Vivek Arya

Bill acknowledged optics will grow faster off a smaller base while AECs grow more slowly after doubling/tripling in prior years, but gave no explicit 1H/2H split.

Dan attributed the beat to broad strength, all top three customers grew sequentially, with the largest at 39% of revenue, rather than a one-off item.

Bill listed front-end network, scale-out back-end and switch-rack applications already in production, and said reliability keeps co-packaged optics from displacing AECs anytime soon.

Asked how large the AEC market can become and whether it cannibalizes copper or optical; Brennan said the TAM is expanding and, for the first time, copper is replacing some optical connections.

Pressed on whether Credo is investing enough to diversify beyond AECs; Fleming said more than half of R&D OpEx is tied to optical projects and forthcoming growth pillars.

Fleming disclosed exact concentration (61% / 12% / 11%) and reiterated three to four 10% customers plus two new hyperscalers in H2 FY26.

Fleming said margin expansion won't always be linear given product mix, and Q1 was guided at 65% at the midpoint without a new long-term model.

Dan gave the 86% concentration figure, framed Q3 as a bit of an outlier, and pointed to expected diversification across hyperscalers.

Bill said the inflection reflects long-planned, frozen architectures deploying, not a shift to leading-edge parts like Blackwell; Credo is GPU-agnostic.

Dan sized the three 10% customers (Microsoft ~11%, second AEC hyperscaler 33%, emerging hyperscaler 14%), noting the 33% customer fell sequentially.

Described relative intensity across HBM, conventional DRAM, logic and greenfield fabs but declined to give an explicit rank order or share quantification.

CEO explained the moat directly (high mix, algorithms, 1600-1700 applications engineers) but did not quantify any China share scenario.

Bren explained the difference is definitional: core WFE growing from $110B to low $120B, plus advanced packaging from $11B to $12B+, totaling mid-$130B which is consistent with the peer. Rick added packaging growth is constrained by customer facility readiness, not demand.

Bren guided China flattish to slightly positive in 2026, with KLA's China revenue in the mid-20% to high-20% range and total China WFE in the mid to high $30B range. On supply constraints, the company is virtually sold out in the first half.

Bren explained leading-edge is actually ticking up but being offset by China normalization from 39% in September, with DRAM also upticking. Export control impact on December quarter is fairly immaterial as slots were rearranged.

Bren guided to a Q4 gross margin uptick from mix, with tariff mitigation through free trade zones, supply chain optimization, and potential longer-term pricing adjustments.

Bren quantified total DRAM intensity increase at about 200 bps from pre-EUV to current HBM era, with an incremental 100 bps from HBM on top of the EUV step-up.

Rick noted customers are still building capacity below stated demand, with leading-edge logic, HBM, and advanced packaging all still in early stages, making 2026 outlook encouraging absent macro disruption.

Bren maintained full-year China at around 30%, with Q1 at 26% being lumpy. Export control impact still estimated at $500M ± $100M with 65-70% systems.

Wallace reframed it as a demand question driven by AI across training and inference, not a function of the number of leading-edge foundries.

Jensen argued the labs remain long-term NVIDIA customers; Colette added the commitments are supply commitments, but no specific FY2028 cash figure was given.

Jensen confirmed $20B is standalone CPU revenue, explained Vera has 4 use cases (Vera Rubin pair, standalone, storage with CX9, confidential computing), and expects supply constraints throughout Vera Rubin's life.

Jensen framed customer cash flow and CapEx growth as directly tied to token/inference demand, which he argued is at an inflection point and compounding.

Jensen said NVIDIA's share per gigawatt increases each generation (Hopper ~$20-25B, Grace Blackwell ~$30B, Rubin higher) and argued hyperscaler CapEx is cash-flow funded, not reliant on vendor financing.

Colette said H20 shipments depend on geopolitical resolution and additional licenses. Jensen argued ASICs rarely reach production and touted NVIDIA's full-stack co-design, architectural ubiquity across clouds, and perf-per-watt leadership.

Colette sized the H20 Q1 recognized revenue at $4.6B plus $2.5B unshipped, and pegged the inaccessible China TAM at ~$50B. Jensen framed AI as essential infrastructure across cloud, enterprise, industrial and sovereign, still at the very beginning.

Colette framed the GM trough as tied to Blackwell ramp execution and confirmed mid-seventies target later in the year. Jensen cited PO visibility, hyperscaler CapEx trajectory, and the long-term shift to accelerated computing as demand-sustainability evidence.

