Compare korea memory selloff CVLT SNDK KLAC AMKR GLW

Date: 2026-07-28 (prices intraday, ~12:48 ET — market open) Command: /analyze CVLT SNDK KLAC AMKR GLW — "recent south korean tech meltdown is dragging a lot of top names down with it. Any opportunities in this?" Framework: analysis_notes.md §0–§5

⚠️ Data notes. Yahoo Finance MCP and roic.ai MCP are not connected this session. Quantitative data is from .mcp/fin.py (direct yfinance — selftest passed) for annual statements, a direct yfinance pull for quarterly statements (fin.py is annual-only), and live Public.com quotes. Qualitative from WebSearch. Where this limits depth (8-yr CAGRs, pre-computed ROIC) it is flagged inline.

⚠️ Wire-service error caught. StockStory/FinancialContent reported Commvault guided next-quarter revenue to "$266M, 13.7% below estimates" and that "ARR missed." Both are wrong — $266M is the subscription revenue guide being compared against a total revenue consensus, and $1.05B is subscription ARR being compared against a total ARR estimate. Verified against the primary press release. See §5.

1. First: the premise needs correcting

The user's framing is that a Korean tech meltdown is "dragging top names down." That is half right, and the half that's wrong matters more.

What actually happened to these five in the last 48 hours:

Ticker	1-day	Cause	Korea-driven?
AMKR	−24.3%	Own Q3 guide ($1.95–2.05B vs $2.12B cons.)	Partly — SiP migration out of Korea + memory-driven Android weakness
GLW	−16.1%	Own Q2 revenue miss + soft Q3 revenue guide	No — own results
CVLT	−16.1%	Own Q1 FY27 print — multiple reset, not a miss	Indirectly — memory shortage deferring on-prem deals
SNDK	−14.0%	CXMT IPO + memory de-rating	Yes — direct
KLAC	−5.5%	Sector sympathy; reports FQ4 tonight	Partly — memory = 15–25% of revenue

Four of the five fell on their own earnings, reported within the last 48 hours. This is earnings season, not contagion.

And critically — every one of these is still up enormously over twelve months:

Ticker	~1yr ago	2026 peak	Now	Off peak	1yr total
GLW	$55.41	$271.78	$120.27	−55.7%	+117%
AMKR	$21.23	$96.68	$45.93	−52.5%	+116%
KLAC	~$87	$307.37	$192.12	−37.5%	+120%
SNDK	~$43	$2,354	$1,098.37	−53.3%	+2,450%
CVLT	~$190	$200.68	$125.36	−37.5%	−34%

This is not a dislocation creating value in quality names. It is the first serious crack in a parabolic 2026 melt-up. The framework's §0 test — "has the market already priced that in?" — mostly still answers yes, even after a 50% drawdown, because the drawdown is measured from a price that had tripled.

CVLT is the one exception: it is genuinely below where it was a year ago.

The tell that this isn't broad AI panic: on the same day, AVGO closed flat (+0.02%). AMD −7.1%, INTC −5.3%. The selling is concentrated in memory and memory-adjacent hardware, not AI broadly.

2. What actually happened in Korea (and why it matters)

A genuine memory supercycle ran through 2026 — Gartner has NAND pricing +234% in 2026, with relief not expected until late 2027. It broke in July:

Date	Event
Jul 2	KOSPI −6–8%; SK Hynix −14.6%, Samsung −9.1%. Emergency trading pause. AI-capex sustainability fears.
Jul 10	SK Hynix lists on Nasdaq — $26.5B raised at $149/ADS, largest-ever US listing by a foreign company. Closes day 1 at $168 (+13%).
Jul 13	SK Hynix weak outlook → MU/SNDK/WDC −6%.
Jul 15–24	Successive profit-taking waves. SK Hynix ADR falls below its $149 IPO price (~$137, ~−30% from post-debut high).
Jul 27	CXMT (ChangXin Memory) IPO on Shanghai STAR — raises $8.6B, closes +466–531%, becoming the most valuable mainland-listed company. China's #4 global DRAM player (7.67% share) is now capitalized with a war chest.
Jul 28	Combined Samsung + SK Hynix lose ~$290B in a day. Memory complex enters a bear market (>20% off highs).

