked whether all tool decisions for 2-nanometer gate-all-around have been finalized at leading foundries.

Answer

EVP and CFO Doug Bittinger confirmed that spending from global multinationals in China grew over 90% quarter-over-quarter but did not attribute it to a specific reason. President and CEO Timothy Archer, without commenting on specific customer timing, stated that Lam is already focused on inflections beyond 2-nanometer, such as Moly adoption and dry resist, where the company is well-engaged with customers.

Krish Sankar referenced TSMC's past discussions about revenue opportunity per smartphone during the smartphone era. He asked if TSMC could quantify the revenue opportunity or wafer demand for TSMC from one gigawatt of AI data center capacity in today's AI-driven market. He then asked if the source of AI demand (GPU versus ASIC) matters to TSMC from a revenue or gross margin mix perspective, given TSMC's long-term forecasting and capacity planning.

Answer

CEO Che-Chia Wei stated that customers indicate a $50 billion investment for one gigawatt of AI data center capacity. However, he noted that TSMC is not ready to share specific wafer demand or revenue figures, as it involves multiple chips forming a system, not just one. Che-Chia Wei responded that for TSMC, there is no differentiation between GPU and ASIC demand. Both utilize leading-edge technologies, and TSMC supports all types, working with customers who are expected to grow strongly in the coming years.

Krish Sankar inquired about the linearity of ASML's revenues or orders in 2026, considering recent AI investment strength and long lead times, and any early implications for 2027. He also asked about the longer-term impact on EUV layer count when DRAM architecture transitions from 6F sq to 4F sq, addressing concerns about a potential negative effect.

Answer

CFO Roger Dassen stated it's too early to comment on 2027 and that orders are typically lumpy, not linear, despite healthy intake in recent quarters. CEO Christophe Fouquet clarified that the transition from 6F sq to 4F sq in DRAM is not expected to reduce EUV layer count; in fact, it's anticipated to grow, and 4F sq is not considered bad news for ASML.

Krish Sankar of TD Cowen asked what quarterly bookings run rate would be needed for ASML to gain conviction in 2026 growth. He also requested a breakdown of Q2 EUV bookings between logic and DRAM, and the shipment timing for the DUV orders booked in the quarter.

Answer

EVP & CFO Roger Dassen declined to provide a target bookings run rate for 2026, stating it would be an indirect form of guidance and that bookings are too lumpy to be a reliable short-term indicator. He disclosed that Q2 EUV bookings were skewed towards logic and that ASML is now virtually fully booked for DUV systems for 2025, with the balance of new orders shipping in 2026.

Krish Sankar asked if Micron could quantify the HBM supply opportunity for calendar 2026, assuming it's fully sold out, and if supply could be increased if demand exceeds expectations. He also inquired about the expected mix between Micron's in-house base die and the customized TSMC logic die for HBM4, and the ease of switching between the two based on customer demand.

Answer

Sanjay Mehrotra, CEO, stated that Micron is not breaking down supply volumes but expects to conclude HBM4 and total 2026 HBM supply agreements in the coming months, following HBM3E pricing and volume agreements. He reiterated Micron's industry-leading HBM4 product specifications and confirmed flexibility to manage the portfolio mix between HBM and non-HBM, considering ROI and disciplined investments, due to fungibility at the front end (using the same 1-beta wafer) and sufficient assembly and test capacity. He clarified that HBM4 uses Micron's internal base die, while HBM4E (a 2027 product) will offer standard and customized products in partnership with TSMC for the base logic die. He expects customization in HBM4E to yield higher gross margins, emphasizing that Micron's own CMOS in its HBM base die provides unique advantages for industry-leading performance.

Krish Sankar asked for a quantification of Micron's HBM supply opportunity for calendar 2026, assuming a fully sold-out scenario, and whether the company has the flexibility to increase HBM supply if demand surpasses current expectations. He also sought clarification on the distinction between HBM4 and HBM4E, inquiring about the expected mix of in-house versus TSMC-manufactured base dies for HBM4E, and the ease of switching between these options based on customer demand.

Answer

Sanjay Mehrotra, Chairman, President & CEO, did not quantify HBM supply volumes but confirmed HBM3E pricing and volumes are largely fixed for 2026, with HBM4 agreements expected soon. He emphasized Micron's flexibility to manage the mix between HBM and non-HBM based on ROI and disciplined investments, leveraging fungibility at the front end. He clarified that HBM4 uses Micron's internal base die, while HBM4E (a 2027 product) will offer standard and customized options with TSMC partnership, expecting higher margins from customization.

An analyst on behalf of Krish Sankar from TD Cowen asked why Micron was not providing 2026 HBM guidance at this time, unlike their approach for 2025 last year, and questioned when they expect to gain clarity on 2026 customer demand.

Answer

EVP and Chief Business Officer Sumit Sadana explained that Micron is now an established, large-scale HBM supplier, unlike last year when it was still ramping. He noted that customers are still finalizing complex 2026 platform transition plans (e.g., HBM3E to HBM4), which dictates their purchasing mix. He expressed confidence in Micron's position and deep customer engagements for future HBM roadmaps.

