Basis Report/Resources/Investor Foundations
4 sections20 entries

Semiconductor stock valuation: how to underwrite the cycle without getting caught at the peak

The multiple should come after you understand where the cycle sits. Before you anchor on 20x forward earnings or 8x EV/EBITDA, you need a clear view on inventory days, end-market demand mix, and what normalized earnings look like outside of peak cycle.

✓Build a through-cycle revenue and earnings model before anchoring on any multiple — peak-cycle inputs produce meaningless outputs.
✓Know the end-market mix: AI datacenter, PC, mobile, auto, industrial, and where each sits in its own inventory cycle.
✓For IDMs, model capex as a percentage of revenue across the cycle — the capital intensity is part of the valuation story.
✓Compare valuations on normalized EPS, not current-quarter EPS — the spread between the two tells you where the cycle sits.
Go to frameworkSee a sample report
When to use this

Use this before initiating or sizing any fabless, IDM, or semiconductor equipment name. The framework is most useful when the stock looks either optically cheap on trailing earnings or optically expensive on depressed estimates — both of which happen routinely across the semiconductor cycle.

Why it matters now

The generative AI wave created an inventory super-cycle in AI accelerators while legacy end markets — PC, mobile, industrial — sat in prolonged correction. Investors who built models on blended growth rates missed the bifurcation. Understanding exactly where each revenue dollar comes from and which cycle that market is in is more important in this environment than at any point in the past decade.

Where theses break

The playbook breaks when investors use peak-cycle EBITDA as the normalization anchor, mistake AI accelerator demand strength for a broader semiconductor upcycle, and apply fabless software-like multiples to IDMs with capital-intensive balance sheets that require sustained capex to maintain competitive process nodes.

Full framework

4 sections · 20 entries — work through each before you size a position.

Most semiconductor valuation mistakes are made at the top of the inventory upcycle — investors pay peak multiples on peak earnings and call the combination 'a reasonable entry.' The real work is normalizing through-cycle earnings power, understanding what separates a structurally higher-margin story from a temporary mix benefit, and stress-testing the model against a 30% revenue decline before it becomes the obvious answer.

20 entries in view

Set the cycle-adjusted valuation lens before the model opens

The most consequential valuation decision in semiconductors is whether the earnings you are using reflect mid-cycle economics or a cyclical distortion. Getting that wrong before the model starts makes every number that follows directionally wrong.

Identify the cycle position before selecting a valuation metric

Semiconductor earnings can swing 50–80% between peak and trough across the inventory cycle. Using current-quarter earnings as the valuation anchor without understanding where the cycle sits is the same mistake as valuing a mining company on peak-cycle commodity prices. Before choosing a multiple or building a DCF, establish the cycle position in each major end market the business serves — PC, mobile, industrial, auto, datacenter — and determine which markets are at peak, mid-cycle, or trough. The valuation anchor should be normalized earnings, not the most recent print.

Why it matters

Every significant semiconductor drawdown in the past two decades — 2001, 2008, 2015, 2019, 2022 — was partially caused by investors paying peak-cycle multiples on peak-cycle earnings and then experiencing simultaneous earnings derisks and multiple compression when the cycle turned.

When it matters

Before initiating any position and whenever a stock re-rates sharply on a single strong quarterly report without a clear structural change in the underlying business.

Investor take

Label the cycle position explicitly in any valuation work. State: 'This company is in an upcycle driven by [specific end market], and I am using normalized earnings rather than current earnings as the valuation anchor.' If you cannot write that sentence clearly, the valuation is not ready.

Separate structural growth from cycle-driven revenue by end-market category

An AI accelerator business growing 60% revenue is not in the same cycle as an industrial microcontroller business in the same company declining 20%. Mixed-end-market companies like Broadcom, Texas Instruments, and Qualcomm have multiple simultaneous cycles running at different phases. Blending the growth rates without understanding the end-market composition produces a model that is accurate at neither the upside nor the downside. Build revenue by end-market category and assign a cycle position to each before summing to the enterprise total.

Why it matters

Companies that benefit from AI-related demand have used that tailwind to offset significant weakness in other end markets without breaking out the math clearly. Investors who did not build the end-market decomposition missed the underlying weakness and overpaid for what looked like broad-based strength.

When it matters

When a semiconductor company reports strong headline revenue growth in a period where several major end markets are clearly in inventory correction. The strong headline is almost always driven by one or two markets masking broad weakness.

Investor take

Build a revenue waterfall by end-market segment for the trailing four quarters. If one segment explains the majority of the growth story and that segment is AI or datacenter, model the scenario where AI demand pauses and ask what the headline revenue growth number looks like without it.