Two distinct things are being repriced, and they are opposite in sign:

(a) The durability of memory-maker margins. SanDisk's gross margin went 22.5% → 78.3% in four quarters on pure price. CXMT's IPO capitalizes the marginal supplier. The market is now asking whether 78% is a plateau or a peak. (It is a peak. See §6.)

(b) The memory shortage as a cost tax on everyone else. This is the underappreciated half, and it explains three of these five names:

AMKR: Android revenue −20% — management explicitly blames memory supply/pricing dynamics crushing phone unit volumes.
CVLT: term-software net-new ARR flat sequentially — management cites "hardware supply constraints and shorter deal terms."
IBM (7/14, from Watchlist.md): pre-announced Q2 miss because clients "shifted late-June capex to supply-constrained memory/servers."

The insight worth keeping: high memory prices are a transfer from the entire computing supply chain to five memory makers. When memory normalizes, SNDK/MU/SK Hynix give back the margin — and the victims (server OEMs, Android, on-prem software budgets) get relief. Positioning for the memory crack means owning the victims, not the winners. None of these five except CVLT is a clean expression of that.

3. Fundamentals scorecard (§1)

Quarterly, because the entire story is 2026 quarterly dynamics.

Revenue & margin trajectory (last 5 reported quarters)

	Q1'25	Q2'25	Q3'25	Q4'25	Q1'26	Latest (Q2'26)
SNDK rev	$1.70B	$1.90B	$2.31B	$3.02B	$5.95B	Aug 5
SNDK gross margin	22.5%	26.2%	29.7%	51.0%	78.3%	—
KLAC rev	$3.06B	$3.17B	$3.21B	$3.30B	$3.42B	tonight
KLAC op margin	42.5%	42.6%	41.7%	41.2%	41.2%	—
AMKR rev	$1.32B	$1.51B	$1.99B	$1.89B	$1.68B	$1.90B
AMKR gross margin	11.9%	12.0%	14.3%	16.7%	14.2%	16.8%
GLW rev	$3.45B	$3.86B	$4.10B	$4.21B	$4.14B	$4.51B
GLW op margin	12.9%	14.8%	14.4%	16.0%	15.4%	15.5%
CVLT rev	$275M	$282M	$276M	$314M	$312M	$314M
CVLT gross margin	82.7%	82.0%	80.0%	81.1%	81.4%	81.6%

Free cash flow — the framework's backbone (§1.1)

Ticker	FY2025 FCF	FY2025 FCF margin	TTM/run-rate FCF	2026 capex plan	Verdict
KLAC	$3.74B	30.8%	~$4.05B	~$390M (3% of rev)	🟢 Elite — asset-light, 31% FCF margin
GLW	$1.41B	9.0%	~$2.39B	~$1.5B	🟢 Inflecting — Q2'26 FCF margin 28.7% vs 11.1%
CVLT	$237M	20.1%	~$250–260M (guide)	~$8M	🟢 Clean — asset-light, FCF +71% YoY
SNDK	−$120M	neg.	$2.99B in one quarter	~$180M	🟡 Cycle-peak — real cash, not durable
AMKR	$191M	2.8%	negative	$2.5–3.0B	🔴 Broken — see below

AMKR is the standout problem. FY2025 FCF was $191M on $6.71B revenue — a 2.8% FCF margin in a good year. Management has now guided 2026 capex to $2.5–3.0B, versus $905M in 2025 — roughly a 3× step-up, against OCF of maybe $1.4–1.6B. That implies FCF of roughly −$1.0B to −$1.6B in 2026, and Arizona Phase 1 spending is still mostly ahead (completion 2027, production 2028). The company confirms $1.8B cash / $2.9B liquidity / 1.1x debt-to-EBITDA — it can fund it, but there is no owner's cash for years.

Per-share & capital allocation (§1.4, §1.6):

Ticker	Share count trend	Buybacks FY25	Dividend	Read
KLAC	−4.1%/yr	$2.15B	0.45% yld, 22% payout	🟢 Best allocator here
CVLT	−0.0%/yr (3y)	$545.7M (FY26)	none	🟡 Aggressive — debt-funded, see §5
AMKR	+0.3%/yr	none	0.55% yld	🔴 Diluting slightly, capex-starved
GLW	+0.5%/yr	$163M	0.78% yld, 54% payout	🟡 Buyback is token; dilutes
SNDK	+8% (145→157M dil.)	none	none	🔴 Diluting into the boom

Framework flag (§1.4, "per-share is what I own"): only KLAC is meaningfully shrinking its share count. SNDK has grown diluted shares ~8% in five quarters — during the single greatest earnings boom in its history, shareholders own less of it.

