Karl Ackerman

Irving said nearline and non-nearline gross margins are both in the mid-50s with little differentiation now.

Kris clarified purchase orders placed ~1 year in advance, with LTA frameworks governing periods beyond the first year with more variability.

Irving stated focus is on areal density and platter count (up to 14 disks) rather than unit capacity. Will invest in heads/media capacity only if economically justified.

Irving confirmed HAMR qualification started with one hyperscaler this month, second customer qualification imminent. Higher capacity drives key to meeting strong exabyte demand.

Irving confirmed inference driving significant data generation. Hyperscalers manage economics across SSD/HDD/tape tiers, and inference data volumes favor HDD economics. Conversations with customers indicate inference will drive significant HDD storage requirements.

Irving clarified price increase letter targets channel customers (client/consumer/low-end nearline), roughly 10-15% of business. Hyperscale LTA/PO terms unaffected.

Kris confirmed 7.5M SanDisk shares remain, to be monetized before February 21 anniversary. May use debt-for-equity exchange again.

Irving confirmed third hyperscaler qualified on UltraSMR, fourth in process. Technology extensible to HAMR. Legacy workloads use lower capacity, but new workloads all on highest capacity.

Irving explained 12-month lead times driven by 9-month head wafer production. LTA visibility extending from 12 to 18 months. Investment focused on head/media for areal density.

Irving sees opportunities in sovereign and private clouds. No pull-forward behavior observed; linearity consistent with historical patterns.

Irving explained exabyte growth driven by areal density and Ultra SMR technology without CapEx. New 28TB and 36TB platforms coming in next few months.

Irving confirmed qualifications completing in CQ1, with customers excited about the new drives. Nearline outlook remains healthy.

Asked about HDD capacity expansion room; management declined to quantify capacity but described sizing to demand, partial.

Robbins credited the broader platform approach rather than Cisco Cloud Control alone for driving the campus demand.

Robbins said demand exceeds supply for 51.2T products. Huge acceleration in 800G optics and Acacia portfolio. Additional design wins in Q2.

Robbins detailed 1,000 new customers for new security products (up 100% QoQ), 4,000 total since launch. Third consecutive quarter of double-digit firewall unit growth. Organic security exiting Q4 near double-digit revenue growth.

Robbins said guide raise reflects both hyperscale AI and campus refresh strength, didn't quantify exact split.

Patterson showed inventory plus advanced purchase commitments up $1B in 90 days and 38% ($3B+) YoY, demonstrating capacity investment to meet accelerating hyperscale demand.

Robbins confirmed intent to drive Silicon One adoption as fast as possible across 800G Nexus and smart switches, said half of switch ASICs in three years is approximately right.

Robbins ranked: 1) Service provider/cloud, 2) Enterprise, 3) Public sector. Ex-federal, rest of world grew 10%. Federal expected to return to growth in FY26 but below FY25 levels.

Robbins said enterprise AI orders are in the hundreds of millions with acceleration but noted NVIDIA partnership solutions start rolling out in ~60 days, so upside is still ahead.

Robbins said CPO was demonstrated in 2023, team actively working on it. Adoption driven by customer demand for power benefits and speed requirements.

Robbins and Herren confirmed demand is from normalized ordering patterns, not from extended lead times like the supply shortage era.

Scott said restructuring was about reallocation not cost cuts. Investment focused on AI and security. YoY OpEx growth mostly driven by Splunk addition.

Chuck emphasized the distributed nature of AI infrastructure demand across multiple hyperscalers, not concentrated in one customer.

Jim attributed the $3 billion target primarily to data center and communications, spanning 800G and 1.6T transceivers, OCS, CPO, multi-rail, and LTA-linked pricing.

Jim sees R&D scaling slightly below revenue to maximize growth opportunities. Sherri confirmed SG&A target of 8% of revenue (Investor Day), R&D target of 10%. ERP consolidation and low-cost region G&A initiatives showing benefits.

Jim confirmed strong semi-cap orders pickup in December quarter expected to drive sequential industrial revenue growth from June quarter. Too early to call broad industrial recovery beyond semi-cap.

Jim confirmed record bookings for telecom and DCI components as well as transceivers. Customers booking over a year in advance for supply coverage, with large customers providing forecast visibility through calendar 2028.

Jim corrected that InP serves multiple products beyond EMLs (CW lasers, photodiodes). Will maintain internal/external mix for supply chain resilience.

Jim confirmed targeting Investor Day growth range or better, actively adding capacity to support strong demand signals from datacenter customers.

