Screen tech sweep — Financebot

🔴 SUPERSEDED IN PART, same day. The full /analyze on TSM found this screen's forward multiple was wrong: 19.3x came from a vendor forward EPS of ~$21.60 against an actual 2026 consensus of $15.91 → the true multiple is 26.2x. TSM's verdict therefore moves ACCUMULATE → WATCH, fair value $432–518 → $350–480, entry ≤$400 → ≤$360. See analyze-2026-08-04.md and Knowledge/Playbook/pitfall-vendor-forward-eps-is-the-wrong-fiscal-year.md. The relative call in §3 (TSM decisively over UMC) is unaffected — UMC's multiple came from the same field and is overstated in the same direction, so the gap between them is wider, not narrower. GDDY was separately confirmed by full /analyze at conviction 6.5.

Date: 2026-08-04 · Analyst: Financebot (Fundamentals + Moat + Valuation + Sentiment)

The list is not one theme. It contains three mature-node foundries, one leading-edge foundry, a tower REIT, an auto supplier mid-spinoff, an RF chip company losing its main customer, a domain registrar, and a superconductor micro-cap. They are graded on their own terms below.

1. Verdict Table

Ticker	Price	Fwd P/E	Verdict	Conviction	Fair Value	One-line reason
GDDY	$89.09	8.1x	ACCUMULATE	6.5	$112–157	8x free cash flow, 4.5%/yr share count shrink, AI fear is aimed at the wrong segment
TSM ⚠️	$417.17	~~19.3x~~ 26.2x	~~ACCUMULATE~~ WATCH	~~7.0~~ 6.5	~~$432–518~~ $350–480	🔴 Revised by full `/analyze` same day — the 19.3x was a wrong-fiscal-year vendor figure. Best business here; now at fair value, not below it
CCI	$77.66	26.7x	HOLD / WATCH	5.5	$70–88	5.5% yield, real, but 92% payout means the dividend cannot grow until 2028
APTV	$47.72	7.1x	WATCH	5.0	unreadable	7x forward is real, but every per-share number on file is pre-spinoff and wrong
UMC	$20.63	23.5x	AVOID	3.0	$12–16	Costs more than TSMC on forward earnings while it shrinks
SWKS	$66.79	13.4x	AVOID	2.5	—	Earnings −69%, payout 147%. The 4.6% yield is bait
GFS	$52.02	20.7x	AVOID	3.5	—	Same mature-node fight as UMC, earnings −53%, nothing UMC does not have
TOELY	$188.11	—	SKIP — bad data	—	—	49x at the equipment-cycle peak, and the ADR feed is broken
AMSC	$33.45	23.8x	AVOID — accounting trap	2.0	—	The "P/E of 11" is a deferred-tax release, not profit

2. The Three Traps in This List

Read this section before the analysis. Three of these nine names look cheap only because of a data artifact.

2.1 AMSC — the "P/E 11" is not earnings

metric	value
Operating margin	5%
Net margin	45%
P/E (trailing)	10.97
P/E (forward)	23.81
EV/EBITDA	53.56
Beta	3.28

A net margin nine times the operating margin is not a business result. It is a release of a deferred tax asset valuation allowance — a one-time non-cash credit that flows through net income and destroys the P/E for four quarters. The knowledge base already carries this trap as Knowledge/Playbook/pitfall-tax-valuation-allowance-round-trip-breaks-eps.md.

The honest numbers are the forward P/E of 23.8 and EV/EBITDA of 53.6. That is a full-priced growth micro-cap with a beta of 3.28, already down 53% from its 52-week high. Do not screen this on trailing P/E.

2.2 APTV — every per-share figure on file is pre-spinoff

Aptiv completed the separation of its Electrical Distribution Systems business on 1 April 2026. That business now trades as Versigent (VGNT). Aptiv also received a $1.9B cash dividend from the transaction.

figure	what the data feed says	what is actually true
Revenue	$20.40B (FY2025)	~$13.2B — Q2 2026 was $3.3B annualised
FCF/share	$6.93	Overstated. Includes EDS cash flow
Revenue/share	$92.40	Overstated by roughly 35%
Book value/share	$43.48	Includes $4.6B of pre-spin goodwill

Tangible book is closer to $21.70/share ($9.21B equity less $4.6B goodwill, over 212M shares) — so the stock trades at 2.2x tangible book, not the 1.10x price-to-book the screen shows. The forward P/E of 7.06 is the only clean number in the file, and it is a single analyst estimate.

2.3 TOELY — the ADR feed is broken

Enterprise value of −$238.56B, price-to-sales of 0.07, EV/EBITDA of −0.31. These are impossible for a profitable company. The ADR ratio is being applied to some fields and not others. Nothing here can be valued without going to the Tokyo listing directly.