Balance sheet (§1.3)

Ticker	Debt/Assets	Net cash/(debt)	Note
SNDK	15.7%	~net cash	Cleanest
AMKR	19.2%	~neutral ($1.8B cash)	But $2.5–3B capex ahead
KLAC	37.9%	−$4.0B	Comfortable at 31% FCF margin
GLW	30.3%	−$7.9B	Manageable, funds Springboard
CVLT	48.6%	−$17.5M	⚠️ Was 1% a year ago — see §5

4. Valuation (§3) — the decisive section

Graham's IV is computed for completeness but is misleading for four of these five and I'm saying so rather than hiding it: KLAC (BVPS $4.46) and CVLT (BVPS $0.18) have had book value destroyed by buybacks, and GLW/SNDK have GAAP EPS distorted by charges/cycle. Graham is only meaningfully weighted for AMKR, the one asset-heavy, book-backed name.

	CVLT	SNDK	KLAC	AMKR	GLW
Price	$125.36	$1,098.37	$192.12	$45.93	$120.27
Market cap	$5.2B	$163.5B	$251.0B	$11.4B	$103.5B
P/E (ttm)	79.5x	37.7x	54.3x	26.5x	57.8x
P/E (fwd)	20.9x	5.2x	37.5x	16.1x	28.0x
P/B	690x*	11.8x	43x*	2.46x	9.2x
FCF yield	~4.9%	(peak)	1.6%	negative	2.3%
Graham IV	$2.54 ✗	$247.60 ✗	$18.85 ✗	$27.10 ✓	$24.68 ✗
Analyst target (mean)	$159	$2,218	$234	$79	$215
Analysts	16	22	28	9	15

* P/B meaningless — equity near zero from buybacks.

Dividend Yield Theory (§3, dividend payers only):

GLW: current yield 0.78% vs. 5-yr average 2.58%. DYT implies fair value near $36–45. I weight this low — Corning's revenue mix has genuinely changed (Optical is now 46% of revenue growing 32%+, so the historical yield band prices a different company) — but it is a real signal that even after −56%, GLW is nowhere near its own historical valuation floor. Named as a risk factor rather than buried.
KLAC: 0.45% vs. 0.93% 5-yr average. Same direction, less extreme.
CVLT, SNDK: no dividend — N/A.

Bogle expected return (§3), 5-yr, stated assumptions:

	Div yield	+ Earnings growth	± P/E change	= Expected
KLAC	0.45%	~15%	37x→30x = −4.1%/yr	~11%/yr
GLW	0.78%	~19% (Springboard)	28x→22x = −4.7%/yr	~15%/yr
CVLT	0	~16%	21x→18x = −3.0%/yr	~13%/yr
AMKR	0.55%	~10% (low visibility)	16x→14x = −2.6%/yr	~8%/yr
SNDK	0	undefined	undefined	not modelable

5. Per-name analysis

🟡 CVLT — Commvault | | the only real "multiple reset on an intact business"

The reported story was wrong. Wire services said CVLT missed on ARR and guided revenue 13.7% light. Verified against the primary release, the actual Q1 FY2027 print (Jul 28) was:

Metric	Actual	YoY
Total revenue	$314.1M	+11% (beat $310.5M)
Subscription revenue	$267.0M	+16%
SaaS revenue	$100.6M	+39%
Subscription ARR	$1,054.3M	+22%
SaaS ARR	$424.3M	—
Gross margin	81.6%	stable
Non-GAAP EBIT margin	22.8%	expanding
Free cash flow	$51.1M	+71%
Non-GAAP EPS	$1.42	vs $1.16 est (+22%)

Guidance was raised, not cut: FY27 subscription revenue raised to $1,119–1,129M (~16%); subscription ARR reiterated $1,200–1,210M (~19%); non-GAAP EBIT margin raised 50bps to ~21%; FCF $250–260M. The "$266M" that wires called a miss is the Q2 subscription revenue guide ($264–268M), compared against a total revenue consensus.

