99% of You Will Miss the Memory Stack (My Full Map)

Reasoning: Supplies the widest slice of equipment for HBM production, including deposition, etching, and layering tools. It runs a joint research center with SK Hynix, ensuring its machines are designed into next-generation memory. HBM revenue grew from almost nothing to $1.5 billion in one year and is targeting $3 billion next. Brian sees it as a stable, cash-rich giant with a tripling memory 'afterburner.' He allocates $25 out of $100 to the 'tools' layer, including Applied Materials and Lam Research.

Reasoning: This ETF focuses exclusively on memory makers such as Micron, SK Hynix, and Samsung, without including any suppliers from the underlying memory stack. Brian describes it as covering a 'completely different part of the same machine' compared to HBMX, noting that it only buys the makers. He implies it does not align with his 'pure picks and shovels' investment strategy.

Reasoning: Launched the Magnum 7H specifically for high-bandwidth memory testing, as stacked memory chips require up to 10 times more testing than ordinary ones. While total company revenue slipped 14% over four years, its DRAM test revenue exploded from $80 million to $350 million (over 330% growth). AI-related work has climbed past 60% of its business and is heading towards 70%. Brian highlights that the part of Teradyne that is winning is precisely the part tied to the exploding memory market. He allocates $22 out of $100 to the 'testers' layer, including Teradyne and Advantest.

Reasoning: Controls 97% of a market most investors haven't heard of. Manufactures the registering clock driver, a chip essential for DDR5 modules and keeping data clean at AI speeds. It also owns underlying memory patents, earning royalties from manufacturers, effectively getting paid twice per chip. Over 4 years, revenue doubled, operating income exploded almost 800%, and operating margin reached 37%. Brian notes an open antitrust inquiry. He allocates $8 out of $100 due to its narrow focus and legal question, calling it the 'smallest bet' despite a 'beautiful' royalty model.

Reasoning: Considered a purer memory bet. Its tools are essential for carving deep vertical channels and filling them with copper, which is how memory is stacked into tall HBM chips. After a downturn, revenue snapped back to an all-time record, climbing 24% in a year, with net income jumping 40%. Brian states that when memory moves, Lam 'turns harder than almost anyone' due to its exposure to memory. He allocates $25 out of $100 to the 'tools' layer, including Applied Materials and Lam Research.

Reasoning: Specializes in high-speed imaging to detect hidden defects in hybrid-stacked memory, which have up to 1,000 times more connections than older methods. Its new Dragonfly G5 system was selected by a leading memory maker for inspecting next-gen HBM4. A single customer locked in a purchase agreement worth over $240 million through 2027, representing nearly a quarter of Onto's annual revenue and providing a guaranteed forward business. He allocates $25 out of $100 to the 'inspectors' layer, including Onto Innovation and KLA.

Reasoning: Holds a near-monopoly in finding nanoscale defects across the industry, with systems like its broadband plasma line unmatched by rivals. As chips become more advanced, the number of places for defects to hide increases, making KLA's role even more critical. Over four years, revenue grew 76%, operating income climbed 110%, and operating margin increased from 36% to 43%. Brian emphasizes that there is no real number two, making KLA the primary beneficiary when the industry needs more inspection. He allocates $25 out of $100 to the 'inspectors' layer, including Onto Innovation and KLA.

Reasoning: Holds approximately two-thirds of the entire chip testing market, making it effectively a 'toll booth' for the testing layer. Its family of testers checks both AI processors and the high-bandwidth memory stacked around them (e.g., T5835 memory tester). In its most recent fiscal year, sales grew about 45%, and operating profit jumped almost 120%. Brian notes that orders keep compounding once a company becomes the industry standard. He also mentions that it trades over-the-counter in the US, making it a bit 'thinner to buy.' He allocates $22 out of $100 to the 'testers' layer, including Teradyne and Advantest.

Reasoning: A leader in hybrid bonding, a technology essential for HBM4 that welds memory dies directly copper-to-copper, rather than using solder bumps. Besi builds these bonding lines jointly with Applied Materials and achieves high accuracy in chip placement. Last year, its revenue grew 28%, but new orders dramatically more than doubled (up 104%). Brian emphasizes that in this business, orders typically precede revenue by a year or two, and memory makers are placing hybrid bonding orders now for 2027 production, indicating strong future demand. He notes it is an over-the-counter name in the US. He allocates $20 out of $100 to the 'foundation' layer, including Besi, Kulicke & Soffa, and Integrys.

Reasoning: Pioneered a cleaner method to weld stacked chips using heat and pressure instead of solder, pushing towards advanced stacking needed for HBM with its Apptura platform. Brian is transparent that its numbers require honesty: revenue has been cut by more than half (down 57%) over four years, and the company is currently around break-even. He states that the entire case is forward-looking, with its advanced packaging line just starting to ramp and real HBM volume expected as a 2027 story, with its first system still in qualification. Brian identifies it as a 'speculative patient higher risk pick' based on where memory packaging is heading. He allocates $20 out of $100 to the 'foundation' layer, including Besi, Kulicke & Soffa, and Integrys.

Reasoning: Supplies ultra-pure chemicals, filters, and polishing materials essential for chip fabrication. Once qualified into a production line, customers tend to stay for years due to the slow and risky process of switching suppliers. Brian highlights that advanced AI and HBM chips, while only 5% of all wafers made, drive nearly 30% of Entegris's revenue because they consume significantly more material, with this intensity roughly doubling with every new node. Revenue is up 39% over four years, with operating margin pushing to 29%. Brian concludes that Entegris wins on complexity, which memory chips are guaranteed to keep adding. He allocates $20 out of $100 to the 'foundation' layer, including Besi, Kulicke & Soffa, and Integrys.

Reasoning: A newer fund that provides a blend of investments, owning both memory makers and reaching down into the supplier stack, including equipment, packaging, and materials. Brian states that it 'buys you the blend of the whole stack,' making it an option for viewers who prefer an ETF that includes some of the underlying 'picks and shovels' components he discussed, rather than solely focusing on memory makers like the DRAM ETF.