Match the valuation metric to the capital intensity of the business model

A fabless chip designer with 70% gross margins and minimal capex should be valued on normalized P/E or EV/FCF, which accurately reflect the cash economics. An IDM that spends 30–40% of revenue on capex should be valued on EV/EBITDA with explicit capex modeling alongside it — EBITDA without capex context understates the capital burden. A foundry like TSMC should be valued on EV/EBITDA plus price-to-book against ROIC because the asset base is the business. Applying the wrong lens for the business model is the first error in semiconductor valuation; the subsequent analysis is built on the wrong foundation.

Why it matters

The semiconductor sector includes business models with radically different capital intensities. A fabless company and a foundry with identical EBITDA margins have completely different intrinsic values because the foundry's EBITDA is producing a much lower return on the capital invested to generate it.

When it matters

When initiating coverage of any semiconductor name, and whenever a company's business model changes — moving from fabless to IDM through acquisition, or outsourcing more manufacturing to reduce capex intensity.

Investor take

For any semiconductor investment, write three sentences: (1) the primary revenue model, (2) the capex/revenue ratio at mid-cycle, and (3) the resulting FCF margin at normalized revenue. Those three facts determine which valuation multiple is appropriate and what a reasonable intrinsic value range looks like.

Check where the stock has traded relative to its own through-cycle valuation history

Each semiconductor business has a characteristic historical multiple range across the cycle. Understanding that range is more useful than any comparison to sector averages, which blend companies at different cycle phases and with different capital intensities. A business that has historically traded between 12x and 22x normalized earnings offers a very different entry thesis at 14x versus 21x. The historical range tells you what the market has been willing to pay for this specific combination of cycle exposure, gross margin quality, and management credibility.

Why it matters

Cross-sector comparisons are often misleading in semiconductors because the relevant comparable is the company's own through-cycle history, not what a peer in a different end-market is trading at today.

When it matters

Before initiating or adding to any position, and after a large market-wide selloff that has brought the stock to historically low multiples without a company-specific deterioration.

Investor take

Build a 10-year chart of the stock's normalized P/E or EV/EBITDA and mark the cycle troughs and peaks on it. Locate the current multiple and identify whether it represents a genuine cycle low or whether current depressed earnings are still inflated relative to the prior trough.

Assess the AI structural argument separately from the cyclical base case

The thesis for NVDA, AMD, Marvell, and others includes both a cyclical recovery in the base business and a structural step-change from AI accelerator demand. These are different analytical problems. The cyclical base case is a traditional semiconductor valuation question: where is the inventory cycle, what does normalized earnings power look like, what multiple does the business deserve on mid-cycle earnings? The AI structural argument is a different question: how large is the incremental TAM, how defensible is the company's position, and over what timeframe does the earnings inflection materialize? Conflating the two makes the valuation look cleaner than it is.

Why it matters

Investors who do not separate the cyclical and structural components of a semiconductor thesis often overweight the structural case in moments of excitement and then discover the cyclical deterioration at the most inconvenient moment — when AI sentiment has softened and the recovery in the base business has not yet arrived.

When it matters

When evaluating any semiconductor company trading at a material premium to its historical multiple range and the bull case centers on a transformational demand driver — AI, EV content, IoT, 5G.

Investor take

Model two scenarios explicitly: (1) the base business at mid-cycle without AI revenue increments, and (2) the incremental AI earnings contribution at your best estimate of steady-state penetration. The combined value of both is the thesis — but each should be defensible on its own terms.

Normalize earnings through the cycle before assigning a multiple

Reported semiconductor earnings contain cycle-period noise that makes current-quarter results a poor input for any valuation decision. Normalization is the step that separates the business's earning power from the inventory cycle's temporary effects on utilization, pricing, and operating leverage.

Normalize gross margins to mid-cycle utilization before running any earnings model

For IDMs and companies exposed to fab capacity, gross margin is directly tied to utilization. At full utilization, fixed manufacturing costs are spread over maximum revenue, and gross margins can expand 8–12 percentage points above mid-cycle levels. At trough utilization — when customers are burning through excess inventory — gross margins compress by the same amount. A company reporting 65% gross margins at peak utilization may normalize to 53% at mid-cycle. Applying a premium multiple to the peak margin structure produces a valuation that is valid only if utilization never falls — which it always does.

Why it matters

Gross margin normalization is the single most important earnings adjustment in semiconductor valuation because the operating leverage is severe and the cyclical swings are large. Missing it produces earnings estimates that are wrong at both ends of the cycle.

