Theme Leaders: A PEG/TAM Valuation Pass — Cheap, Fair, or Rich?

A companion pass to Semis, Quantum, Rare Earth, Nuclear, Space, Robotics: Broken Thesis or Just Oversold? — that piece asked which of 95 beaten-down names still have an intact thesis. This one asks a different question about the ~56 theme leaders across twelve themes (Data Storage added Aug 11, 2026): at today's price, is the leader Cheap, Fair, or Rich relative to its own growth?

Method

Adjusted-PEG methodology (short version): Score = Forward P/E (or, for loss-making names, Price/Sales) divided by (trailing growth rate x TAM-runway factor x quality factor), following the TAM-Adj-PEG framework. Runway and quality factors are the same bottleneck/moat judgments used to pick these names as theme leaders in the first place — a 15-year monopoly (ASML's EUV lock, TSM's foundry share) earns a bigger multiple than a 5-6 year cyclical franchise (Teradyne, Cognex). Score <0.5 = Very Cheap, 0.5-0.8 = Cheap, 0.8-1.2 = Fair, 1.2-1.8 = Fair-Rich, 1.8-2.5 = Rich, >2.5 = Very Rich. Growth input is trailing revenue or EPS growth (FMP-sourced), not a forward consensus CAGR — treat this as a screening pass, not a full multi-year model. Loss-making names (marked P/S) get a Price/Sales-vs-growth read instead of PEG, and pre-revenue/data-distorted names are marked Option-like rather than forced into a score.

Two data-quality bugs were flagged going into this pass (same family as the HDB/TGS/BE issues already on file): TSM's on-file forward P/E read 0.84 instead of ~25x (a TWD/USD currency mismatch), and RCAT's on-file revenue growth read -59% against an actual Q1 +849% YoY print. Both are corrected below. Checking those two led to a broader spot-check of the price_to_sales field for every loss-making name in this set (cross-referenced against external sources) — 6 of the 15 names checked were off by 30% or more, worst case Symbotic at 44x understated (1.8x on file vs. ~78x actual, which flips its verdict from Very Cheap to Very Rich). All corrected figures and the resulting verdict changes are flagged inline. This looks like a systemic issue with how price_to_sales is computed/stored for smaller or newly-scaling names, not a one-off — worth a proper data audit outside of this pass. Each table also carries the closing price used for this pass (Price, July 23, 2026) alongside a Live column that re-queries the current price and day change every time this page loads — so the valuation case above is frozen to July 23, but the price itself won't be.

Semiconductors

TSM: DB's on-file forward_pe (0.84) is a currency/units bug — net income is booked in TWD against a USD market cap. Corrected here via cross-check (GuruFocus): TTM PE ~27.5x, forward PE ~25.3x.

AVGO: Trailing EPS growth (+287%) reflects an easy prior-year comp (VMware integration costs a year ago), not a repeatable growth rate — verdict uses revenue growth (+24%) instead, which reads Fair.

Optical Interconnect

Quantum Computing

IONQ: Headline trailing PE (46.7x) is flattered by a positive TTM net income figure that is very likely a non-operating mark (warrant/investment fair-value gain), not core profitability — IonQ is still operationally loss-making per its own guidance. Read the 'Very Cheap' PEG score here with real skepticism; this is still an option-like, pre-profit name.

IBM: Trailing EPS growth (+74%) looks like one-time-item noise for a company that normally grows earnings mid-single-digits — verdict uses revenue growth (+8%) instead, which reads Rich. IBM's dividend/FCF ballast and quantum optionality aren't captured by a pure PEG screen, so treat 'Rich' here as a caution flag, not a sell signal.

RGTI: Revenue is tripling off a tiny base (+199% YoY per Q1) but the on-file YoY comp reads -34% against last year's one-time system-sale comp. Too early-stage for a multiple; hardware-roadmap/CHIPS-LOI optionality only.

