Management beat by 33% on earnings and then guided Q1 gross margin down 630 basis points in the same breath.
Thesis: The DRAM bottleneck is physical and management quantified it: $150-200M of ion implant capital per 100K wafer starts across announced mega fabs that won't absorb tools until early 2027. But 2026 is a bridge year of flat revenue with gross margins falling from 45.2% to low-to-mid 40s, because the segment that grows (memory at lower system margins) offsets the segments that decline (power and general mature). The stock's edge depends entirely on whether forecast-driven DRAM POs convert to bookings before Q2. Memory books and ships in the same quarter, so backlog gives zero visibility into the thesis.
Verdict: HOLD — Conviction: MEDIUM
Catalyst: DRAM tool orders materializing as clean room space opens across SK Hynix Yongin and other mega fabs in early 2027, visible as sequential bookings acceleration beyond $128M in Q2-Q3 2026.
Key Risk: Q2 systems pushouts extend beyond one quarter as clean room readiness slips, pushing 2026 flat revenue into a decline while gross margin lands below low-to-mid 40s as memory mix grows faster than modeled.
The Tell: Russell volunteered unprompted concern about the DRAM cycle itself: 'whenever I hear about long-term contracts being signed, it always makes me a bit worried because I have kind of seen these happen before... they call it RAMageddon.' The CEO bullish on DRAM simultaneously flagging cycle-overheat risk in the same breath. He also volunteered 'I was in Korea last week' before Jamie's multi-quarter forecast confidence, showing he's building internal conviction faster than external guidance reflects.
Friction Level: MODERATE_FRICTION — Both sides agree 2026 is flat revenue with declining margins. Bull says buy the pullback for 2027 DRAM tool orders from clean room space releasing. Bear says the growth segment is the margin headwind and memory revenue never shows in backlog. Disagreement is on whether forecast-driven POs justify a full position now.
2025Q4 · Feb 17, 2026
4 of 5 readings agreed4 HOLD · 1 LONGRead the full call
Flat 2026 Guide, DRAM Promise in 2027, Margins Falling Either Way
The DRAM bottleneck is physical and management quantified it: $150-200M of ion implant capital per 100K wafer starts across announced mega fabs that won't absorb tools until early 2027. But 2026 is a bridge year of flat revenue with gross margins falling from 45.2% to low-to-mid 40s, because the segment that grows (memory at lower system margins) offsets the segments that decline (power and general mature). The stock's edge depends entirely on whether forecast-driven DRAM POs convert to bookings before Q2. Memory books and ships in the same quarter, so backlog gives zero visibility into the thesis.
Russell volunteered unprompted concern about the DRAM cycle itself: 'whenever I hear about long-term contracts being signed, it always makes me a bit worried because I have kind of seen these happen before... they call it RAMageddon.' The CEO bullish on DRAM simultaneously flagging cycle-overheat risk in the same breath. He also volunteered 'I was in Korea last week' before Jamie's multi-quarter forecast confidence, showing he's building internal conviction faster than external guidance reflects.
Russell confirmed 'memory typically books and ships in the same quarter,' meaning the $457M backlog excludes the DRAM segment management claims will drive growth. Backlog is general mature and power, both declining. The thesis stock has zero visibility in the number the bulls cite.
GM guides from 47.3% to 41% in Q1 and low-to-mid 40s for FY2026 versus 45.2% in 2025. Management attributes decline to memory mix, CS&I seasonality, and tariffs. Memory growth is simultaneously the growth story and the margin headwind. Jamie quantified it precisely: higher memory volume in Q1 drives the guide down.
Q4 CS&I hit a record $82M but Jamie disclosed pull-forward from Q1, a non-recurring large customer SiC upgrade, and systems pushouts on clean room readiness. Three separate sources of artificial Q4 strength disclosed in the Q1 guide breakdown.
China fell from 46% to 32% of revenue sequentially while the street credited the beat to non-China strength. FY2026 China guided 'relatively flat to down slightly' from a 42% base. Full-year 2025 China was 42% of revenue with no structural replacement market identified.
Management frames rising utilization in general mature as positive ('once those tools are up to high utilization rates, that is when they start to buy more tools'), but also admits 'some of our customers state that their CapEx may not be that high this year.' The ceiling is real: utilization must run high before capex follows, and that lag means 2026 tool orders stay soft.
Q4 beat on revenue ($238M vs implied trajectory), non-GAAP EPS ($1.49 vs $1.12), and gross margin (47.3% vs 43% guide). CS&I at $82M record. FCF $107M full year on flat-to-down revenue. Management sandbagged and the pattern holds.
New high-current win at North American memory manufacturer fanning out across applications, above the prior $90M DRAM high-water mark that excluded this customer. SK Hynix Yongin alone is 4 mega fabs at 200K wafer starts/week with $150-200M implant capex per 100K wafer starts. Announced, not speculative.