Jensen pushed back strongly, arguing no digestion phase until the trillion-dollar data center installed base is modernized for machine learning, plus a new AI factory industry is being created.

Colette directly confirmed mid-70s margins in H2 2025 is a reasonable goal, dependent on mix of ramp.

Haviv framed a long runway with revenue still below the 2022 peak, and said Q3 growth is mostly unit-driven with almost insignificant pricing contribution.

Haviv acknowledged 2025 had a strong start that didn't accelerate in second half. Cautious on macro/geopolitics but optimistic on secular growth. Playing it quarter-by-quarter.

Haviv pushed back on double-ordering concerns, noting industrial is still 15% below 2022 peak. Highlighted broad recovery across all sectors, regions, and customer sizes including tail customers.

Haviv said no direct impact seen from memory pricing but speculated customers may be replenishing inventory in response. Mike added that customers sometimes come in quickly to order parts when they complete their BOM, which could be related.

Haviv confirmed Q1 strength is not pricing-related; it's order-driven from industrial recovery and data center now moving the numbers. Full-year 2025 pricing was low single digit down as expected, and that's the 2026 assumption.

Management declined to comment on Q4, noting Q2/Q3 are typically stronger quarters. Mike confirmed Q4 and Q1 are seasonally lower but deferred any guidance.

Haviv admitted industrial ran hot at ~15% sequential growth with China up 32% YoY, warranting caution for Q3. Mike provided China details showing broad growth except auto.

Rafael explained GPM beat was from higher revenue and industrial mix, loadings were down but higher than base case, and guided Q2 gross margins up with loadings increasing slightly.

Haviv pointed to continued industrial strength as the main driver, with automotive also growing sequentially and YoY but still last to recover from the shallow downcycle.

Roche attributed strength to defense/AI super cycles, Maxim synergies, cyclical tailwind, and pricing, guiding to a brisk growth year in fiscal '27.

Roche explained pricing offsets inflation and is sticky due to long product life cycles with zero competitive substitution post design-in. Puccio added pricing came through as expected and adds a couple of points to FY26 growth.

Rich quantified AI business at $2B+ run rate: 40% ATE, 60% data center (split evenly between power and optical). Vince expects all areas to grow double digits over the next several years.

Rich detailed Q1 by segment: industrial up mid-singles, auto down mid-singles, comms up 10%, consumer down low double digits. Vince expects industrial and comms to lead FY26 with all markets up.

Rich confirmed expectation to grow at seasonal in Q1, noting Q1 typically declines low single digits. Vince expects industrial to be a very strong part of momentum in the coming year.

Rich estimated still shipping 10%+ below consumption, but expects to ship to end demand in Q3. Vince confirmed cyclical and idiosyncratic tailwinds driving recovery across the franchise.

Rich confirmed 7-10% long-term growth model with potential upside. Vince confirmed normalization across direct and distribution channels, all segments including industrial, and the broad market customer base.

Rich acknowledged Q4 industrial was below expectations due to broad market weakness and channel inventory reduction, but pointed to two quarters of sequential growth and 20% under-shipment as recovery support.

Rene pushed back directly, citing industry-wide tightness and noting peers with their own fabs have also been unable to supply demand.

Rene pushed back on share concerns, noting all 3 largest GPU providers now pair with Arm CPUs. Highlighted customers like Meta, SAP, and OpenAI who will use AGI CPU rather than design their own.

Jason confirmed data center is in the teens-to-20% range and growing much faster than all other segments. SoftBank contribution now $200M per quarter as a full-quarter run rate, expected to be durable. Data center projected to match or exceed mobile (40-45% of total) within 2-3 years.

Jason quantified SoftBank at ~$178M/quarter (license + design services), up $52M QoQ. Called it a good run rate going forward. Noted eventual product could cannibalize licensing but stack royalties on top.

Jason attributed the low-end royalty result to slightly weaker smartphone growth, though still an order of magnitude above market. Rest of year sequential forecast largely unchanged.

Jason pushed back that Q4 licensing was where expected internally. ACV at 15% is above LT target of mid-to-high single-digits; sustainable double-digit ACV possible but not guaranteed.

Jason framed the flat v9 percentage as positive, v8 grew stronger than expected while v9 absolute dollars still grew triple digits. Still expects v9 to reach 60-70% of royalties.

Rene detailed v9 value-based pricing and CSS royalty uplift and said mix will rise over multiple quarters, but did not directly bridge the royalty guide-down to high teens.