Space

ASTS: DB's on-file price_to_sales read 180x; real P/S (cross-checked) is closer to 377x. Stays Very Cheap on this framework only because trailing revenue growth (+1,505% YoY) is treated at face value — but that figure is a near-zero-base artifact (2025 revenue went from $4.4M to $70.9M), the same distortion flagged for UEC below. Read this one with real skepticism; the honest label is option-like, not a clean PEG win.

RDW: DB's on-file price_to_sales read 3.0x; real P/S (cross-checked) is ~4.2x. Corrected verdict moves from Very Cheap to Cheap.

PL: DB's on-file price_to_sales read 20.3x; external sources range 24-45x (use 30x). Corrected verdict moves from Fair to Fair-Rich.

LUNR: On-file revenue_growth_yoy (-8%) undercounts the story — FY26 guide is $900M-1B on the Lanteris acquisition plus new NASA contracts (South Pole $180M, Nova-C $148M). PSG isn't usable off the stale trailing number.

Rare Earth & Critical Minerals

MP: DB's on-file price_to_sales read 21.9x; real P/S (cross-checked) is ~31x. Score moves from 0.34 to 0.48 — stays Very Cheap, but closer to the Cheap line than the raw DB figure suggested.

USAR: revenue_growth_yoy reads 0.0% — Stillwater Phase 1a just commissioned, ramping toward 600 MTPA by Q4; this is a pre-revenue-to-revenue transition the trailing field can't capture yet.

UUUU: Revenue -16% YoY is real, not a data artifact — flagged Broken in the correction article. No PEG case here; still working through inventory/pricing softness.

Nuclear/Uranium

CCJ: Trailing EPS growth (+238%) is a low-base snapback off last year's depressed uranium pricing, not durable — verdict here uses revenue growth (+11%) instead, which reads Rich. That said, CCJ's real driver is multi-year contract book re-pricing at higher uranium prices as legacy low-price contracts roll off — a plain PEG screen understates that structural tailwind.

GEV: Trailing EPS growth (+218%) is a turnaround snapback off a depressed prior-year base in a lumpy power-equipment order book — verdict uses revenue growth (+9%) instead, which reads Rich.

LEU: Screens Very Rich on trailing numbers (60x forward PE against 1.5% revenue growth), but the fresh $900M DOE HALEU contract signed 7/1/26 (total $1.07B w/ options) isn't in the trailing print yet — the correction article flagged this exact gap. Revisit once the new contract starts flowing through revenue.

UEC: revenue_growth_yoy reads +29,738% — a near-zero prior-year base artifact (Burke Hollow mine just came online), not a usable growth rate. Treat as option-like on production ramp, not a PEG name yet.

Power/AI Data Center

VST: revenue_growth_yoy reads -12%, which the correction article already traced to a fuel-cost/hedge accounting pass-through quirk — actual Q1 revenue was +43% YoY. No usable PEG/PSG off the stale field.

Robotics

SYM: DB's on-file price_to_sales read 1.8x — the single worst data error found in this pass. Real P/S (cross-checked externally) is ~78x. Corrected, this flips the verdict from Very Cheap to Very Rich: back-to-back profitable quarters and a $22.7B backlog are real, but the stock is not cheap on that basis.

Defense Drones

AVAV: DB's on-file price_to_sales read 2.2x; real P/S (cross-checked) is ~3.7x. Stays Very Cheap after correction, but the multiple itself was materially understated.

RCAT: On-file revenue_growth_yoy reads -59% — this is the same stale/mismatched TTM comp the correction article already flagged (Bottom line #5). Actual Q1 2026 revenue was +849% YoY (Army SRR program win replacing Skydio) and industry trackers show Q2 tracking +742% YoY ahead of the Aug 6 print. TTM PE/PSG isn't usable here — real growth is clearly positive, just not mechanically computable from this field.