So why did it fall 16%? Three real reasons, none of which is a fundamental break:

Valuation reset. The stock had run +66% in six months ($77.89 in March → $141.73 in June) into a 94x GAAP P/E. The beat was margin-driven, not growth-accelerating — software markets punish that at a premium multiple.
AI monetization deferred. Management said outright that "AI-related workloads are not yet a major contributor to guidance." AI Studio / Data Activate / AI Protect have no near-term revenue attached.
The memory tax. Term software net-new ARR flat sequentially, attributed to hardware supply constraints and shorter deal terms — the Korea/memory link, arriving as a cost problem for customers.

🚩 Two genuine red flags the drop does not explain away:

Securities class action (D.N.J.) against the company, CEO Sanjay Mirchandani, and former CFO Jennifer DiRico. Class period Apr 29 2025 – Jan 26 2026. Allegation: ARR guidance failed to account for how the SaaS-vs-term-license mix shift suppresses net-new ARR. On Jan 27 2026 net-new ARR came in at $39M vs. $45M guided and the stock fell >31%. Lead plaintiff deadline was Jul 17 2026. This is a live credibility problem about the exact metric that is still soft today. CFO has since changed (Gary Merrill, Apr 2026).
Balance sheet transformed in one year. Debt went $11M → $917.5M; equity $325M → $7.5M; Debt/Assets 1% → 48.6% — a $545.7M debt-funded buyback at an average price well above today's. Framework §1.2 explicitly warns on "buybacks at rich prices." This one was executed near the highs.

Valuation. At $125.36 / $5.2B cap against FY27 FCF guide of $250–260M → ~20x FCF for a business growing subscription ARR 22% at a 22.8% EBIT margin. That is reasonable, not cheap. Fwd P/E 20.9x. Analyst mean target $159 (range $100–200).

Fair value: $115–150 (base ~$132). Verdict: HOLD. Do not add here. Add only $100–110 — and only if Q2 FY27 shows term-software net-new ARR turning up. Trim $160+.

Conviction: [6.0] — down from an implied ~6.5–7. Business is fine; governance and the ARR-metric credibility gap are not.

A portfolio-specific passage was removed from the public build.

🟢 KLAC — KLA Corp | the best business, and now in its own accumulate zone | ⚠️ REPORTS TONIGHT

Continuity with prior work: compare-ASML-KLAC-LRCX-2026-07-15.md rated KLAC business quality 9.0/10, conviction [8.0], "the one I'd most want to own, at a better price," with an accumulate zone of $180–200 set when the stock was near $300 at 44x forward.

KLAC is now $192.12 — inside that zone, at ~37.5x forward. That is the single most concrete development across all five names.

Business quality is not in question (§2). FY2026 (ending Jun'26): revenue ~$13.5B (+11%), operating margin 41.2%, gross margin 61%, ROE 95%, FCF ~$4.05B at a 31% FCF margin. It holds a majority share of semiconductor process control — the inspection/metrology toll booth every fab must pay, and process-control intensity rises with each node. Management expects process control to grow >20% in 2026, outpacing broader WFE. Share count −4.1%/yr. This is a genuine wide-moat compounder.

Adversarial stress-test (§2.2): a well-funded rival cannot meaningfully attack this. Process control requires decades of accumulated defect-library data, fab-floor integration, and trust at the leading edge; ASML doesn't compete here, and Applied/Hitachi have tried for years without displacing KLA. The moat is intact. The threat is not competitive — it is cyclical and geographic.

What changed since 7/15 — and why I'm tightening the zone:

Memory is 15–25% of KLA's revenue, led by SK Hynix, Samsung and Micron. South Korea is 15–20% of geographic revenue. That is precisely the customer base whose equity value just fell ~$290B in a day and whose marginal competitor (CXMT) just IPO'd with $8.6B.
If memory capex plans get trimmed into 2027, KLA's FY27 consensus EPS of ~$5.13 (+39% on FY26's ~$3.60) is the number at risk. The multiple isn't the fragile part — the estimate is.
China export-control exposure remains an open risk.

Valuation. At $192.12: 53x trailing FY26 GAAP EPS (~$3.60), 37.5x forward (~$5.13). Context: sector median forward P/E 29.6x; KLAC's own 5-yr trailing average 30.5x, 3-yr 33.1x. FCF yield 1.6%. Analyst mean target $234 (28 analysts, range $150–325). So: still a premium multiple, but ~15% cheaper than 13 days ago and no longer at target.