Speaking on behalf of Krish Sankar, an analyst asked why Micron was not providing specific 2026 HBM guidance, unlike its approach to 2025 last year, and questioned when the company expects to get clarity on 2026 customer demand.

Answer

EVP and Chief Business Officer Sumit Sadana explained that Micron is now a scaled, trusted HBM supplier with deep customer engagements for HBM4 and HBM4E. He noted that customers are still finalizing their complex 2026 platform transition timelines, which involve shifts from 8-high to 12-high HBM3E and the adoption of HBM4. As these plans solidify, agreements will fall into place, and Micron is confident in its HBM business trajectory.

Krish Sankar from TD Cowen asked about the expected die trade ratio for HBM4 and whether there is a difference in the margin profile for HBM sold to GPU versus ASIC customers.

Answer

Chairman, President & CEO Sanjay Mehrotra stated that the HBM4 trade ratio is greater than three, compared to approximately three for HBM3E, and will trend towards four with HBM4E. This increasing ratio puts pressure on non-HBM DRAM supply. Regarding margins, he explained that HBM commands high value in both GPU and ASIC-based accelerators and its value is growing, but did not provide a specific margin differentiation between the two customer types.

Krish Sankar from TD Cowen posed a two-part question, asking about the AI server mix between liquid and air cooling and its margin implications, and why the full-year EPS guidance was raised by a much smaller percentage than the revenue guidance.

Answer

Jeff Clarke, COO, answered that the AI backlog and pipeline are biased towards direct liquid cooling for large-scale deployments. CFO Yvonne McGill addressed the guidance question by explaining that the revenue increase is primarily driven by $5 billion in lower-margin AI servers, with a smaller contribution from the core business. She noted that the EPS lift is based on numerous drivers, including mix and operational efficiencies across the company.

Krish Sankar of Cowen and Company followed up on the storage attach rate for AI. He asked if Dell sees private storage companies taking market share and whether it's necessary for Dell to be designed into the AI training phase to capture the subsequent AI inference business.

Answer

Jeff Clarke, Vice Chairman and COO, explained that training and inference are distinct use cases and Dell aims to win both. He emphasized the significant data consumption in enterprise inference workloads, creating a large storage opportunity. He highlighted Dell's focus on building out a portfolio of high-performance, disaggregated storage systems like PowerScale, ObjectScale, and Project Lightning, which he believes is the bridge to improving storage attach rates for AI workloads.

Krish Sankar asked for clarification on whether China sales were pushed from Q2 to Q3 and questioned how much of the Q4 China weakness is due to the licensing assumption, as well as the potential revenue upside if licenses are approved.

Answer

CFO Brice Hill confirmed that the higher Q3 China revenue was not a push from Q2, stating that the Q2 and Q4 levels are more indicative of the expected run-rate. He emphasized that investors should not expect immediate upside to the Q4 outlook even if licenses are approved, as it would take time to plan and build the equipment. He declined to quantify the license backlog.

Krish Sankar of TD Cowen questioned the sustainability of the recovery in the enterprise segment given tepid IT spending and asked about expectations for December and March revenue trends and seasonality. He also inquired about Synaptics' potential to pursue the industrial IoT market.

Answer

SVP & CFO Ken Rizvi noted healthy visibility for the September quarter and building visibility for December, while acknowledging that the March quarter typically sees some seasonality. President & CEO Rahul Patel addressed the industrial IoT question, explaining that Synaptics' platforms, particularly its processors and mixed-signal capabilities, are designed to scale across a wide landscape from consumer to industrial applications like factory control and EV charging.

Krish Sankar from TD Cowen inquired about the semiconductor revenue outlook beyond Q3, particularly the sustainability of outperformance versus WFE given NAND lumpiness, and whether the strength in the PCB business was driven by smartphone-related pull-ins.

Answer

President and CEO John Lee explained that the base WFE portfolio is growing faster than the market, with the NAND upgrade cycle acting as a lumpy overlay that can cause quarterly fluctuations. He also noted that any pull-in for the PCB business was minimal, with strength in Q2 and the Q3 outlook being driven primarily by AI applications for both equipment and chemistry.

Krish Sankar of TD Cowen asked about the sustainability of Advanced Energy's high data center revenue run rate, questioning if it's a structural change driven by AI or related to market share gains.

Answer

President and CEO Steve Kelley confirmed that the current data center revenue level is sustainable into 2026. He attributed this to high, ongoing investment rates by hyperscalers, the increasing power needs of new GPUs which drives content growth, and AE's strong design win rate. He also highlighted the company's willingness to invest in factory capacity to support this demand.

Krish Sankar inquired about the primary sources of demand for Q3 and the second half, given the noted weakness in EUV and from a major US manufacturer. He also asked about broader gross margin pressures, such as customer pricing, beyond the immediate operational challenges.