When it matters

Whenever the most recent quarterly gross margin is more than 5 percentage points above or below the company's 5-year median. That gap is usually a signal that utilization or pricing is at an extreme, not that the margin structure has permanently changed.

Investor take

Build a gross margin sensitivity model: estimate the business's normalized gross margin at 75% fab utilization (typical mid-cycle), 90% (upcycle), and 60% (downcycle). Map those three scenarios to three revenue assumptions. That is the earnings envelope before any modeling of operating leverage in R&D or SG&A.

Separate unit volume, pricing, and mix in the revenue model before normalizing earnings

Semiconductor revenue changes are driven by three variables that move at different speeds: unit volume (tied to end-market demand), average selling prices (tied to supply/demand balance and competitive dynamics), and product mix (tied to design-win ramp and customer spending shifts). Conflating these into a single revenue line prevents you from understanding whether a revenue growth period is driven by durable content expansion or temporary pricing power that will erode when supply catches up.

Why it matters

Revenue models that do not decompose volume, price, and mix will not catch the inflection point where volume is growing but ASPs are already declining — which is typically the first signal of an inventory oversupply cycle forming.

When it matters

During periods of strong industry demand when it is easy to assume all three drivers are moving favorably in the same direction. The cycle turns when ASPs soften first, before unit volumes decline.

Investor take

For each major product category, track the three-quarter trend in ASP separately from unit volumes. Falling ASPs on rising units is a warning sign. Rising ASPs on flat units is a pricing tailwind that may not persist. Durable revenue growth requires both units and mix moving favorably.

Model inventory days at the customer as a leading indicator before finalizing forward estimates

Semiconductor customer inventory builds ahead of normal demand, then corrects, driving order patterns that amplify the underlying end-market cycle. When customer inventory days for a major chip category exceed historical averages — typically 60–90 days for consumer electronics, 90–120 days for industrial — semiconductor companies typically see order weakness 2–3 quarters before the end market itself weakens. Building forward revenue estimates without modeling customer inventory days produces guidance that misses the cycle inflection.

Why it matters

The inventory correction cycle in semiconductors is well-documented and still routinely underestimated because customers are not required to disclose inventory data in real time, and management guidance tends to lag the deterioration until it is unavoidable to acknowledge.

When it matters

Every quarter before finalizing forward estimates, and during periods when distributor and OEM inventory levels are reported above normal by trade publications, channel checks, or management commentary.

Investor take

Track customer inventory for the top three end-market categories as a separate model input. When inventory days in any major end market crosses 1.3x the historical average, reduce the forward revenue estimate for that segment by 15–20% as a working assumption until the data shows de-stocking progress.

Normalize R&D spending to the mid-cycle investment level before computing margin structure

Semiconductor companies accelerate R&D investment during strong revenue periods and cut it during downturns. This means reported R&D as a percentage of revenue fluctuates with the cycle rather than reflecting a steady-state commitment. A company that reports 18% R&D/revenue at peak cycle and 28% at trough has a mid-cycle level somewhere between — and that mid-cycle level is the appropriate input for any normalized earnings estimate. Using trough-cycle R&D efficiency overstates normalized margins; using peak-cycle intensity understates them.

Why it matters

R&D normalization matters most for fabless companies where R&D is the primary operating cost driver and the absolute investment level must hold above a minimum threshold to maintain competitive design capability regardless of the revenue cycle.

When it matters

When normalizing earnings for any semiconductor company and when comparing operating margin structures across peers at different phases of the investment cycle.

Investor take

Find the company's R&D/revenue ratio for each of the past three years and identify the mid-cycle level. Use that as the normalization anchor. If the current quarter's R&D/revenue is significantly below mid-cycle, the company may be under-investing competitively; if significantly above, margins will improve as revenue recovers.

Account for the design-win-to-revenue lag before forecasting growth from new programs

Semiconductor revenue from new design wins typically materializes 12–24 months after the design win is secured, and the ramp from initial production to peak volume adds another 6–12 months. Investors who build revenue forecasts based on design-win announcements without modeling this lag tend to front-load revenue by 4–6 quarters, producing near-term estimates that are too high and then missing the ramp when program volume actually arrives.

Why it matters

The design-win pipeline is the correct leading indicator for semiconductor revenue potential, but the timing between pipeline announcement and revenue contribution is longer and more variable than most financial models assume.

When it matters

When evaluating any semiconductor company that is winning design awards or announcing new customer programs as a primary argument for valuation premium, and when building forward estimates beyond the next two quarters.