Murphy said the new program is part of the more-than-doubling custom setup for next year, judged conservatively, with the most meaningful part of the FY28 raise actually coming from scale-up optics and switching; declined to quantify the single program.

Murphy confirmed $10B+ FY29 custom target, noting it derives from a $55B TAM at 20% share; said new tier-1 XPU program is about a third of FY28 custom growth and on track.

Murphy sized XPU attach at couple hundred million in FY2026, doubling in FY2027, potentially over-doubling to ~$1B in FY2028. Total XPU attach TAM sized at $15B in calendar 2028 with 20% market share goal (~$3B). On new XPU customer exclusivity, emphasized deep multi-generational engagement with strong confidence in sustainability and revenue targets.

Murphy agreed optics is tied to AI acceleration growing faster than CapEx; used CapEx as broad proxy. FY2027 custom growth mostly from existing programs and lead customer transition, with second bigger XPU customer contributing more in FY2028.

Murphy declined annual guide but highlighted strong tailwinds: custom up H2 vs H1, optics up double digits in Q3, enterprise/carrier recovery from $900M to $1.7B run rate. Positive setup for FY2027.

Murphy firmly addressed competitive noise, stating Asia supply chain sources have 'incomplete view.' Confirmed multi-year revenue growth commitment regardless of customer's other relationships. Acknowledged customers may pursue multiple paths.

Murphy broke down data center mix: ~50% Electro-Optics, ~25% custom, balance other. AI crossed majority of data center. Provided sequential growth context (25% Q2-Q3, 25% Q3-Q4).

Matt confirmed tracking significantly ahead of $1.5B AI target by hundreds of millions, both this year and next year. Attributed outperformance to demand, not supply, with strong visibility across custom AI, optics, and switching.

Gregory confirmed an ARM mix shift in server CPUs positions Teradyne for greater share gain while the company also works to gain share in x86.

Gregory said the correlation holds over three- to five-year periods with roughly a year lag from fab equipment spend to test revenue, and is not a good year-by-year predictor.

Greg explained tester lead times match wafer production lead times, not front-end equipment. Customers hold back test equipment orders until wafers are flowing, creating structural visibility gap.

Greg explained the gating factor is fast-follower execution speed and eventual tester-agnostic development. Only low double-digit GPU share needed for the $6B target model.

Confirmed Q4 2025 will be stronger than Q4 2024 at the company level, not just compute. Significant compute growth expected Q3 over Q2 and Q4 over Q3.

AI compute roughly 20% of SOC in Q2. GPU opportunity would have modest positive impact in 2026 if successful, building over time.

Second-half strength driven by share growth plus back-half-weighted recovery in Auto, Industrial and Mobile segments rather than Compute alone.

Equipment demand doesn't track end demand linearly, experience curve improvements reduce test times on maturing devices, limiting sequential capacity adds.

Amie guided margin one quarter at a time and flagged memory content dilution; Hock argued the frontier labs will become first-party hyperscalers running their own data centers, making Broadcom's investment economically compelling.

Hock argued the 6 LLM customers are strategic multi-year partnerships, not transactional, visibility comes from roadmap commitments.

Hock confirmed $73B is current backlog and pushed back strongly on COT threat, citing the hardware performance advantage of purpose-built XPUs over general GPUs.

Hock said AI growth rate accelerates in fiscal 2026, declined to quantify. Networking share of AI mix expected to decline as XPU volumes rise faster.

Hock dismissed the XPU softness read-through, said XPU is tracking expectations. Declined to update FY27 numbers on prospects until FY26.

Hock clarified $60-90B SAM already includes both training and inference XPUs with different architectures; training is larger portion of dollars within same SAM.

Hock estimated the 2024 AI SAM at $15-20B going to $60-90B by 2027. On margins, he acknowledged gross margin dilution from AI mix but emphasized operating margin improvement through revenue leverage.

Wall said Hexagon is delivering as expected and warned against annualizing IP; Devgan confirmed most of the IP growth is organic.

Devgan cited marked improvement over last 3-6 months. Customers may use multiple foundries for capacity, creating more design activity. Agentic solutions welcomed for productivity gains.

Anirudh explained 3-layer cake framework, AI tools call Cadence base software to get jobs done. Agentic flows use more software, not less. Workload is exponential (chips going from 100M to 1T transistors). No customer discussions about reducing usage.

Anirudh cited three reasons for IP strength: focus on AI/HPC at advanced nodes, four major foundries at leading nodes (TSMC, Samsung, Intel, Rapidus), and improving PPA driving customer demand. Expects IP to grow better than Cadence average going forward.