Security Software

CRWD: DB's on-file price_to_sales read 9.3x; real P/S (cross-checked) is ~21x. Corrected verdict moves from Very Cheap to Cheap — still favorable, materially less so.

Cloud Software

Data Storage

Added Aug 11, 2026 — later than the July 23 batch above, so prices/multiples here are pulled as of this date, not the 7/23 close used everywhere else. Same Adjusted-PEG methodology. Nvidia's Jensen Huang called enterprise/AI data storage "completely unserved" at CES 2026, and the tape has agreed: WDC is +202% YTD and SNDK +469% YTD on a hyperscaler-driven HDD/NAND shortage that's now running into 2027 forward-pricing contracts. This theme splits cleanly into two camps that don't compete with each other so much as sit on top of one another: legacy media/hardware incumbents (WDC, STX, SNDK, NTAP, CVLT) who make or have long made the physical disks, flash, and backup infrastructure now capacity-constrained, versus new software-defined/cloud-native players (P, NTNX, RBRK, BLZE) who sell the data-management and resilience layer built on top of whatever media sits underneath — several of them explicitly monetizing the shift away from spinning disk and away from legacy incumbents like Broadcom/VMware and Dell EMC.

Legacy Storage

WDC/STX: Trailing EPS growth (WDC +303%, STX +333%) is operating-leverage snapback off the prior cyclical trough — HDD pricing was still in a multi-year downturn as recently as 2023 — not a repeatable growth rate; verdict for both uses revenue growth instead. Worth flagging on its own: WDC's on-file trailing P/E (16.0x) sits far below its forward P/E (48.5x), meaning the Street is pricing today's enterprise-nearline shortage as a peak, not a new baseline. The same commodity-cycle risk that hit these names in 2011 (Thailand floods spike, then crash) and 2018-2020 (oversupply bust) applies in reverse once hyperscaler capex normalizes.

SNDK: on-file trailing P/E (16.0x) is identical, digit for digit, to WDC's — almost certainly a stale/duplicated field carried over from the Feb 2025 spinoff rather than an independent SanDisk print, the same kind of data-quality issue that caught the TSM currency bug and Symbotic's price_to_sales error elsewhere in this piece. Trailing revenue growth (+10%) also straddles the spinoff boundary and understates the current NAND-shortage quarter — SNDK is +469% YTD on the same AI-storage demand shock driving WDC/STX, but neither field here is clean enough for a real PEG read yet; marked Option-like rather than forced into a score.

NTAP: the 5% growth figure is real, not a data artifact — NetApp is the slowest-growing, most mature name in this set, re-rated on AI-storage enthusiasm spilling over from faster-growing peers rather than its own numbers.

New / Cloud-Native Storage

P: formerly Pure Storage — renamed Everpure, Inc. in February 2026, now trading under ticker P rather than PSTG. Trailing EPS growth (+48%) is a profitability-inflection snapback (on-file trailing P/E reads 144x off a still-thin earnings base) rather than a repeatable growth rate; verdict uses revenue growth (16%) instead, which reads Rich. The Meta hyperscaler qualification and Evergreen//One subscription mix are the real re-rating catalyst — the stock is already pricing a good chunk of that in.

NTNX: trailing EPS growth (+237%) is a GAAP-profitability-inflection snapback, not repeatable; verdict uses revenue growth (18%) instead.

RBRK: newest name in this set (IPO'd April 2024), still unprofitable — forward P/E (567x) is meaningless off a near-breakeven base, so it's scored on Price/Sales instead, same treatment as CRWD/ZS above. Screens Very Cheap on 48% trailing growth, but per the framework's own caution on sub-0.5 scores: verify that growth isn't decelerating before treating this as a clean entry — hyper-growth SaaS multiples compress fast once ARR growth slows into the 20s-30s, and Rubrik is still GAAP-unprofitable with real dilution risk.