At 30–36x FY27E $5.13 → fair value $155–185 (base ~$170).

Verdict: 🟡 WATCH → STAGED BUY, but not today. FQ4 FY2026 results are released after the close tonight (Jul 28, 2 p.m. PT), with guidance for the September quarter. Buying a 37x-forward name hours before a print — during a sector de-rating that directly hits 15–25% of its customers — is not a risk worth taking for a long-term holder. Wait for the print.

Accumulate $165–190 (tightened from the prior $180–200 — memory-capex risk is materially higher than on 7/15)
Strong buy <$160 (200-day was ~$156)
Trim $260+
Thesis break: process-control revenue growth <10% for 2 quarters, memory customers publicly cutting 2027 capex, or FY27 EPS consensus falling below ~$4.50.

Conviction: [7.5] — down from [8.0]. The business earns 9/10; the estimate risk from memory capex is new and real.

🟡 GLW — Corning | a genuinely transformed business at a fair, not cheap, price

Q2 CY2026 (Jul 28): revenue $4.51B (+11.4%) — missed $4.65B consensus by 3%. Adjusted EPS $0.78 beat ($0.75). Operating margin 15.5% (from 14.2%). FCF margin 28.7%, up from 11.1%. Q3 guide: revenue $4.95B (vs $5.04B est — a 1.7% miss) but adj EPS $0.87 vs $0.85 est — a beat.

Read carefully: Corning missed on revenue and beat on earnings, both quarters. The market sold a 3% revenue miss in a stock that had tripled. The profitability trajectory is improving, not deteriorating.

The moat has genuinely changed shape (§2). This is no longer primarily a display-glass company:

Optical Communications: $2.07B, +32%, now 46% of revenue — with Enterprise Networks +65% and Gen-AI product sales growing faster still.
NVIDIA partnership: Corning to expand US optical connectivity manufacturing 10× and US fiber capacity +50% for AI factory buildouts.
Meta fiber expansion agreement.
New Photonics market-access platform targeting a $10B revenue stream by 2030.
Springboard plan (upgraded and extended May 2026): $20B annualized run-rate by end-2026 → $30B end-2028 → $40B end-2030, a 19% sales CAGR Q4'26→Q4'30, with earnings growing faster than sales.

Corning's moat sources are real: process intangibles (fusion draw, low-loss fiber IP), efficient scale, and now genuine customer co-investment lock-in with NVIDIA/Meta. Fiber is a hard business to enter — but it is also historically a brutally cyclical, capacity-driven commodity whenever the buildout pauses. That is the disruption vector, and it is a real one: this is the second time in 25 years Corning has been priced as an infinite-optical-demand story.

Korea exposure: Display Technologies (32.5% of revenue) runs through Corning's large Korean operation tied to Samsung Display, and management flagged weak consumer demand and inflationary pressure compressing Display margins. Real but secondary.

Valuation. At $120.27: ~36.6x FY26E adj EPS (~$3.28), ~28.0x FY27E (~$4.29). TTM FCF ~$2.39B → 2.3% FCF yield. Analyst mean target $215.47 with a low of $155 across 15 analysts — i.e. every published target sits above the current price (though most predate today's −16%).

The tension I won't bury: DYT says fair value $36–45. Bogle says ~15%/yr expected return. A 28x FY27 multiple on a 19%-CAGR guide says "fair." These disagree violently. Weighting by company type (§5.2) — Corning is now a growth-capex industrial, not a dividend stalwart — I weight Bogle and forward FCF over DYT, but the DYT reading is a legitimate warning that this stock has never been valued this way before.

Fair value: $105–135 (base ~$120). The stock is at fair value, not below it.

Verdict: 🟡 WATCH — add to Watchlist, do not buy yet. Entry $95–110; strong buy <$90. Trim $180+. Thesis break: Optical Communications growth <15% for 2 quarters, Springboard run-rate targets pushed out, or an NVIDIA/Meta capacity agreement reduced.

Conviction: [6.5] — good business, real AI franchise, no margin of safety at $120.

🔴 AMKR — Amkor | the value trap of the group

Down 52.5% from its high and still not cheap — because the "E" in its 16x P/E is not backed by cash.

