LRCXLam Research Corporation

“we said in early 2025 that roughly $40 billion in conversion spending would be required over several years to enable existing NAND installed wafer capacity to produce devices with more than 200 layers. We now anticipate that this conversion will be pulled forward with the majority of spending occurring before the end of calendar year 2027.”

“We now expect WFE of $140 billion with a bias to the upside as the industry continues to work through various constraints. We believe this sets the stage for another year of compelling WFE growth in 2027.”

“Lam's advanced packaging revenue growth is expected to exceed 50% in calendar year 2026.”

“clearly, there is increased demand for NAND coming from AI data centers, and that's helpful. But also, if you think of the -- on a relative basis, what under investment in that area, partly as customers make choices about clean room allocation and obviously some other devices like HBM were so hot during that period. Also, going back to what we said early last year, the installed base had gotten a little bit behind in terms of the state-of-the-art technology. And so most of the installed base at that time, early 2025, about 2/3 of it was still running in the [ 1xx ] 100-plus layer technologies really when you need to get those incremental bits out now, you need to be 200-layer plus.”

“We are also gaining momentum with our Dextro cobots, which deliver an unprecedented level of automated tool maintenance precision and repeatability. Customers using Dextro and production are benefiting from higher output and in some cases, improved yield from existing capacity. In the March quarter, we expanded Dextro coverage to 8 Lam tool types up from 6 last quarter. We also introduced the next generation of Dextro, which packs 10x more compute power than the first generation into a smaller footprint.”

“In NAND, AI transformation is moving beyond compute and into the storage layer. [ Hoken ] economics are driving changes to the memory hierarchy used in AI data centers including rising adoption of higher layer count QLC-based NAND devices for SSDs. We expect total data center bits this year to be greater than both PC and mobile segments combined with growing -- continuing growth in data center mix into the future.”

“As the industry moves to 1C nodes, we see our total dielectric deposition SAM and DRAM growing more than 20%. With innovations like Stryker ALD, we believe Lam is well positioned to gain share within this expanding opportunity.”

“In 2026, we see Lam's served available market or SAM, percent of WFE, expanding to slightly more than the mid-30s percent level, well on track towards our stated goal of high 30s percent over the next few years.”

“overall industry installed wafer capacity is expected to decline more than 20% from prior highs by the end of this year.”

“Equipment Intelligence, if you think about what it does is it allows us to look at massive amounts of data coming from our tools on every single wafer, and that shortens troubleshooting time if there is a problem with the tool. It helps us with the time to ramp those tools either on new process or as they start up, helps us to match tools, better tool to tool chamber to chamber, all those things can yield -- lead to those tiny little improvements in yield that really do matter for the customer.”

“In DRAM, AI's power and efficiency requirements are driving an industry transition to 1C generation devices. As feature dimensions shrink, the industry is shifting from traditional silicon nitride base dielectric films deposited using furnace, the more advanced ALD silicon carbide [ loc ] layers to achieve bit line capacity in production.”

“Highlights included a new agreement with a leading foundry/logic customer to deploy our equipment intelligence services for critical deposition applications. The top memory customers also set to utilize our Equipment Intelligence capabilities in R&D to enable faster ramps of new nodes for NAND and DRAM production.”

“For Lam, the AI-driven demand environment is creating an ideal setup for continued outperformance. Semiconductor technology inflections required to meet escalating AI compute needs are driving higher deposition and etch intensity.”

“memory is just so critical in all of these accelerated compute architectures to feed the parallel compute, you need just data coming in to keep the machine going. And so the criticality of it maybe is more than it's ever been from my point of view.”

“these are being driven by the growing importance of NAND as we see it within the AI memory hierarchy. And so again, we think it's something that in a matter of time, this kind of capability is likely needed”

“Our guidance for the June quarter points to Lam's strong momentum in an accelerating AI-driven semiconductor demand environment.”

“it does seem like AI is driving memory demand growth a lot faster than what we've seen historically.”

“Is there a way to segment was the bigger driver NAND? Was it CPU tightness? Or was it just AI strength?”