Memory remains the video’s core trade: DRAM, NAND, HBM, hard drives, and the equipment/test stack are in a supply squeeze that has already driven outsized gains and still looks under-extended on valuation for the top makers. MU, SK Hynix, SNDK, WDC, STX, LRCX, AMAT, KLAC, and the niche suppliers are framed as the main ways to play a shortage that could run into 2028 and even beyond 2030, while the small caps are presented as higher-beta but far riskier. The other major setup is BSEM, a freshly uplisted Nasdaq microcap rebuilt in 1H26 around hospital allografts, with the next 12 months hinging on hospital expansion, trial data, 510(k) launch, and 2027 in-house manufacturing. The throughline: prefer the strongest names in the memory stack for long-term exposure, and treat BSEM as a speculative due-diligence watchlist name with material dilution and execution risk.
$100 in MU 1 year ago would be $818 today; DRAM contract prices rose in the low 60% range in a single quarter and NAND in the mid-80s, showing how extreme the memory squeeze has been.
HBM, the memory inside every AI chip, is sold out through 2027 at all three makers; hard drives are sold out through 2027; fabs are spending about $150 billion this year on capacity that lands late 2027 into 2028.
MU makes DRAM and NAND and is one of only three HBM makers; HBM is sold out through end-2027, MU last quarter revenue was $41.5 billion, up 346% YoY, gross margin was 85%, and the next quarter guide was $50 on the September 30 report.
MU trades at about 7x forward earnings, was a ZipTrader 25 pick, and was called a top pick in January at $363; the current stance stays constructive on it.
SK Hynix makes DRAM, NAND, and HBM, was first to get HBM approved inside NVDA products, and holds about 50%-55% HBM share.
SNDK, WDC’s NAND spinout, did $20.2 billion of revenue last fiscal year, up 175%, with data-center revenue doubling in one quarter, EPS of $39 last quarter, 4 years of contract visibility, and an August investor day guide for double-digit growth through 2030 and an 80% gross margin.
STX and WDC are the two remaining hard-drive makers at scale; hyperscalers are buying every high-capacity drive they can build, WDC says all 2026 production is sold out on multi-year purchase agreements, and WDC trades around 19x earnings.
LRCX gets 46% of systems revenue from memory, split evenly between DRAM and NAND; fiscal revenue was $23.2 billion, up 26%, EPS rose 39%, the September quarter guide was $8.1 billion, and the spending forecast was raised by $10 billion.
AMAT’s HBM-specific revenue rose from almost nothing to about $1.5 billion in a year, and management is targeting $3 billion of HBM-specific revenue.
KLAC remains the inspection monopoly for defects; stacked memory can require up to 10x the testing of a normal chip because one bad layer in a 12-high stack kills the part.
FormFactor’s probe cards test every layer before shipment, revenue was up 18% and net income up 163%, and margins should expand as HBM becomes a bigger share of its mix.
TER’s DRAM test revenue rose from $80 million to $350 million, and it launched an HBM-specific tester; BESI’s orders were up 104%, revenue up 28%, and HBM 4 increasingly needs its bonding niche.
ALAB revenue is up 98% and net income up 269%; its Leo CXL memory-expansion product is slated to ship in volume to two US hyperscalers in 2027.
SIMO is described as the cleanest small-cap name: it is the number one NAND controller maker, revenue is up 71%, it trades at 18x forward earnings, and it pays a dividend.
PLAB has about $500 million in cash on a $1.7 billion market cap, trades at 10x trailing earnings, and its Texas facility is finally shipping after being flagged in March.
ATOM’s trailing revenue is only $233,000, but it says it is in talks with large memory makers for next-gen DRAM and HPM and could sign deals in 2026; if it lands one, the stock could rise 5x to 10x, and if not it could drift to zero.
BSEM is BioStem Technologies on NASDAQ, and in January 2026 it bought BioTissue’s surgical and wound-care assets, shifting from a physician-office wound business to a hospital/surgery-center model with direct sales, GPO access, and 87% procedure-based reimbursement.
BSEM says the US soft-tissue allograft opportunity is about $26 billion, it has 4 GPO contracts covering about 2/3 of US hospital beds, more than 400 publications, and more than 1.2 million product applications backing the platform.
BSEM’s Pompano Beach facility has about 4x the capacity it uses today, the acquired Neox and Clerix products are still made by BioTissue under cost-plus supply, and management targets bringing that manufacturing in-house in 1H27.
BSEM issued 8 new US design patents in 2Q26, but the name is still tiny, thinly traded, cash-burning, and likely to need another capital raise, so dilution and execution risk remain central.