Texas Instruments runs a 80,000-SKU analog and embedded chip catalog with a 36.7% 10-year average ROIC and a 30-year religion of growing free cash flow per share. The current six-year capex super-cycle has crushed near-term FCF, leaving the stock priced near base-case IV ($157) with real upside if 300mm utilization normalizes.

Plain English

Texas Instruments makes 80,000 different small chips that put electricity, sound, and signals where engineers need them. Every car, factory machine, toaster and smartphone needs them. They sell parts that stay in catalog for 15 years, in their own factories on big 300mm wafers that are cheaper than competitors'. The factories cost a lot to build, so right now profits look smaller than usual. Once the new factories fill up, profits should be higher per share than before. The job: buy when the price is well below their long-run worth, and wait.

Thesis

Texas Instruments designs and manufactures analog and embedded processing semiconductors. Analog (79% of 2025 revenue at $14.0B) takes real-world signals (temperature, sound, pressure, current) and conditions or converts them; embedded processing (15%, $2.7B) is the small digital brain that runs microcontrollers in cars, appliances, factory equipment and infrastructure. The catalog is roughly 80,000 distinct part numbers sold across more than 100,000 customers, with no single end product driving the franchise. Average product life is 10 to 20 years; gross margins ride 60-65% through-cycle.

Why it might compound: TXN earns a 10-year average ROIC of 36.73%, generates owner earnings of ~$5.5B TTM (period ending 2026-03-31), converts roughly 68% of net income to free cash flow over the trailing five years (depressed by the capex super-cycle), runs net debt to EBITDA of just 1.26x, and has shrunk shares outstanding modestly (-1.16% over ten years; the heavy years were earlier). Management measures itself in free cash flow per share growth, full stop. The four sources of edge — internal 300mm wafer manufacturing, a broad catalog, an enormous direct sales channel and the longevity of analog SKUs — compound rather than decay.

Why now: TXN is mid-cycle in industrial and auto demand and late in a deliberate six-year elevated capex cycle (Sherman SM1/SM2, Lehi, Richardson RFAB2). Reported FCF and current ROIIC look weak; the scorer flags ROIIC as not meaningful and notes maintenance capex carries a >50% spread, which is why the IV range is wide.

Valuation math: scorer IV range is $89 / $157 / $237. With current price unspecified in the input, I anchor recommendation thresholds to that range. Margin of safety becomes meaningful at roughly 0.83x base IV, or ~$130. The bull-case IV of $237 already prices a clean 300mm cost-down cycle and a full industrial recovery; above ~$215 (about 0.91x bull IV) the asymmetry inverts. Position is Buy on weakness, scale in.

Moat

Five-moat scan, with Damodaran's framework as scaffolding [1][3][4].

1. Cost advantage (the load-bearing pillar). Analog manufacturing on internal 300mm wafers is roughly 40% cheaper per packaged die than 200mm because a 300mm wafer yields ~2.3x more die area. TXN owns the largest internal 300mm analog footprint in the world: Richardson RFAB1, RFAB2, Lehi (acquired from Micron), and the new Sherman SM1/SM2 campus. Smaller analog peers (ADI, ON, NXP, Microchip, Infineon, STM) are predominantly fab-lite or 200mm-heavy. This is the textbook scale-driven cost advantage Damodaran describes — bigger firms reduce unit cost in capital-intensive process industries [4]. Stress test: if a $10B competitor tried to replicate over five years, the capital is feasible, but the 30-year process learning curve, the depreciation schedule already absorbed by TXN, and the product-mix amortization across 80,000 SKUs are not. Erosion risk: if the industry shifts decisively to fab-light/foundry-fabricated analog at TSMC's specialty nodes, this advantage compresses. Today, TSMC analog capacity is constrained and pricier; the moat holds.

2. Switching costs (real, narrow per design-in but powerful in aggregate). Each TI part is designed into a customer board; redesign costs the customer engineering hours, requalification, and often automotive AEC-Q100 or industrial functional-safety recertification. Damodaran's Microsoft analogy applies in spirit: switching is rarely catastrophic per chip but creates enormous cumulative inertia across an installed base of millions of boards [1][4]. Embedded processing strengthens this: customers write their own software on top of TI MCUs and processors, and that software is sticky across product generations (filing language confirms this). Stress test: a $10B challenger over five years can bid below price on individual sockets, but it cannot make the customer redo a decade of board-level qualification on tens of thousands of designs simultaneously. Erosion risk: greenfield Chinese analog competitors winning new sockets at lower price tiers in PE and consumer; this is happening at the low end and is a real bear point.