Investor take

For each major design win, note the announced production start date and the expected ramp timeline. Discount the near-term contribution if production start is more than 12 months away. Build a probability-weighted production ramp schedule rather than assuming 100% capture of the stated volume from the first quarter of availability.

Build a DCF that survives the downturn, not just the upcycle

A semiconductor DCF that assumes current tailwinds persist through the terminal period will almost always produce a bullish output. The value of the model is forcing the cycle into the analysis — explicitly modeling a downturn, a recovery, and a normalized long-run state before committing to an intrinsic value range.

Build a three-phase DCF that explicitly includes a downturn phase

A standard two-phase DCF — growth period then terminal value — is particularly misleading for semiconductor businesses because it assumes current growth persists through the first phase without any cycle interruption. The correct model for a semiconductor business has three phases: the current cycle position (upcycle, mid-cycle, or trough), a normalization period where revenue and margins revert toward mid-cycle levels, and then a long-run steady state. The normalization period is where most of the analytical work happens — it forces a specific view on how deep the next correction is, how long it lasts, and what the recovery path looks like.

Why it matters

Semiconductor DCF models built without a normalization phase consistently produce intrinsic values that are too high because they embed cycle-peak margins in the growth phase and then compound them at terminal growth. The terminal value captures most of the value in a high-multiple business, and if it is anchored to peak margins, the DCF is validating optimism rather than testing it.

When it matters

Before any DCF model is finalized for a semiconductor name, and whenever the current revenue growth rate is above 20% in a period that appears cyclically supported rather than structurally permanent.

Investor take

Structure the DCF with three explicit phases: Phase 1 is the current cycle period (1–3 years), Phase 2 is the normalization and recovery period (3–7 years), Phase 3 is the terminal steady state. The terminal growth rate should be applied only to Phase 3 earnings, not to the cyclically elevated Phase 1 run-rate.

Set the terminal value on through-cycle FCF margin, not on current-period margins

Terminal value in a semiconductor DCF typically represents 60–75% of total intrinsic value. If the FCF margin embedded in the terminal value is based on peak-cycle gross margins, the terminal value is overstated by the difference between the peak margin and the normalized margin — which can be 10–15 percentage points for an IDM with significant manufacturing leverage. A 5 percentage point reduction in terminal FCF margin assumptions can reduce intrinsic value by 20–30% in a standard DCF structure. Always label the terminal FCF margin assumption and verify it against the company's 5-year median, not against the most recent two quarters.

Why it matters

Terminal value errors are the most consequential modeling errors in DCF analysis because they dominate the output, they compound over the explicit forecast period, and they are the easiest errors to obscure in a well-formatted model.

When it matters

Every time a semiconductor DCF is built, and particularly after a period of strong reported margins that may not be representative of long-run economics.

Investor take

Explicitly state in any DCF: 'Terminal FCF margin assumption is X%, which corresponds to mid-cycle gross margin of Y% minus normalized R&D intensity of Z% and normalized G&A of W%. This margin has been achieved in [period] and failed to sustain above [higher level] for more than [duration].'

Stress-test the model against a 25–30% revenue decline before sizing the position

Semiconductor revenue declines of 20–35% in major inventory correction cycles are not tail risks — they are historically consistent outcomes that occur roughly every four to six years. An IDM with 70% gross margins at peak utilization and significant fixed costs can see EBIT swing from positive to negative in such a cycle. The stress test should model: a 30% revenue decline over four to six quarters, gross margins compressing by the operating leverage math, and the resulting free cash flow trajectory — including whether the company needs to draw on the balance sheet, reduce capex plans, or access capital markets.

Why it matters

Semiconductor companies that appear conservatively valued at 12x normalized earnings can become deeply distressed investments in a severe downturn cycle if the balance sheet is weak or capex commitments are inflexible. Discovering this after the downturn begins is not diligence — it is damage control.

When it matters

Before setting the initial position size in any semiconductor name, and after any quarter where guidance implies the cycle may be inflecting downward.

Investor take

Run the 30% revenue decline stress test with a gross margin compression equal to the historical trough gross margin from the prior downturn cycle. Calculate the resulting FCF and divide by market cap to get the stress-case FCF yield. If the stress-case FCF is negative and the company carries significant debt or capex obligations, reduce the position size accordingly.

Quantify the combined earnings and multiple compression in the downside scenario

When a semiconductor cycle turns negative, two simultaneous forces compound the stock's decline: earnings estimates fall as revenue and margins disappoint, and the multiple the market assigns to those estimates contracts as confidence in the recovery timeline decreases. A stock trading at 20x normalized earnings that disappoints and guides down does not re-rate to 18x earnings on reduced estimates — it re-rates to 14x on earnings that are themselves 25% lower. The combined effect is a 40–50% stock decline from peak to trough that looks obvious in hindsight but is routinely underestimated in advance.