Q2 2026 was genuinely a record: revenue $1.9B (+26% YoY, +13% sequentially), EPS $0.70 (beat by 49%), gross margin expanded 250bps to 16.8%. Advanced packaging (High-Density Fan-Out) is real, with TSMC and NVIDIA partnerships and AI advanced-packaging revenue guided to triple in 2026. 2030 targets: $11B revenue at 22% gross margin.

And yet:

16.8% gross margin at a record quarter. OSAT is structurally low-margin subcontracting. Compare KLAC at 61%, CVLT at 81.6%. This business has almost no pricing power — its customers (TSMC, NVIDIA, Apple) are far stronger than it is, and TSMC is simultaneously its partner and its most dangerous competitor in advanced packaging (CoWoS in-house). Framework §2.2 adversarial test: a well-funded rival — or a customer — can attack this, and TSMC already is.
Capex $2.5–3.0B in 2026 vs. $905M in 2025 — roughly 35–40% of revenue. FCF goes to roughly −$1.0B to −$1.6B, and Arizona Phase 1 doesn't produce until 2028. There is no owner's cash for years.
Q3 guide $1.95–2.05B vs. $2.12B consensus. Management attributes the shortfall 50% to market factors (memory constraints, build patterns) and 50% to its own SiP migration from Korea to Vietnam — a self-inflicted disruption now extending into late 2026 and H1 2027.
Android revenue −20%, directly attributed to memory supply dynamics. The memory tax lands here hardest.

Valuation. 16.1x forward EPS (~$2.86) looks superficially cheap, and it is the one name where Graham applies (asset-backed, book-real): Graham IV $27.10 vs. price $45.93 — the stock is 70% above Graham's intrinsic value. P/B 2.46x. Analyst mean $79 (only 9 analysts; B. Riley just cut to $65, Neutral) — these targets look stale against a −24% day and a guide-down.

Fair value: $32–45 (base ~$38), using 13–16x a normalized ~$2.50 EPS with a discount for negative FCF and the capex cycle.

Verdict: 🔴 AVOID. This is the framework's "poor business, cheap-looking price" = value trap quadrant (§0). A −52% drawdown in a low-margin, cash-consuming subcontractor mid-way through a three-year capex build is not a bargain. Revisit only if it reaches high-$20s–low-$30s and FCF inflects positive.

Conviction: [4.0]

🔴 SNDK — SanDisk | not an investment, a cycle bet at the peak

The numbers are genuinely spectacular, and that is exactly the problem.

Quarter	Revenue	Gross margin	Op margin	EPS
Q1 CY2025	$1.70B	22.5%	−2.5%	−$13.33
Q2 CY2025	$1.90B	26.2%	2.7%	−$0.16
Q3 CY2025	$2.31B	29.7%	8.3%	$0.75
Q4 CY2025	$3.02B	51.0%	35.4%	$5.15
Q1 CY2026	$5.95B	78.3%	69.9%	$23.03

Revenue +251% YoY on roughly flat units. That is essentially all price. Gross margin went from 22.5% to 78.3% in four quarters. FCF was $2.99B in a single quarter.

A memory company does not earn a 78% gross margin through a cycle. The 2017–18 NAND peak topped out around 45–50%. The same operating leverage that produced $23.03 of quarterly EPS runs in reverse with equal violence — Q1 CY2025, fifteen months ago, lost $1.93B.

Supply is coming. Kioxia/SanDisk are raising capex +41% YoY to $4.5B; YMTC is aggressively adding wafer capacity; and CXMT just raised $8.6B. The bull rebuttal is real — Korean and US makers are freezing NAND wafer input to prioritize HBM/DRAM, and Kioxia insists 2027 capacity won't break balance — but "the marginal supplier is disciplined" is the assumption every commodity cycle dies on.

Valuation is unresolvable, and I won't fake precision (§0, §3):

Method	Implied value
5.2x consensus forward EPS (~$213)	market says cheap
Goldman: 20× normalized EPS $110	$2,200
11.9x annualized peak EPS ($92)	current price
Graham IV √(22.5 × 29.27 × 93.09)	$247.60
Full normalization (30–35% GM, ~$10–16B rev, 12–15x)	$200–350
P/B 11.8× (memory makers historically 1–3×)	deeply expensive
Analyst range (22 analysts)	$1,000 – $3,169

The honest fair-value range is roughly $250 to $2,200 — a 9× spread. Any point estimate inside it is fiction. The 52-week range ($40.10 → $2,354, a 59× band) says the same thing.