Sherri noted no significant direct P&L impact from restructuring in Q3; benefits more long-term, though R&D shift to growth engines already visible.

Jim confirmed strong short-term and long-term demand signals from datacenter customers, believes company gained share over last 2-3 quarters and focused on continuing gains.

Jim confirmed InP capacity tripled YoY with CHIPS Act support for Sherman, TX expansion. Uses multisource strategy with external EML partners alongside internal production.

Sherri explained OpEx increase driven by higher R&D spend for long-term growth, while SG&A was reduced sequentially. Deferred long-term model to Investor Day.

Jim confirmed supply chain resilience increasingly important to datacenter customers, highlighted geographic diversity and vertical integration as key differentiators.

Bill explained hyperscalers point module customers to Credo components and that component and module sales will remain complementary long term.

CEO said training, inference and agentic all deploy differently but each is a strong connectivity opportunity; agentic drives more front-end connections.

Dan attributed the guide-down to conservative forecasting and quarterly mix, noting Credo is running at or above the high end of its 63-65% long-term range.

Bill said Credo filed an ITC case to protect AEC IP built over 7-8 years, has reached licensing deals with 3 parties and more in flight, with no competitive change.

Dan attributed the raised outlook to broad, industry-wide forecast strengthening over recent quarters rather than a single customer or program.

Asked whether AECs have a larger advantage in scale-up Ethernet vs PCIe; Brennan said the near-term scale-up opportunity is with PCIe and that Credo is protocol-agnostic on the 200 gig per lane SerDes.

Asked whether new hyperscalers are adopting 100 gig per lane AECs across multiple programs; Brennan confirmed recent ramps are at 100 gig per lane and noted a pull-in versus the prior schedule.

Fleming said CapEx is driven by 3nm production mask-set tape-outs, disconnected from customer contracts rather than take-or-pay commitments.

Dan attributed the margin uptick principally to scale, with product gross margin ex-services at 62.4%, plus the Amazon warrant contra-revenue rolling off.

Dan said the additional four 5%+ customers are broad-based across products, while the 10%+ customers are largely concentrated in AEC.

Bill reiterated Credo is fundamentally a product company; IP will shrink as a percentage and remains lumpy, and he expects it below 10% this quarter.

Jayshree said NVLink dominates scale-up today but Arista is more engaged in non-NVIDIA accelerators (AMD MI, Google TPUs) for scale-up and scale-out.

Jayshree firmly stated ANET remains product-led and does not expect services as a percentage of revenue to increase despite deeper design engagement with customers.

Jayshree confirmed $3B+ AI target includes front-end demand tied to back-end clusters, with agentic AI applications expected to drive adoption across enterprise and XPU diversity.

Ken Duda highlighted 15 years of both architectures in production, offering choice between congestion-free scheduled fabric and lower-cost unscheduled, both running same EOS.

Jayshree confirmed all four titans expected in back-end by 2026, with Neoclouds also adopting combined back-end and front-end deployments.

Jayshree confirmed 800G activity picking up in 2025 following Etherlink portfolio introduction in late 2024, with some reflected in shipments and some in deferred.

Chantelle declined to guide services separately, noting timing effects from catching up to product post-COVID and suggesting historical trends as a guide.

Michael confirmed line of sight to $250M quarterly transceiver run rate. Wants to grow in margin-rich profitable fashion rather than unbounded. Added manufacturing capacity in Thailand to support bounded growth. Other growth drivers (CPO, OCS) provide additional upside beyond transceivers.

Michael dismissed pull-forward concerns, demand has been a constant race to keep up. Wupen confirmed ramps tied to dominant GPU and accelerator introductions but demand is broad-based.

Wajid clarified DCI tunable products moving in-house. CapEx going in this quarter will add 50% more DCI capacity by end of CY2025. Contract manufacturers kept at full speed with incremental supply from Nava.

Wajid confirmed roughly matched CapEx in second half, primarily for EML capacity expansion in Japan. Thailand infrastructure mostly in place, needing some equipment orders.

Alan confirmed new customer wins are sub-1.6T. 1.6T with multiple customers but CY2026 timeframe. CW laser shortage impacting own transceiver production this quarter.

Chris clarified wins are new opportunities, not reallocations, but U.S. headquarters and non-China manufacturing makes them more competitive. Scale-up optics still 1-1.5 years out.

Chris declined to discuss customer architectures. Noted initial transceivers won't have Lumentum EMLs but future SKUs will increase vertical integration as they earn more share.