Why it matters

Every semiconductor investing cycle contains investors who were intellectually aware of the cyclicality but underweighted its impact on total returns because they modeled earnings compression without modeling the simultaneous multiple compression that amplifies it.

When it matters

Before entering any semiconductor position trading above its historical mid-cycle multiple on current-period earnings, and before adding to a position after a sharp run-up driven by earnings estimate revisions rather than multiple re-expansion.

Investor take

Model the downside scenario explicitly as: (lower earnings estimate) multiplied by (historical trough multiple). That product — not a percentage off the current price — is the defensible downside. If the upside scenario from current levels divided by the downside loss is less than 3:1, the position size should be small.

Use the reverse DCF to check whether current earnings growth is already in the price

Back into the earnings growth rate and FCF margin the current price requires, using normalized — not current — FCF as the starting point. If the price requires 15% compound EPS growth for 10 years and the company has achieved that growth in only two of the past ten years, the stock is not cheap — it is priced for consistent execution on the best historical pace the company has demonstrated. A reverse DCF with realistic cycle assumptions, including at least one downturn in the explicit forecast period, usually produces a required growth rate that is sobering relative to what the business has actually demonstrated.

Why it matters

The reverse DCF is the most useful valuation tool in semiconductor investing precisely because it converts the question from 'what is the stock worth?' — which is easy to answer optimistically — to 'what does the stock require?' — which is harder to dismiss when the answer is uncomfortable.

When it matters

Before initiating any position in a semiconductor stock that has re-rated significantly from its prior trough, and whenever an analyst's price target implies a return that appears attractive but requires heroic assumptions about the length and magnitude of the current upcycle.

Investor take

Build the reverse DCF on normalized FCF, not current-period FCF. Specify the discount rate, the terminal growth rate, and the explicit forecast period. Present the required growth rate and margin trajectory and ask: has this company demonstrated the ability to produce these outcomes through a full cycle, including at least one inventory correction?

Use relative valuation as a cycle check, not a price target

Comparable analysis in semiconductors tells you whether a stock is cheap or expensive within a peer group at a given point in the cycle. It cannot tell you whether the entire sector is priced correctly for the cycle position, and it should not be used as a primary valuation anchor when the sector is simultaneously at a cyclical extreme.

Compare peers on normalized earnings multiples, not current-quarter results

The most misleading comparable analysis in semiconductors is built during an inventory upcycle, when all companies in the peer group have elevated earnings simultaneously and multiples look reasonable relative to each other. The correct comparison is normalized multiples — price divided by mid-cycle earnings — across the peer group. This comparison is more informative because it adjusts for each company's cycle position and reveals whether the current relative pricing reflects genuine quality differences or simply the market rewarding the companies in the earliest-cycle recovery.

Why it matters

Peer comparisons during an upcycle systematically overvalue the most cyclically leveraged companies because they appear cheapest on the most recently reported earnings, which are the most cyclically elevated. This creates a value trap that appears internally consistent until the cycle turns.

When it matters

During periods of strong industry performance when the peer group appears broadly attractive, and before initiating any new position in a semiconductor company that looks cheap on a trailing or near-term forward P/E basis.

Investor take

Build the peer comparison on normalized P/E rather than LTM or NTM P/E. Use a mid-cycle revenue and margin assumption for each company, compute the normalized EPS, and compare current prices against normalized earnings. The ranking that emerges is a more honest picture of relative value.

Use gross margin dispersion across the peer group to identify genuine quality premiums

Gross margin is the most important determinant of long-run FCF potential in semiconductors after capital intensity. A company with 65% gross margins at mid-cycle deserves a structurally higher multiple than one with 50% gross margins because the higher-margin business has more room to generate FCF as scale increases and R&D/SG&A become smaller percentages of revenue. Peer comparisons that do not adjust for gross margin dispersion mislead investors into believing that a lower-gross-margin, lower-priced company is more attractive than the fundamental economics warrant.

Why it matters

Gross margin differences across semiconductor companies are often driven by product positioning, IP intensity, and customer mix — and those differences are stable through the cycle. The market frequently underweights this stability and allows gross-margin leaders to trade at smaller premiums than the FCF differential justifies.

When it matters

When comparing two semiconductor companies with similar reported revenue growth rates but different gross margin profiles, and when evaluating whether a premium multiple is justified or represents overextension.