3. Intangibles / catalog scale. TI's 80,000-product catalog plus its direct field sales force and ti.com self-serve channel is itself an intangible asset. A small customer wanting one part for a prototype gets it next-day with reference designs, EVMs, simulation models and application notes. No competitor matches this distribution density. Brand here is not consumer brand; it is engineer-trust brand, accumulated over 90+ years (since 1930). Damodaran warns that brand value is the consequence of behavior, not the cause of returns [3]; TI's behavior — keeping parts in catalog for 15-20 years, supporting legacy designs — is what builds the trust.

4. Network effects. Weak/none directly. Indirect: more designs in catalog -> more reference designs -> more engineers default to TI -> more sockets won -> more revenue to fund more catalog. Call this a flywheel rather than a true network effect.

5. Pricing power. Limited at the part level; TI generally does not price-gouge. Pricing power is expressed through mix shift to higher-value catalog parts and through gross margin durability across cycles (62-65% through-cycle). Buffett-style pricing power (raise prices without losing volume) exists in narrow pockets like aerospace and medical sockets but is not the franchise's defining trait.

Competitor stress test summary: ADI is the closest peer, strong in signal chain and high-performance analog after the LTC and MAXIM deals, but lacks TI's 300mm depth and broad-line catalog density. Chinese entrants (3Peak, SG Micro, Will Semi) are real but operate primarily at the low end and in the consumer/PE markets that are 21% of TI revenue. The industrial and automotive 66% of revenue is structurally harder to dislodge because of qualification cycles.

Moat verdict: WIDE.

Management

TXN's capital allocation is the stuff of HBS case studies, but the past five years require honest re-grading.

Reinvestment. From 2016 to 2025 TI allocated $109B of capital, with roughly $24B going to capex during a deliberate six-year elevated investment cycle in 300mm capacity (RFAB2 Richardson, Lehi acquired from Micron, Sherman SM1 online and SM2/SM3/SM4 staged). The strategic logic is impeccable: build owned 300mm capacity at trough valuations to widen the structural cost moat for the next 30 years and reduce geopolitical risk via U.S. soil. The execution risk is real: capex peaked in the wrong year of the cycle, depreciation is now climbing fast, and the ROIIC over the past five years is not meaningful (the scorer flags this explicitly). Management chose capacity over near-term per-share earnings; this is the right Buffett-Munger choice if and only if the demand returns. Grade on this dimension alone: B+ for strategy, C+ for timing.

Acquisitions. TXN's record is famously disciplined and rare: National Semiconductor (2011, ~$6.5B) is the standout, integrated cleanly into the catalog and amortized into superior returns within five years. The Lehi 300mm fab purchase from Micron (2021, ~$900M for a working fab plus future tools) is one of the savviest semis deals of the decade. Buffett would approve: small number of large, careful, value-creating decisions rather than serial M&A.

Debt. Net debt to EBITDA is 1.26x — investment grade, prudent, with long-dated maturities at attractive coupons. Interest coverage is not in the input, but TI's bonds trade tight; the balance sheet is not the problem.

Buybacks. This is where the scorecard tells a more nuanced story than the legend. Share count change over ten years is only -1.16%; the heavy buyback decade was 2004-2014 (when TI bought back 40%+ of shares at low single-digit P/E ratios). The past decade saw repurchases moderate and shift more toward dividends. Buying at average prices that often exceeded conservative IV during 2018-2021 — when the stock traded at 25-30x earnings — is a real capital-allocation flaw if measured against Buffett's discipline. To management's credit, they explicitly state the goal is "accretive capture of future free cash flow for long-term investors," which is the right language. Whether they actually waited for prices below IV is a fair criticism for the 2018-2021 vintage. Grade on buyback discipline: B-.

Dividends. TI has raised the dividend for 21 consecutive years, with payout growing roughly in line with FCF/share over a full cycle. The current cycle has stretched the payout ratio to uncomfortable levels (north of 70% of trailing FCF) precisely because capex consumed FCF. Management has been explicit that it will not cut the dividend; this is a quiet implicit promise that owner earnings of $5.5B normalizes higher. Grade: A on consistency, B on absolute level given the capex squeeze.

Communication. TI's quarterly capital management call and annual capital management presentation are unusually candid for a public semiconductor company. They publish a written capital allocation framework, name FCF/share as the metric that matters, and accept poor cyclical quarters with public equanimity rather than excuses. The 10-K language echoes Buffett ("act like owners who will own the company for decades").

The weakness: the elevated capex cycle has eaten the past five years of per-share growth. A bear can fairly say management bet large on a structural demand thesis that is, so far, only partially confirmed by orders. A bull can fairly say the time to build counter-cyclical 300mm capacity is precisely when peers are pulling back.