John Wall upgraded China from flat to slightly up for FY. Anirudh said rest of world doing even better than China, so China percentage may decline over time.

Anirudh acknowledged Intel as historical weak spot. Pleased with Lip-Bu appointment. Discussions already started on expanding relationship.

John Wall attributed recurring revenue deceleration to China headwind and mix shift toward upfront revenue businesses (hardware, IP) growing faster than average.

Tim said longer-term customer conversations improve visibility and enable disciplined supply chain buildout; Doug added down payments are unnecessary given ample free cash flow.

Tim emphasized margin improvements are rooted in fundamental value delivery and operational efficiency rather than transactional pricing, but declined a long-term margin forecast.

Tim confirmed plans to increase WFE share again in 2026, driven by accelerating technology transitions and higher deposition and etch intensity. China WFE expected flattish year over year.

Tim framed Lam's AI data center opportunity through SAM expansion from low 30s to high 30s of WFE at leading-edge nodes, without quantifying semiconductor content share.

Doug attributed the slight first-half weighting to customer plan timing and lumpiness rather than a single customer, plus the $700M second-half-weighted lost China business.

Doug cited Reliant declining from a strong 2024 as the drag, offset by upgrade spending, and said OpEx growth will be slower than revenue growth.

Management said China is trending lower as a percent of revenue as other markets strengthen, but is not going away.

Sanjay described large agreements as multiyear with floor and ceiling price bands at CQ2 levels but declined to break out data-center specifics.

Sanjay reiterated that Micron reached its HBM-DRAM share parity target in 2025 and will manage HBM vs non-HBM mix as part of the overall portfolio, supplying both HBM3E and HBM4 in calendar 2026 without breaking out quarterly share.

Mark pushed back on reversion to historical margin means, citing AI as a transformational secular driver, memory's increased strategic value, supply constraints (HBM trade ratio, bits-per-wafer decline, greenfield lead times), and customer willingness to enter SCAs.

Sanjay acknowledged possible unit-demand impact and customer mix adjustments in smartphones and PCs given memory prices, said these are already accounted for in Micron's forecast, and emphasized AI content-per-device continues to drive memory demand across edge devices.

Mark declined to quantify full-year margins but committed to Q1→Q2 sequential gross margin improvement driven by tight DRAM supply, improving NAND, favorable mix, and good cost performance.

Sanjay said HBM4 first production shipments start in calendar Q2 2026 and ramp through the year in line with customer demand, while declining to comment on HBM3E pricing direction.

Mark attributed the Q3 beat to stronger-than-expected pricing and good cost performance, and said Q3→Q4 favorable mix (more DRAM, more data center) is driving the 42% guide.

Jim reaffirmed second-half calendar 27 for scale-up CPO revenue, noting nearly every strategic customer is engaged on CPO or NPO depending on architecture.

Sherri noted FX drove Q4 margin decline, not expected to recur. Apple partnership expected to benefit gross margin. Very confident in achieving long-term 42%+ target.

Jim acknowledged quarterly fluctuation as normal, pointed to full-year 60%+ datacenter growth with share gains. Strong demand signals across 800G, 1.6T, OCS, and DCI.

CEO cautioned against quarter-by-quarter IP analysis due to lumpiness. IP headwinds derisked. CFO added IP will be back-half loaded due to title availability timing.

CEO expects challenging China environment to persist in FY26. Not assuming positive changes. Comps get easier beyond FY26 if no additional restrictions. Long-term double-digit EDA commitment maintained.

CEO provided qualitative color on hyperscaler silicon investment acceleration over 3-5 years but avoided specific design start growth figures.

Michael noted only $75M incremental needed from $525M midpoint. Components, OCS will contribute alongside modules. All 3 customers contributing revenue. No new customers expected near-term.

Wajid expects higher end of 37-40% GM range at $600M. Noted 150bps quarterly NPI headwind from customer ramps not yet generating revenue. As volumes grow, NPI as percentage of revenue decreases.

Cristiano confirmed both merchant and custom silicon approaches, with bespoke IP configurations for different hyperscaler solutions. Akash added the custom engagement is expected to be accretive at operating margin level.

Akash confirmed Apple royalty expected to remain at current scale pending renegotiation, operating as independent business from chip relationship.

CFO said Scorpio (mostly P-Series) broke above 15% of 2025 revenue and X-Series ramps materially in the back half of 2026.