Investor take

Sort the peer group by gross margin and check whether the market's multiple ordering roughly tracks the margin ordering. Outliers — companies at a discount despite superior gross margins — are worth investigating. Those situations often arise when the market is discounting near-term earnings weakness, creating an attractive entry for investors who model the normalized earnings correctly.

Cross-check relative valuations against the distribution of acquisition premiums in the sector

Strategic acquirers in semiconductors — Intel, Qualcomm, Broadcom, Marvell — have demonstrated willingness to pay specific transaction multiples for companies with demonstrated IP, proven design-win pipelines, and strategic market positions. Those acquisition multiples provide a floor that is often more grounded than public market comparables in periods of sector dislocation. If a semiconductor business with strong IP, diversified end-market exposure, and a track record of design-win success trades at or below recent transaction multiples for comparable assets, the public market discount deserves an explanation.

Why it matters

The semiconductor M&A market has been active and consistently produces transaction multiples higher than public market consensus for assets with strategic value to an acquiring platform. Understanding the transaction comp set contextualizes public market valuations in a way that pure equity comparables do not.

When it matters

When a semiconductor stock has declined materially from prior highs and the business quality appears intact, and when evaluating whether the stock has reached a level that might attract strategic interest.

Investor take

Build a transaction comparable table for M&A in the specific semiconductor sub-category over the past three years. Calculate EV/Revenue and EV/EBITDA for each transaction. If the current public market price is below the lowest transaction multiple in the set without a company-specific deterioration, that disconnect is analytically significant.

Track the premium or discount to book value as a cycle sentiment indicator

For capital-intensive semiconductor businesses, price-to-book value is one of the clearest cycle indicators available. At cycle trough — when earnings are depressed and confidence is low — semiconductor stocks typically trade between 1.0x and 2.0x book. At cycle peak — when earnings are at record levels — the same stocks often trade between 4.0x and 6.0x book. The position within the historical range is the signal: a stock at 1.5x book on a normalized ROIC of 20% is a very different situation than the same stock at 5.0x book on a 20% ROIC.

Why it matters

Price-to-book is one of the few valuation metrics that remains meaningful at cycle trough when earnings-based multiples are either negative or artificially high because the earnings base has collapsed. It anchors the valuation to the replacement value of the asset base rather than to a temporarily distorted income statement.

When it matters

When earnings have declined significantly and P/E-based valuations are no longer informative, and when assessing whether a cycle trough has been reached in a capital-intensive semiconductor name.

Investor take

Calculate the company's book value per share and the current price-to-book multiple. Compare both to the historical range across the past two full cycles. If the stock trades below its historical trough P/B on a sustainable ROIC that exceeds WACC, the implied pessimism is worth testing with a recovery scenario.

Do not use sector-wide relative value to justify a position when the entire sector is at a cycle peak

The most dangerous conclusion in semiconductor peer analysis is 'this stock looks cheap relative to sector.' If the entire sector is at or near peak-cycle earnings and peak-cycle multiples, the cheapest stock in the peer group is still expensive relative to mid-cycle economics. Sector-relative cheapness is useful when at least some portion of the peer group is at trough — it helps identify which recovery plays offer the best risk-reward. It is not useful when the entire sector is priced for the continuation of the current upcycle without any cycle correction.

Why it matters

Sector-relative valuation creates a form of collective rationalization: every company looks fairly priced because they all look similar relative to each other, and investors convince themselves that the peer group average represents a reasonable anchor. But if the anchor is set at a cyclical high, the relative cheapness disappears when the cycle normalizes.

When it matters

During periods of strong sector momentum when a rising tide is lifting all multiples, and whenever the most common bull argument for a semiconductor name is that it trades at a discount to a peer group that is itself at a historical premium.

Investor take

Ask the explicit question before using sector-relative valuation as a primary argument: is the sector itself fairly priced on normalized earnings, or is the sector-wide multiple at or near a historical peak? If the answer is the latter, reduce the weight given to peer comparisons and increase the weight given to the absolute intrinsic value estimate based on through-cycle normalized earnings.

Evidence

Semiconductor valuation inputs scorecard

The six valuation inputs that change across the semiconductor cycle

No single metric captures semiconductor value in isolation. Each reads differently at cycle top versus cycle bottom — know which regime you are in before drawing a conclusion.