Capital allocator: B+.

Industry

Porter's Five Forces on the analog/embedded semiconductor industry as TI competes in it.

1. Rivalry among existing competitors: Moderate-to-high but unusually civilized. The named peer set — Analog Devices, Infineon, NXP, STMicroelectronics, ON Semiconductor, Microchip, Renesas, Maxim (now ADI), and a long tail of Chinese entrants — competes hard on individual sockets but rarely engages in destructive industry-wide price wars. Why: products are differentiated at the part-number level; switching costs and qualification cycles slow share migration; and the largest players (TI, ADI, Infineon) prioritize through-cycle gross margins over share. The exception is at the low end of consumer and personal electronics, where Chinese entrants have compressed pricing meaningfully over the past three years.

2. Threat of new entrants: Low at the high end, real and growing at the low end. Greenfield 300mm analog capacity costs $20-30B per fab and takes five-to-seven years to ramp; this gates serious entry. China is funding entrants aggressively, but they currently win primarily on consumer and commodity power management, not on industrial or automotive sockets that demand AEC-Q100 qualification, 15-year supply commitments, and functional-safety pedigrees. The barrier is not technology alone — it is decades of catalog accumulation and customer relationships.

3. Bargaining power of suppliers: Low. TI is largely vertically integrated. It owns its wafer fabs, runs its own assembly and test (the analog tail does not require leading-edge nodes), buys silicon wafers from a competitive supplier base (Shin-Etsu, SUMCO, Siltronic) and tools from ASML, Applied Materials, KLA and Lam (oligopolistic but spread across enough vendors that no single one squeezes margin). The biggest supplier-side risk is geopolitical (semiconductor export controls, China retaliation, Taiwan), not commercial.

4. Bargaining power of buyers: Moderate. Customers are fragmented — 100,000+ end customers across industrial and automotive — and no single customer drives more than a low single-digit percentage of revenue. Apple-style concentrated customer pressure does not exist here. The exception: very large auto OEMs and Tier-1s during shortages cut hard pricing deals; and during gluts, distributors push back aggressively on pricing. Net: buyers have less leverage than in most tech subsectors.

5. Threat of substitutes: Low to medium. Analog functions are not easily substituted in the physical world (you still need amplifiers, ADCs/DACs, power management). Substitution risk is intra-category: integrated SoCs from foundry-fabricated competitors absorbing functions, or software-defined radios replacing discrete signal chains in some niches. The bigger structural shift is embedded processing migrating toward higher-performance edge AI chips, where TI is competitive but not the leader.

Value pool location and trajectory: industrial (33% of TI revenue) and automotive (33%) — together two-thirds of the franchise — sit in the sweet spot of long-cycle demand: electrification of vehicles, factory automation, energy infrastructure, semiconductor content per car rising from ~$300 in 2010 to ~$1,500 today. Personal electronics and communications are commodity-like and will continue to compress. Data center (9%) is small but growing as analog content per AI server (power delivery, cooling control, signal conditioning) accelerates.

Industry Verdict: Good. Not Excellent because the low-end Chinese pressure and the cyclical capex obligation mean the industry rewards scale operators rather than offering universally attractive economics; not Average because the structural demand drivers (electrification, automation) and the high entry barriers protect long-term ROIC for incumbents.

Inversion

Mandatory inversion. Playing the short-seller in good faith, no hedging.

The single event that kills this. A multi-year industrial recession (think 2001-style, not 2008-style) overlaps with full ramp of Sherman SM1, SM2 and the Lehi conversion. Utilization sits at 30-40% on a depreciation base that has roughly doubled. Gross margin collapses from 62% to 45%. Free cash flow goes negative for two consecutive years while the dividend (currently roughly $5B annual) becomes uncovered. Management is forced to either cut the dividend (breaking a 21-year streak and inviting forced selling by income funds) or take on $10-15B of incremental debt at a much higher cost of capital. The stock breaks below the IV-low of $89, into the $60s. The narrative shifts from "counter-cyclical compounder" to "value trap with stranded capacity."

Why the moat is narrower than bulls think. The 300mm cost moat is real but is being matched, not eroded — ADI is also building 300mm; Infineon's Villach 300mm power fab is operational; ST and Bosch are committing to silicon carbide and gallium nitride at 300mm. The bull narrative implicitly assumes TI is alone in its capacity build. It is not. Worse, the Chinese government has explicitly funded ~$50B for analog capacity buildout; those fabs do not need to earn TI's 36% ROIC, they need to absorb domestic demand. China is currently 20%+ of analog end-demand. As Chinese substitution takes 5-10 percentage points of share at the low end of TI's 21% personal-electronics exposure, that is a 1-2 percentage point hit to total revenue and a far larger hit to incremental margin. Switching costs at the part-number level are real but exaggerated: a competent applications engineer can pin-swap a generic LDO in an afternoon. Switching costs are an asset class, not a fortress.