Bottom line

1. Split across the full basket: 11 Very Cheap, 4 Cheap, 9 Fair, 7 Fair-Rich, 2 Rich, 15 Very Rich, 8 Option-like/pre-revenue. The correction article's core claim — that the drawdown was mostly valuation, not fundamentals — still reads true for the biggest names: the two hardest-hit, best-documented bottleneck monopolies (NVDA, TSM) screen as the cheapest large names on this whole list, not the most expensive.
2. The two structural bottlenecks (NVDA, TSM) are the cheapest large-caps in the set — both under 0.25 on the adjusted score despite trading at $2-5T market caps, because trailing growth (50-65%+) is still outrunning a mid-20s forward P/E. AVGO screens Fair once the VMware-comp snapback in its trailing EPS growth is stripped out.
3. Software and cyclical-industrial multiples are priced well ahead of trailing growth almost across the board — PANW, FTNT, OKTA, DDOG, KTOS, TER, CGNX, ROK, FN, QBTS, and (after the price_to_sales correction below) SYM all screen Very Rich on trailing numbers. That doesn't mean broken (PANW just closed the $25B CyberArk deal, DDOG's observability moat is real, Symbotic just posted back-to-back profitable quarters) — it means the market is already paying for acceleration that hasn't shown up in the trailing print yet, so the risk is multiple compression on any guide that merely meets rather than beats.
4. A price_to_sales data-quality issue changed real conclusions, not just cosmetics — Symbotic's on-file P/S (1.8x) was 44x understated versus its real ~78x, which flipped the verdict from Very Cheap to Very Rich; CrowdStrike and Redwire both moved from Very Cheap to Cheap on the same kind of correction. This was caught by spot-checking every loss-making name externally rather than trusting the field as-is — a systemic check worth running before any future valuation pass off this table, not just a one-time fix.
5. Pre-revenue/early-stage names cluster into two honest buckets, not one: names where the trailing growth field is simply the wrong tool (RCAT, LUNR, UEC, USAR, RGTI — all have real, well-documented growth stories a TTM comparison can't see) versus names that are Rich even on a corrected Price/Sales read (QBTS, SYM) or Option-like because the data is genuinely too thin either way. Don't average these into one theme-level verdict. ASTS is a special case: it screens Very Cheap only because its +1,505% trailing revenue growth is taken at face value, but that figure is a near-zero-base artifact (2025 revenue went from $4.4M to $70.9M) — treat it as option-like, not a clean PEG win, despite the number.
6. Nuclear/Uranium is the most split single theme — CCJ, GEV, and LEU all screen Rich-to-Very-Rich on trailing growth despite real catalysts (LEU's fresh $900M DOE contract, CCJ's contract-book re-pricing) that the trailing-growth field hasn't caught up to yet. This is a case where the PEG screen is arguably behind the news, not ahead of it.
7. Cross-reference with the correction article before acting on any single 'Very Rich' flag — a name screening expensive here (e.g., KTOS, FTNT) was independently confirmed Thesis Intact there on guidance/contract grounds. Rich means 'priced for a lot to keep going right,' not 'broken.' Cheap names with a genuinely broken thesis don't exist in this set by construction — these are theme leaders, selected for quality first.
8. Data Storage splits along the same fault line as the rest of the basket, with a twist: the AI-driven HDD/NAND shortage (WDC +202% YTD, SNDK +469% YTD) is real and structural, yet WDC screens the only outright Fair name in the group — its trailing P/E (16x) sitting far below its forward P/E (48x) shows the market already treating today's earnings as a cyclical peak, not a new floor. The cloud-native/software layer sitting on top of that hardware (P, NTNX) is pricing in more of the good news already (Rich, Fair-Rich) than the physical media itself, while the theme's cheapest name (RBRK, Very Cheap) and its richest (NTAP, Very Rich despite being the slowest grower in the set) make the same point from opposite directions — growth rate matters more than sector narrative once the framework adjusts for it.