Even taking the clean figures at face value — 49x trailing earnings, 13.5x book — this is a wafer-fab-equipment name priced at the top of a capex cycle that TSMC has just guided to $60–64B. That is the good news already in the price.

A fourth, milder version of the same problem affects TSM: the reported price-to-book of 87 is wrong. It divides the USD ADR price by a book value per ordinary share. Real book value is roughly $32.80 per ADR (NT$5.36T equity ÷ 5.19B ADRs ÷ ~31.5 TWD/USD), so the true multiple is about 12.7x book. High, but not absurd. Graham's intrinsic value for TSM recalculates to about $93, not the $35 shown — and is still useless for a 40%-ROE business.

3. UMC vs TSMC — the comparison you asked for

Both are Taiwanese foundries carrying identical geopolitical tail risk. That makes the comparison unusually clean: the risk cancels, and only the business is left.

metric	TSM	UMC	winner
Revenue CAGR (3yr)	+18.9%	−5.2%	TSM
Net income CAGR (3yr)	+19.6%	−23.3%	TSM
FCF CAGR (3yr)	+24.0%	−7.9%	TSM
Gross margin	64%	31%	TSM
Operating margin	60%	22%	TSM
Return on equity	40%	21%	TSM
Free cash flow (FY25)	NT$992B	NT$49B	TSM
Capex trend	NT$1.28T, rising	NT$50.7B, cut 45%	—
Forward P/E	19.3x	23.5x	TSM
Analyst consensus	strong buy	underperform	TSM
Mean target vs spot	$540 (+29%)	$18.49 (−10%)	TSM

The single fact that settles it: UMC is more expensive than TSMC on forward earnings while its revenue, earnings and free cash flow all shrink.

Two further details make it worse:

UMC's forward P/E (23.5x) is higher than its trailing P/E (20.2x). That is consensus saying earnings per share will fall from here. Compare TSM, where forward (19.3x) is far below trailing (35.8x) — earnings rising.
UMC's net margin (33%) exceeds its operating margin (22%). As with AMSC, profit is arriving from below the operating line — non-operating income, most likely FX and investment gains. That is not repeatable.

The UMC bull case, stated fairly: the mature-node downturn has found a floor. Utilisation is heading above 90% in Q3 2026, UMC has formally notified customers of price increases for 2H 2026 after demanding 15% cost cuts from its own suppliers, and there is genuine optionality in the Intel 12nm partnership (design kits due by year-end) and silicon photonics in production.

Why it still fails: the stock has already run from $6.56 to $20.63 — up 214%. The recovery is fully priced and then some. Graham intrinsic value is $11.19 against a $20.63 price. You are paying a premium to TSMC's multiple for a commodity business whose main competitors (SMIC, Hua Hong) are running at 96–109% utilisation and will add capacity into any price strength.

Verdict: TSM wins decisively. UMC is an AVOID at $20.63. Revisit UMC below $13 — near Graham value, and where the price hikes would be free.

The same reasoning kills GFS ($52.02, forward 20.7x, earnings −53%). It is UMC's fight with a weaker balance sheet position and a 16% short interest. If you want mature-node exposure, this list does not contain a good way to buy it.

4. TSM — ACCUMULATE, conviction 7.0

The business. TSMC produces roughly all of the world's leading-edge logic. Q2 2026 revenue was $40.2B at a 67.7% gross margin, and management raised full-year 2026 revenue growth guidance to above 40%. Capex went up to $60–64B, with a further $100B committed to Arizona (total US programme now $265B).

Health. Faultless. Revenue +18.9%, net income +19.6%, FCF +24.0% (3yr CAGRs). Debt/assets 13.4%. Cash NT$2.77T against NT$1.06T of debt. Share count dead flat for four years — no dilution, and dividends have grown from NT$285B to NT$467B.

Moat. As close to structural as exists in hardware. At N2 the capital cost of a competing fab is beyond any single private balance sheet, and the customer list (Apple, Nvidia, AMD, Qualcomm, Broadcom) has nowhere else to go at volume. The genuine attack vectors are Intel 18A/14A and Samsung Foundry — both real, neither at parity today. TSMC has also announced price increases of up to 10% starting 2027, which is what pricing power looks like when it is stated out loud.

Valuation.

model	result	note
Forward earnings	Forward EPS ≈ $21.60 at 19.3x	The anchor
Fair value at 20–24x forward	$432–518	Spot sits just below the low end
Bogle expected return	0.93% yield + ~20% earnings growth, flat multiple	~+20%/yr
Bogle, multiple compressing to 15x	—	~+10%/yr
Graham intrinsic value	~$93 (corrected)	Discard. Useless for a 40% ROE business
Analyst mean target	$540	+29%

The one risk. Taiwan. That is the entire discount, and it is not diversifiable — it is a binary that would take the position to near zero. The margin of safety is the business quality, not the price.