Normalized P/E
Price / Through-cycle EPS
The most useful lens for fabless companies with stable capital needs. Normalize EPS to mid-cycle revenue and margins before comparing to historical ranges. A 25x normalized P/E on a business with 20% EPS growth through the cycle and strong IP moats is different from 25x on peak earnings that may revert by 40%. Always specify what 'normalized' means in your model — it should be a defensible mid-cycle assumption, not an average of peak and trough.
EV / EBITDA
Enterprise value / EBITDA
More useful for IDMs and equipment companies where D&A is material and capex cycles do not perfectly track revenue. For IDMs, EBITDA must be viewed alongside the capex commitment that sustains the D&A — an IDM at 8x EBITDA with 40% of revenue going to capex is not cheap. For equipment companies, EBITDA against backlog provides a forward-looking sanity check on the current revenue recognition pace.
FCF Yield (Normalized)
Mid-cycle FCF / Market Cap
Use normalized FCF — mid-cycle revenue, mid-cycle margins, normalized capex intensity — not the FCF reported in the most recent quarter. Peak-cycle FCF can be 2–3x trough FCF for IDMs with operating leverage. An apparent 6% FCF yield calculated on peak FCF may be a 2% FCF yield in a normal year. Always label the cycle position when publishing an FCF yield.
Price/Book vs. ROIC
Market cap / Tangible book
For capital-intensive IDMs and foundries, Price/Book mapped against return on invested capital is a useful cross-check. A business generating sustainable 20%+ ROIC deserves to trade above book; one generating 8% ROIC at peak should not command a significant book premium. When P/B contracts while ROIC holds, look for a valuation opportunity. When P/B expands while ROIC is declining, assume the market is pricing optimism that may not materialize.
Book-to-Bill
Orders / Shipments (trailing 3-month)
Not a valuation metric itself, but essential context for any valuation decision. A book-to-bill above 1.0 signals demand exceeding supply — cycle momentum is favorable. Below 1.0 signals customers are working down inventory — be cautious about underwriting current-quarter revenue as sustainable. A book-to-bill sustained below 0.9 for two or more consecutive quarters has historically preceded revenue step-downs across the industry.
Reverse DCF Implied Growth
Solve for g at current price
Back into the through-cycle growth rate the current price requires. For a semiconductor business, that growth rate must be decomposed: volume growth in existing end markets, ASP stability or erosion, and content growth from new applications. If the implied growth requires content growth from markets still in sample qualification where adoption is 3–5 years away, the stock is pricing optimism that the operating calendar does not yet support.

Business model to valuation lens

The semiconductor business model should dictate the valuation method, not the sector average

Applying a fabless software-like multiple to a capital-intensive IDM is the most common semiconductor valuation error. Map the lens to the economics.

The semiconductor business model should dictate the valuation method, not the sector average
Business modelPrimary valuation lensQuality gates requiredCommon mistake
Fabless (NVDA, AMD, QCOM)Normalized P/E or EV/FCF, calibrated to IP quality and design-win cycleGross margin above 50%, design-win pipeline replacing current revenue, no single customer above 20%Paying peak earnings multiples at peak cycle on the argument that AI exposure makes the current earnings level structural — without underwriting what happens to non-AI revenue in a correction.
IDM (Intel, TXN, STMicro)EV/EBITDA or normalized P/E, plus capex-adjusted FCF yield as a sanity checkFab utilization above 75%, capex declining as a percent of revenue, ROIC above WACCIgnoring the capital intensity embedded in the EBITDA. An IDM at 7x EBITDA with 35% capex/revenue and a fab modernization program in process is not as cheap as the multiple implies.
Semiconductor equipment (AMAT, LRCX, KLAC)EV/EBITDA or P/E, cross-checked against backlog coverage and deferred revenueBacklog above 1.0x trailing 12-month revenue, customer capex plans visible for next 18 monthsBuilding a valuation on current-quarter revenue when backlog dynamics suggest a pull-forward — equipment revenue can peak 6–12 months before the end-market correction becomes visible in shipments.
Memory (Micron equivalent)Price/Book against ROIC and trough-to-peak cycle analysisTrough normalized earnings confirmed, capacity discipline visible in industry capex plans, DRAM/NAND spot pricing stable or inflectingBuying memory stocks on P/E at trough because the multiple looks high — trough earnings are negative or near zero, so P/E is not useful. Price-to-book against the ROIC inflection is the correct entry signal.
Analog and mixed-signal (TXN, ON, ADI)EV/FCF yield or normalized P/E with long-cycle content analysisIndustrial and automotive end-market inventory levels normalizing, long design-win pipeline maintaining diversityUsing short-cycle consumer electronics inventory cycles as a proxy for analog end markets, which are dominated by industrial and automotive programs with 5–7 year design-in lifetimes.