Answer

CEO Jeffrey Andreson identified foundry logic, high-bandwidth memory, and continued NAND investment as the key demand drivers. He stated that the primary headwind to gross margin is the inability to fully ramp internal component supply to meet demand, rather than significant changes in customer pricing pressure. He added that tariff impacts are now better understood and are being passed on.

Krish Sankar of TD Cowen asked if the GPU test opportunity was aided by the networking business or a direct value proposition, and about potential share gains in ASICs. He also sought to size the robotics win in terms of units and clarify if Teradyne was supplying peripherals beyond the robotic arm.

Answer

President & CEO Greg Smith clarified the GPU opportunity is not yet a 'win' but an opportunity to compete, aided by their reputation with the customer in other segments. He noted their focus is on proving value to OSATs, foundries, and hyperscalers. For the robotics win, he declined to provide unit numbers but specified Teradyne is providing the arm and software platform, while the end customer handles end-of-arm tooling and vision systems.

Krish Sankar sought clarification on the timing of inventory pull-ins and asked about the long-term implications of edge AI on smartphone RF content, specifically regarding uplink MIMO and the adoption of Power Class 2.

Answer

SVP & CFO Grant Brown and SVP of Sales and Marketing Dave Fullwood clarified that the estimated $15-30 million inventory buffer is related to components, not finished goods, and is already factored into guidance. Regarding AI, SVP & President of Advanced Cellular, Frank Stewart, agreed that trends associated with 5G-Advanced will continue to drive the need for more complex RF solutions, including additional RF paths and expanded use of higher power levels (Power Class 2), consistent with the company's previous Analyst Day commentary.

Krish Sankar of TD Cowen inquired about the potential risk to UCTT's China revenue from new AI-related rules and the sustainability of its current run rate. He also asked for an update on a previously mentioned European customer issue. In a follow-up, he pressed further on the risk of China revenue going to zero and questioned the state of customer inventory levels.

Answer

Chairman & Interim CEO Clarence Granger expressed confidence in the China business, noting it's a "China for China" strategy without direct tariff implications and that customer relationships are strong. He confirmed the European customer issue was resolved. Answering as Cheryl, VP of Marketing Cheryl Kneffler added that the products sold in China are broadly supported, mitigating risk. On inventory, she stated that while some remains, the situation has significantly improved, with the largest customer working down stock.

Hadi Orabi, on behalf of Krish Sankar from TD Cowen, asked for the potential worst-case margin impact from tariffs in September and whether new customer fab openings historically drive increased HBM probe card demand.

Answer

CFO Shai Shahar stated it's too early to speculate on future tariff impacts beyond the guided 1% headwind for Q2. CEO Mike Slessor explained that new fabs for additional capacity (not replacement) do drive incremental probe card demand, but typically with a 2-3 quarter lag before significant orders materialize.

On behalf of Krish Sankar, an analyst from TD Cowen asked about the long-term scalability of Rigetti's modular tile architecture, including potential physical limitations and whether the target for a 2x reduction in error rates could be outperformed. A follow-up question concerned the accounting treatment of the company's $100 million investment commitment under the Quanta collaboration, specifically whether it would impact OpEx or CapEx.

Answer

CEO Subodh Kulkarni detailed the company's roadmap, emphasizing that the chiplet (tiling) approach is the key strategy for scaling to thousands of qubits, leveraging proven semiconductor industry methods to overcome the limitations of monolithic chips. He expressed confidence in the scaling path and the error rate reduction target. CFO Jeffrey Bertelsen clarified that Rigetti's $100 million commitment is a continuation of its existing R&D efforts and will flow through operating expenses (OpEx), not representing a new or separate capital expenditure.

Speaking on behalf of Krish Sankar, an analyst asked about the next development milestones with Riverlane on quantum error correction (QEC), how the work with NVIDIA and Quantum Machines at IQCC relates to the QEC efforts, and the outlook for quarterly operating expenses.

Answer

CEO Subodh Kulkarni clarified that the NVIDIA/IQCC collaboration is focused on the fundamental QPU-GPU interface for hybrid computing, which highlights the speed advantage of superconducting qubits. The work with Riverlane is a distinct, parallel effort focused on the complex software layers of QEC. CFO Jeffrey Bertelsen added that he expects operating expenses to remain in line or only slightly higher than recent quarters, without a significant increase.

Krish Sankar of TD Cowen requested the specific definition of an "AI PC" and its impact on bill of materials (BOM) cost. He also asked about the time lag for seeing benefits from declining commodity costs, such as memory.

Answer

CEO Enrique Lores explained the 15% figure uses the IDC definition of a PC with an NPU. He emphasized HP's focus is on "next-gen AI PCs" with >40 TOPS NPUs. Regarding costs, he stated commodity cost pressure is expected through the first half of FY25, with mitigation from repricing and cost actions taking effect in the second half. CFO Karen Parkhill added that strategic buys helped initially, and now the focus is on repricing, which takes time.