Position sizing note: this belongs in the 🌏 Structural-Risk sleeve, not 🏛 Evergreen Compounders. It is a top-quality business inside a single-point-of-failure jurisdiction. Size it as the risk, not the quality.

Entry: ≤ $400
Trim: 26x forward (not a fixed dollar — re-set each analysis)
Break: Intel 14A wins a tier-1 external customer at volume · N2 gross margin dilution exceeds the guided 3–4 points · any hard change in the Taiwan Strait posture

5. GDDY — ACCUMULATE, conviction 6.5

The setup. Down 44% from its 52-week high, at $89.09. The market has decided AI will kill the website builder.

Health — and it is very good.

metric	value
Free cash flow (FY25)	$1.58B
FCF/share	$11.20 (diluted) / $12.47 (current share count)
Price / free cash flow	7.1–8.0x
FCF CAGR (3yr)	+19.7%
Net income CAGR (3yr)	+35.4%
Gross margin	64%
Operating margin	27% (from 12.6% in 2022)
Share count CAGR (3yr)	−4.5%/yr
Capex	$23.9M on $4.95B of revenue

That last line is the business in one number. GoDaddy spends 0.5% of revenue on capital expenditure and converts 32% of revenue to free cash flow. It has bought back $1.60B of stock in one year against an $11.28B market cap — a 14% annual repurchase rate.

Two apparent red flags that are not:

Current ratio 0.63 — driven by deferred revenue from prepaid multi-year domain registrations. Customers pay first. This is a working-capital advantage dressed as a liquidity problem.
Debt/equity 57,334 — an artifact. Equity is $215M only because buybacks have consumed the book. Debt is $3.86B against $1.58B of annual FCF and EV/EBITDA of 10.1. That is roughly 2.4x FCF. Not distress.

The bear case, taken seriously. It is real and it is the reason for the price. Applications & Commerce bookings decelerated to +7% year-over-year from +9% in Q1. The Airo transition put a ~100 basis point drag on total bookings. Guidance merely matched consensus. William Blair downgraded to Market Perform citing intensifying AI-disruption concern. Revenue growth is only 6.6% CAGR.

The rebuttal. The AI threat is aimed at website building — the A&C segment. It is not aimed at the domain registry, which is where the moat sits: GoDaddy is the registrar of record at roughly 85% renewal rates, and an AI-generated website still needs a domain name. If anything, more sites means more domains. What is genuinely at risk is the upsell attached to the domain, and that is a margin story, not an existence story.

Valuation.

model	result
At 11–14x FCF	$123–157
At 10x FCF (permanent AI discount priced in)	$112
Owner return, zero growth	12.6% FCF yield + 4.5% share shrink ≈ 17%/yr
Analyst mean target	$104.80

Note the bear case. Even assuming the market is permanently right about AI and GoDaddy never re-rates above 10x free cash flow, fair value is $112 — 26% above spot. The buyback alone compounds free cash flow per share at ~5% a year with no growth in the business at all.

This is the best risk/reward in the list on pure numbers.

Entry: ≤ $95 — in zone now
Trim: 14x FCF
Break: A&C bookings turn negative · renewal rate falls below 82% · buyback pauses to service debt

6. CCI — HOLD / WATCH, conviction 5.5

What changed. On 1 May 2026 Crown Castle closed the $8.4B sale of its small cell and fiber businesses and repaid $7.2B of debt ($5B floating rate, $1.75B unsecured notes). It is now the only publicly traded pure-play US tower operator. It also repurchased $1B of stock at $88.66, retiring over 11M shares and cutting the annual dividend obligation by $47M.

The dividend arithmetic — this is the whole decision.

item	value
2026 AFFO guidance (midpoint, raised $5M)	$1,975M
Shares outstanding (post-buyback)	~426M
AFFO per share	~$4.64
Dividend per share (annualised, $1.0625/qtr)	$4.25
AFFO payout ratio	~92%
Price / AFFO	16.7x
Dividend yield	5.47%

The dividend is covered — ignore the 173% GAAP payout ratio, which is meaningless for a REIT with heavy depreciation. But 92% of AFFO leaves almost nothing for growth. Crown Castle cannot meaningfully raise this dividend until AFFO grows, and AFFO cannot grow while Sprint/T-Mobile decommissioning churn runs off — 2026 site rental revenue is guided down 5% even though organic growth was +3.9% in Q2. The churn ends around 2028.

The models disagree, and the disagreement is informative.

model	fair value	assumption
Dividend Yield Theory	$86.40	$4.25 ÷ 4.92% five-year average yield
Dividend Discount Model	$66.70	2% growth, 8.5% discount rate
Analyst mean target	$95.47	—

Dividend Yield Theory assumes the yield reverts to its historical average. That average was earned by a growing dividend. It is not applicable while growth is structurally zero. The DDM is the more honest model here, and the truth is between them.