Jamie walked through the Q1 guide line by line with three named sources of decline, quantified tariff drag at under 100bps, and explained the margin dip by component. Russell gave specific capex intensity math and fab capacity numbers. No deflection on any direct question.
-25
CAUTIOUS"Repeated deferral of recovery calls: 'we wanted to kind of see it in the data before we kind of were willing to put that out there.' Qualified guidance on memory ('healthy number,' no specific revenue target). Hedged on 2027: 'we will be sort of moderate in our tone.' Volunteered specifics on Q1 guide breakdown but refused quantified memory outlook."
Rich specifics on ion implant capex intensity ($150-200M per 100K wafer starts), SK Hynix Yongin fab details, prior DRAM peak ($90M), and transparent Q1 guide breakdown. Gaps: no quantified 2026 memory revenue target, no disclosed customer concentration data, backlog quality unquantified.
DRAM tool orders materializing as clean room space opens across SK Hynix Yongin and other mega fabs in early 2027, visible as sequential bookings acceleration beyond $128M in Q2-Q3 2026.
Q2 systems pushouts extend beyond one quarter as clean room readiness slips, pushing 2026 flat revenue into a decline while gross margin lands below low-to-mid 40s as memory mix grows faster than modeled.
“Starting on slide 10, fourth quarter revenue was $238 million with systems revenue at $156 million and CS&I revenue reaching another quarterly record of $82 million, both above our expectations for the quarter.”
“We expect revenue in the first quarter of $195 million.”
Figures are extracted from the transcripts and checked in code: every quote appears word for word. Descriptive only: none of this feeds the verdict.
Beat and raise with physical constraint backing it. Ion implant capex of $150-200M per 100K wafer starts across three DRAM manufacturers with announced mega fabs. SK Hynix Yongin alone: 4 fabs at 200K wafer starts per week, tools shipping early 2027. New North American memory customer winning high-current applications, above the prior $90M DRAM peak that excluded this customer. Clean room space is the binding constraint, not tool demand, and it releases in 2027. Backlog $457M, CS&I compounding 14%, $557M cash, $107M FCF buying back stock. Management called the memory recovery correctly last year when the street doubted them. Q1 guide is timing noise: three named non-recurring items (SiC upgrade, CS&I pull-forward, clean room pushout). Multi-quarter forecasts are 'very sound' with requirements sometimes exceeding them.
The guide tells the real story. Q1 revenue drops $43M sequentially across three admitted sources: non-recurring SiC upgrade, CS&I pulled forward from Q1 into Q4, and systems pushouts on clean room readiness. GM falls 630bps to 41%, then full year lands at low-to-mid 40s versus 45.2% in 2025. The growth engine is the margin headwind. Memory books and ships in the same quarter, so the $457M backlog that bulls cite contains zero DRAM visibility. China was 42% of 2025 revenue, guided flat-to-down, power and general mature both decline. FCF was negative $9M in Q4. PO lead times are 'a week, a month, and maybe a quarter' ahead of shipment with no long-term contracts. The Veeco merger burns $5M per quarter in transaction costs. The entire thesis rests on clean room space opening in 2027 and forecast-driven POs converting on time.
Backlog Quality
Russell explicitly stated memory books and ships in same quarter. The $457M backlog cited as bull evidence contains general mature and power, both declining. Zero backlog visibility into the DRAM thesis.
Mix Shift Spin
GM guiding down 630bps sequentially to 41% and low-to-mid 40s for FY2026 versus 45.2% in 2025. Memory growth drives both the top line and the margin compression. Full year 2026 guide has flat revenue with declining profitability.
Say/Do Gap
Q4 CS&I record $82M included pull-forward from Q1, non-recurring SiC upgrade, and budget flush seasonality. Jamie disclosed all three sources in the Q1 guide breakdown. The beat was partially manufactured.
China Revenue Decline Masked
China fell 46% to 32% sequentially in Q4. FY2026 China guided flat to down from 42% base. The beat was credited to non-China strength while the largest single geography declines with no disclosed offsetting region.
Unsustainable Trend Confidence
Q4 gross margin at 47.3% driven by record CS&I volume and upgrade mix, both of which management guides down for Q1. The margin beat was elevated by factors management explicitly says won't repeat.
HOLD. The DRAM constraint is real and quantified but gated: clean room space releases tools in early 2027, and today's POs arrive with week-to-quarter lead times on forecast confidence, not contracted backlog. 2026 is flat revenue with declining gross margins. Asymmetry is not wide enough for a full position. Wait for sequential bookings above $128M in Q2 as confirmation DRAM POs are converting. If Q2 bookings accelerate on DRAM, that's the entry point. If they don't, the 2027 story slips a year while the stock carries declining margin estimates. Hold the position flat, size up on order rate confirmation.