The peak earnings trap

Paying a 'reasonable' multiple on peak earnings is not cheap — it is cycle-timing risk disguised as value

When a semiconductor company reports record earnings in a demand upcycle and the stock trades at 18x forward P/E, the multiple looks rational. What the multiple does not capture is that the forward estimate is itself at a cyclical peak. If end-market demand normalizes — customer inventory destocking, pricing pressure from new wafer supply, or a consumer spending slowdown — the earnings that support the 18x multiple may fall 30–50%. The result is not a 20% stock decline; it is a 40–60% decline, because both the earnings and the multiple contract simultaneously. The correct entry price for peak-cycle earnings is a discount to the historically appropriate mid-cycle multiple on mid-cycle earnings — not a comparison to the current multiple on current earnings.

Common questions

What investors ask about investor foundations for investor foundations stocks.

How do you value a semiconductor stock at the bottom of the inventory cycle?
At cycle trough, reported earnings are depressed and multiples look stretched on a trailing basis — but that is precisely when the valuation work matters most. The right anchor is normalized earnings power: what does the income statement look like in a mid-cycle environment where utilization is normal, pricing is stable, and customers are ordering at sustainable rates rather than working down inflated inventory? For most large-cap semis, that means modeling the revenue run-rate that corresponds to end-market unit volumes at trend, not at cycle-peak or cycle-trough. Once you have a normalized EPS estimate, check where the stock trades relative to its historical mid-cycle multiple. If the stock is at a material discount to that range on normalized earnings, you have the beginning of a valuation argument. If it is at a premium even on normalized earnings, the recovery is already in the price.
Should you use EV/EBITDA or P/E to value semiconductor stocks?
It depends on the business model. For fabless companies with high gross margins and minimal capex, P/E or EV/FCF is often the cleaner lens because earnings are a good proxy for cash generation. For IDMs that spend 30–40% of revenue on capex — Intel, Texas Instruments — EV/EBITDA is more useful because the D&A load is real and meaningful, and EBITDA better captures the economics before the capital intensity decision. For semiconductor equipment companies, EV/EBITDA adjusted for backlog and deferred revenue is standard because revenue recognition can lag orders significantly. Whatever method you use, always sanity-check it against a through-cycle FCF yield — that grounds the relative valuation in actual cash economics rather than accounting constructs.
What is the right way to model semiconductor gross margins across the cycle?
Gross margins in semiconductors are highly sensitive to fab utilization for IDMs, and to pricing power and product mix for fabless companies. For IDMs, fixed manufacturing costs are spread over variable revenue, so a 10% decline in revenue can produce a 15–20 point decline in gross margin — the operating leverage is severe. Model gross margins at three utilization levels: peak (full capacity), mid-cycle (normalized demand), and trough (a major inventory correction). For fabless companies, gross margin is more stable but not immune — customers push back on pricing at cycle troughs, and mix shifts toward lower-margin legacy nodes can compress blended margins even without a unit volume decline. The goal is to understand the margin floor and the margin ceiling before committing to any earnings estimate.
How do you tell the difference between a structural margin expansion and a peak-cycle margin?
Structural margin expansion shows up when margins hold or improve even when revenue growth decelerates, because the business has permanently shifted its product mix toward higher-value content, improved yields, or repriced older products on IP-driven pricing power. Peak-cycle margins appear when utilization is high, pricing is tight, and customers are competing for supply — and then retreat materially when the cycle turns. The test is to look at gross margin through the last full cycle: if margins were 10 percentage points lower at trough versus peak, that is cycle-driven variance, not structural improvement. If the trough in the current cycle is 5 percentage points higher than the trough in the prior cycle on similar revenue declines, that is evidence of structural progress. Never extrapolate the current peak margin without asking whether the last trough already told you what the floor looks like.

Continue research

Keep building the foundation

Semiconductor stock analysis: reading the cycle before repricingInvestor FoundationsTech multiples: what actually deserves a premiumInvestor FoundationsHow to build a DCF that holds under pressureInvestor FoundationsHow to Value a Stock Without Fooling YourselfInvestor FoundationsReturn on invested capital: the incremental-returns testInvestor FoundationsEV/EBITDA: how to use the multiple without fooling yourselfInvestor Foundations

Put it on a live stock

NVDA stock overviewJump to the live ticker page and report flowAMD stock overviewJump to the live ticker page and report flowQCOM stock overviewJump to the live ticker page and report flowAVGO stock overviewJump to the live ticker page and report flow

Category hub

Valuation frameworks hubSee all guides in this category