Add: diluted shares grew 145M → 157M (+8%) through the boom, so per-share ownership is shrinking during the best five quarters in company history (§1.4).

Verdict: 🔴 AVOID — not a fundamental investment. This fails the framework at §0: you cannot answer "has the market priced this in?" when normalized earnings power is unknowable to within an order of magnitude. It may well go higher — NAND pricing has real momentum into 2027 and FQ4 reports Aug 5 — but that is a trade, not a buy-and-hold thesis. Conservative bias (§3): rather miss an opportunity than overpay.

Conviction: [3.5] — as a long-term holding. This says nothing about its near-term direction.

6. Conflict resolution (§5)

Tension 1 — Valuation vs. Sentiment on GLW/KLAC. Sentiment points to every analyst target sitting above the current price (GLW $155–270 vs. $120; KLAC $150–325 vs. $192). Valuation says both are still above historical multiples. Resolution: targets were largely set before today's prints and lag by days-to-weeks; the framework weights price paid over consensus (§0). Valuation wins, but the gap is wide enough to say GLW at $120 is fair rather than expensive.

Tension 2 — Moat vs. Valuation on KLAC. Moat says 9/10, buy the toll booth on any weakness. Valuation says 37x forward with 15–25% of revenue exposed to customers in an active de-rating. Resolution (§5.2 — weight by company type): for a wide-moat compounder, Moat should outweigh a static multiple. But the risk here isn't the multiple, it's the FY27 estimate. Both agree the answer is "yes, but at a price and after the print." → staged accumulate $165–190.

Tension 3 — Fundamentals vs. Sentiment on AMKR. Fundamentals: record quarter, 26% growth, AI packaging tripling. Sentiment/Valuation: negative FCF for three years, 16.8% peak margin, TSMC as competitor. Resolution: for a low-margin capital-intensive subcontractor, cash generation outweighs revenue growth — growth funded by $3B of capex at 16% gross margin destroys value unless the 2030 targets land exactly. Fundamentals-of-cash wins → AVOID.

Tension 4 — CVLT's beat vs. the 16% drop. Fundamentals says the print was good and guidance was raised. The market says −16%. Resolution: both are right — it was a multiple reset on a stock that ran 66% in six months, not a fundamental break. But the class action and the ARR-metric credibility gap are genuine, unresolved, and attach to the exact metric still going soft. → HOLD, don't add.

7. Verdicts

	Business quality	Priced in?	Fair value	Now	Conviction	Action
KLAC	9.0/10	Mostly	$155–185	$192	[7.5]	🟡 WATCH → stage $165–190 AFTER tonight's print
GLW	7.5/10	Yes	$105–135	$120	[6.5]	🟡 WATCH — buy $95–110
CVLT	7.0/10	Roughly	$115–150	$125	[6.0]	🟡 HOLD (held) — add only <$110
AMKR	4.5/10	No — still rich on cash	$32–45	$46	[4.0]	🔴 AVOID — value trap
SNDK	5.5/10	Unknowable	$250–2,200	$1,098	[3.5]	🔴 AVOID as an investment

Direct answer to the question

Mostly no — and one qualified yes.

Four of these five are not Korea-contagion bargains. They fell on their own earnings, in the last 48 hours, from prices that had roughly tripled in twelve months. A 50% drawdown from a 3× move leaves you at 1.5×, not at a discount. Framework §0 still answers "the market has priced it in" for GLW, AMKR and SNDK.

The one qualified yes is KLAC — the highest-quality business of the five (61% gross margin, 41% operating margin, 31% FCF margin, majority share of process control, −4.1%/yr share count), which has now fallen into the $180–200 accumulate zone this agency set on 2026-07-15, at ~15% cheaper than then. But do not buy it today. It reports FQ4 after the close tonight, and the memory customers being repriced are 15–25% of its revenue. Let the print land, then stage in at $165–190.

CVLT is the one where the market got the facts wrong (wires misreported a raised guide as a cut), but the right answer is still hold, don't add — because the class action and the ARR-mix credibility problem are real and attach to a metric that is still soft.

A portfolio-specific passage was removed from the public build.

Sources

Prior related work: compare-ASML-KLAC-LRCX-2026-07-15.md