We use Google Analytics to count anonymous page views and understand which content gets read. No ads, no profiles. Decline keeps you on cookieless mode. Details.
Elite Semiconductor Microelectronics Technology Inc. (TWSE: 3006)
Information Technology · Semiconductors
Structural read: a niche legacy-DRAM designer riding the capacity cascade out of HBM/DDR5 investment, into a DDR4 structural supply gap, down-spec substitution into DDR3, and finally a DDR2 squeeze as Winbond exits that tier. Fabless (foundry: PSMC, unowned) -- the wafer-cost pass-through, not demand, is the swing factor on margin.
- July 2026 revenue hit a monthly record, NT$6,785M (~$213M), +40% MoM and +491% YoY, on niche-DRAM price hikes.
- Q2 2026 net income NT$5.908B / EPS NT$20.21, both all-time highs; Jan-Jul 2026 net income ~$359.6M on ~$879M revenue.
- Balance sheet: ~$455.7M cash (~$395.9M net cash), ~$943.5M assets, ~$596.0M book equity against a ~$2.0B market cap -- real cushion, not a pure multiple story.
- Winbond (its own May 2026 investor deck) frames the DDR4/LPDDR4 supply gap as structural through 2028+, "no near-term resolution," and is gradually exiting DDR2 -- direct share transfer to ESMT.
- Even under a stress-tested wafer-cost headwind (PSMC +45%, 100% pass-through assumed), napkin math still implies a mid-single-digit annualized P/E.
- Does not own its fabs. PSMC raised DRAM wafer-start prices ~45% in July 2026 (flowing into COGS from Nov); as the unowned foundry, PSMC -- not ESMT -- controls how much of any further price hike gets captured upstream. No Kioxia/SanDisk-style vertical alignment.
- The revenue ramp (Feb 2026 $65.6M -> July $212.8M, a ~3.2x run in five months) is a hockey stick; legacy-DRAM upcycles have a history of pull-forward/stockpile buying ahead of anticipated price hikes that unwinds hard once it stops.
Each weekday: the scan's ENTER setups with entry, stop and target, every new ticker teardown, and the bubble shifts that move names like this one. Free. Education, not advice.
No spam. One email per day max. Pro adds Telegram trade alerts and higher AI-assistant limits.
- The July print (and the "~2x P/E" headline built on annualizing it) only became public because TWSE forced a supplemental disclosure after abnormal trading in the shares (2026-08-05) -- regulatory and market attention was already on the name before the number was public.
- Legacy DRAM cycles are historically vicious on the downside (2022-2023 memory crash pattern): if panic/stockpiling buying unwinds or fab capacity reallocates back to legacy nodes, pricing can crater within 2-3 quarters.
No major news in the last 7 days for ESMT - only listicles and opinion pieces, which we filter out by default. .