Bernstein Conference: Lam Research's $40 Billion NAND Upgrade Pulled Forward, Panel Packaging Center Opens in Austria
Bernstein 42nd Annual Strategic Decisions Conference, May 27, 2026 — CEO Tim Archer lays out accelerating demand across every device category and flags a compressed upgrade timeline that changes the WFE growth calculus
The NAND Upgrade Clock Is Running Faster Than Anyone Expected
The single most important new data point from Lam Research's appearance at the Bernstein Strategic Decisions Conference is that the company's $40 billion NAND upgrade cycle — the process of stepping installed tools from roughly 100-layer to 200-plus-layer device capability — will now be substantially complete by the end of 2027. That is a dramatic compression from the "several years" guidance offered at Lam's February 2025 Investor Day, and CEO Tim Archer confirmed it explicitly. "On our most recent call," Archer said, "we said that entire $40 billion upgrade spending would likely be complete by the end of 2027."
The pull-forward is not a demand concern — it is the opposite. It reflects NAND's rising strategic importance inside the AI infrastructure stack, as inference, agentic AI, and KV-cache applications are driving enterprise SSD and high-density storage demand at a pace that was not anticipated even a year ago. The constraint, as Archer acknowledged, has partly been the shortage of clean room space, which forced NAND customers to max out existing fabs through upgrades before greenlighting greenfield construction. The greenfield wave, in his view, is now a 2028-and-beyond event — meaning the upgrade revenue lands first, and a second, structurally larger wave follows.
Critically, Lam captures a disproportionate share of upgrade spending because it dominates the critical etch and deposition steps required to build taller memory stacks. Archer was direct: "If the customer wants to build a taller device, typically the majority of the spending is actually on Lam equipment." And the math compounds — a larger 200-layer installed base simply creates a bigger surface area for the next upgrade cycle to 300 or 400 layers. "That's not a one and done," Archer noted, pointing to recent customer roadmap disclosures targeting 600-, 700-, 800-, and even 900-layer NAND architectures.
Panel-Level Advanced Packaging Is No Longer Theoretical — Austria Center Just Opened
Equally notable is Archer's confirmation that Lam opened a center of excellence for panel-level advanced packaging in Austria just days before this conference. This had not been widely flagged, and it represents a meaningful strategic bet. Lam acquired an Austria-based company with large-panel packaging positions several years ago and has been quietly investing in it. Archer stated the company is "making shipments now" in panel-level packaging and described the transition from wafer-based to panel-based formats as "almost a necessity" as AI device packages continue to grow in physical size.
When pressed on competitive overlap with dedicated packaging equipment players, Archer declined to name names but was clear that Lam intends to lead the transition rather than follow it. The advanced packaging business overall is running at approximately $2 billion in revenue and growing over 50% year-over-year in 2026, by Lam's own sizing — a business that barely registered five years ago.
WFE Is Running at $140 Billion With a Bias to the Upside, Constrained by Clean Rooms
Archer confirmed WFE is tracking at approximately $140 billion for 2026 with an upside bias, and framed the primary constraint not as equipment supply or demand, but as clean room availability. "Clean rooms are constraining the ability to ship more," he said, adding that once fab construction begins, Lam gets roughly two years of demand visibility from that project alone. He declined to give a specific count of fab projects under active tracking, but described the number as "a lot," encompassing both new greenfield sites and continuous technology upgrades at existing facilities. On 2027, he committed to "compelling WFE growth" even under current clean room constraints, and acknowledged that unconstrained demand would be higher still.
The broader setup Archer described is one of sequential demand waves — training drove leading-edge logic and HBM adoption; inference and agentic AI are now pulling forward NAND; and physical AI represents the next leg that Lam is already positioning for. "Every time we come to that next checkpoint, there's more than we thought there was," he said, describing what the company internally calls an "accelerating demand environment."
Dry Resist Is in Production at Two Memory Customers
Lam's Ether dry resist platform — a suite of deposition and etch tools that replaces wet chemical photoresist for the most critical EUV lithography steps — is now running as process of record at two memory manufacturers. This is a meaningful commercial milestone for a technology that was introduced roughly six to seven years ago and spent considerable time in customer qualification. Archer described the suite as consisting of an EUV photoabsorption underlayer, the dry resist material itself, and a dry develop process, all of which move lithography patterning into a fully dry, more controllable environment analogous to standard deposition and etch workflows.
The five-year revenue target of approximately $1.5 billion — back-end weighted — remains unchanged. Foundry logic adoption is described as moving more slowly given the longer qualification cycles at leading-edge nodes, but Archer indicated that as patterning requirements tighten below 2 nanometers, the case for dry resist in logic strengthens. The memory ramp, combined with accelerating EUV adoption in DRAM, provides the near-term revenue base.
Cobots and Equipment Intelligence Are Generating Real Revenue in Customer Fabs
One of the more underappreciated parts of Lam's services narrative is now a commercial reality. The company's Dextro cobot — a robotic system physically servicing Lam tools in customer fabs — is running in production across multiple customer sites. Archer described measurable improvements in first-time-right maintenance outcomes and in on-wafer metrics like edge uniformity after cobot-driven focus ring installation, where tolerances of 50 microns make human repeatability a genuine constraint.
The cobot addresses two converging problems: the scarcity of trained technical resources as fabs multiply globally, and the rising precision demands of AI-driven semiconductor manufacturing. Archer noted that a single cobot can service a large number of tool modules given the infrequency of maintenance cycles, making the economics workable. The business model is service-agreement based, creating recurring revenue and, as Archer put it, a very sticky customer relationship — "you're relying on that piece of equipment to effectively maintain your other equipment."
Alongside cobots, Lam is commercializing what it calls Equipment Intelligence — AI models built on tool telemetry data, including video analysis of wafer and plasma conditions, used to match tool performance across large fleets distributed across geographies. The ability to run the same process at a fab in Asia and one in the U.S. with consistent output is becoming a customer requirement, not a nice-to-have, as manufacturing footprints globalize.
The Business Has Flipped: 60% Foundry Logic, Margins Now in the 50s
Lam's structural transformation over the past five years is now fully reflected in the numbers. Five years ago, roughly 60% of revenue came from memory. Last year, 60% came from foundry logic. Archer described this as a deliberate seven-to-eight-year strategy, executed through targeted R&D redirection toward gate-all-around, advanced packaging, and selective etch and atomic layer deposition applications that foundry logic transitions increasingly require. "We have the broadest, most competitive product portfolio in the company's history," he said.
Gross margins have moved from the high 40s into the low 50s, driven by a combination of manufacturing scale at the Malaysia campus (a second facility of equal size opening in the second half of 2026), mix shift toward technology-enabling tools with higher value capture, and new products like the Acara conductor etch tool with direct-drive RF capability. Archer noted that excluding China — which has come down substantially from a peak of roughly 40% of revenue and is now running roughly flat year-over-year — core business margins have performed even better than the headline numbers suggest. Tariff impacts, he said, are already contemplated in reported figures.
China Is a Managed Plateau, Not a Recurring Controversy
China revenue has stabilized at a significantly lower share of the mix than its peak, with Archer describing the region as "flattish to maybe roughly up a little" in 2026. Lam competes only where regulations permit, which in practice means lagging-edge nodes where local equipment suppliers have absorbed the restricted business. Archer was careful not to dismiss local Chinese competitors — "they are real companies, they're very capable" — but the investment controversy around China sustainability has effectively resolved itself. The company's story has moved on.