Weighted fair value: $70–88, midpoint ~$79. Spot at $77.66 is fairly valued. Not cheap.

The real risk is the balance sheet, not the towers. Debt/assets is 93.8% and book equity is negative $1.64B. That is tolerable for a tower REIT with contracted, escalating, investment-grade-tenant revenue — but it leaves no room for an interest rate surprise, and interest expense is already $941M against $2.09B of operating income.

Verdict. A bond proxy paying 5.5% with a flat dividend for two years, then a plausible 4–5% growth rate once churn clears. If you want that, it is honest. It is not a growth-and-income compounder, and buying it here gives you no discount for waiting.

Entry: ≤ $70 (pushes the yield above 6% and gives a real margin of safety)
Trim: 19x AFFO
Break: AFFO payout exceeds 100% · a further dividend cut · churn extends past 2028

7. APTV — WATCH, conviction 5.0

Why it is not a buy today: the numbers are unreadable. See §2.2. The spinoff completed 1 April 2026, and every historical per-share figure on file still includes the Electrical Distribution Systems business that has since left.

What is knowable:

Q2 2026 revenue $3.3B, +2% year-over-year — an earnings beat, but guidance for both Q3 and the full year came in below consensus, and the stock is at its 52-week low of $46.41.
New Aptiv targets 4–7% revenue CAGR 2025–2028, with non-automotive at 8–10% and software and services at mid-teens. That is the re-rating thesis: a commodity harness supplier turning into a software and sensing business.
Forward P/E 7.06, EV/EBITDA 6.00. Cheap if the estimate is right.
Share count down 6.6%/yr (3yr CAGR) — 283M to 212M, with $4.1B repurchased in 2024 alone. Management is buying its own stock hard.
Net debt roughly $9.0B against a $10.1B market cap. That is close to 1:1, and it is the reason the equity moves violently on small changes in the auto cycle.

The judgment. A supplier guiding 4–7% growth at 7x forward earnings with a 6.6% annual share count reduction is worth owning if the forward estimate survives contact with a clean quarter. It has not yet had one. Buying now means buying an estimate against financials that describe a company that no longer exists.

Wait for the Q3 2026 print — the first fully clean post-spin quarter. Then recompute FCF/share and the real leverage on the standalone entity. If free cash flow per share lands anywhere near $4 on a ~$13B revenue base, this is a buy in the low $40s.

Entry: hold fire until Q3 2026 reports
Break (pre-emptive): net leverage rises above 3x EBITDA · the software and services mid-teens CAGR target is walked back

8. Rejected — the short reasons

SWKS ($66.79). Earnings −69% year-over-year, revenue −3%, operating margin down to 9% from a business that used to run in the high 20s. The 4.63% dividend yield sits on a 147% payout ratio — it is being paid out of the balance sheet, not out of earnings. Short interest is 35% of float, the highest in this list by a wide margin, and consensus is hold with a mean target of $68.25 — effectively spot. This is Apple content loss playing out in public. The yield is the bait on the trap.

GFS ($52.02). Earnings −53%, revenue +3%, forward P/E 20.7x, beta 1.80, short interest 16%. It is fighting the identical mature-node battle as UMC against the same Chinese capacity, with no advantage UMC lacks. If mature-node is the theme you want, neither of these is the vehicle at these prices.

TOELY ($188.11). Broken ADR data (see §2.3), and on the figures that do work, 49x trailing earnings and 13.5x book at the peak of the equipment capex cycle. TSMC's raised $60–64B capex guide is exactly what is already in this price. Buy the customer, not the supplier, at this point in the cycle.

AMSC ($33.45). Accounting trap (see §2.1). Real forward P/E 23.8, EV/EBITDA 53.6, beta 3.28.

9. What to Do

GDDY at $89 — the clearest opportunity here. 8x free cash flow, 14% annual buyback, and the bear case still values it 26% higher. Start a position; add below $80 if AI fear pushes it there.
TSM at $417 — accumulate below $400, sized as a structural-risk position rather than a core compounder. Quality is not in doubt; jurisdiction is.
CCI — watch at $70, not $78. The dividend is safe but frozen. There is no reason to pay fair value for two years of no growth.
APTV — diarise the Q3 2026 print. Nothing to decide before then.
UMC, GFS, SWKS, TOELY, AMSC — pass. Two are value traps, two are cycle-peak, one is an accounting artifact.

Portfolio fit note. Four of the nine names are semiconductor foundry or equipment exposure that correlates tightly with the AI capex cycle. If the existing portfolio already carries AI-infrastructure weight, TSM should be treated as an addition to that cluster, not as diversification away from it.

Sources: .mcp/fin.py (Yahoo Finance) for all financial statement data · Q2 2026 earnings coverage and guidance per company reports as cited inline.