Why management is worse than it appears. The buyback record, examined honestly, is mixed. TXN bought back roughly $30B of stock from 2018 to 2022 at average prices in the $130-180 range, when conservative IV was likely $100-140. That is repurchasing above intrinsic value. The 21-year dividend streak is now a constraint rather than a virtue: management cannot cut without breaking the implicit promise, which means in a downturn the company will defend the dividend by drawing on the balance sheet. Worse, the 2021 decision to commit to the $30B+ Sherman build assumed a structural growth rate (electrification, factory automation) that is real but slower than the capex schedule needs. Management bet that demand would arrive on a four-year clock; if it arrives on an eight-year clock, the per-share economics suffer materially. Rich Templeton was the architect of the FCF/share religion; Haviv Ilan inherited it but inherits it mid-storm.

What bulls are extrapolating that won't hold. (a) The 36.7% 10-year average ROIC includes 2017-2021 peak years and roughly half a decade of pre-capex-cycle returns; forward 10-year ROIC could easily be 22-28% rather than 36% given the doubled invested capital base. (b) The 68% FCF conversion over five years is depressed by capex but the bull case wants you to mentally annualize it back to 90%; if industrial demand is structurally lower-growth, the conversion may settle at 75-80%, not 90%. (c) Owner earnings of $5.5B TTM are a depressed cyclical number, but the bull's normalized $7.5-8B owner earnings number assumes an industrial recovery that may take longer than three years. Working from $5.5B at a 17x multiple (reasonable for a wide-moat compounder) gets to roughly $94B equity value — well below the current likely market cap.

Valuation trap / multiple compression. TXN has historically traded 22-26x earnings during bull phases. If the market re-rates analog as more cyclical and more capex-intensive than it priced in 2018-2021, a 15-17x multiple becomes the new normal. That is roughly 30% multiple compression on top of any earnings disappointment. Combine 25% lower normalized earnings with 30% multiple compression and you have a 50% downside scenario that requires no business catastrophe — only a recalibration of expectations.

If I am right, the stock could be worth $75-90 within 24-36 months.

Lollapalooza Bias Check

Biases active in me right now as I analyze TXN.

Authority bias. TI's reputation as a capital allocation paragon goes back to Rich Templeton's 2004 letter outlining the FCF/share framework. Buffett-Munger investors revere the company, and Bill Nygren / Don Yacktman / Aswath Damodaran have all praised it publicly. I am inclined to give management more benefit of the doubt than I would for a less-storied company. Counter: the past five years of capital allocation deserve fresh scrutiny, not deference to legacy.

Anchoring. The IV range of $89 / $157 / $237 is given by the scorer. I find myself reasoning relative to the $157 base — "is the stock more than 15% below 157?" — rather than independently asking what the franchise is worth. Anchoring on a model output without re-deriving the assumptions is precisely the trap Damodaran warns against in the canon excerpts on simulations and probabilistic risk [4]. Counter: I should remember the scorer also flags "Maintenance capex uncertain (>50% spread)" and "NOPAT declined; ROIIC not meaningful." The IV bands are a range for a reason.

Recency. Headlines about industrial weakness, China subsidies, and the auto cycle have been negative for 18 months. I am at risk of letting the most recent quarter's pain dominate my normalized view. The 10-year ROIC of 36.73% reflects a longer reality. Counter: also be suspicious of the opposite — using a 10-year average that includes a peak as the new baseline.

Confirmation bias. I genuinely admire TI's culture and want to find reasons to be constructive. I notice I am more rigorous in finding bull evidence than bear evidence. The mandatory inversion section above is partly a structural defense against this bias — by forcing myself to write the strongest credible bear case, I expose what I would otherwise gloss over.

Commitment bias / consistency. If I conclude TXN is too hard or unattractive after years of being a Buffett-Munger darling, I am implicitly admitting prior community consensus was wrong. The social cost of saying "this is fine but not right now" is lower than saying "this isn't what people think it is." I notice myself drifting toward the former even when the analysis might warrant the latter.

Deprival super-reaction. The stock has been rangebound for several years. There is a low-grade fear of missing the move when the cycle turns and the multiple re-rates. This bias pushes me toward "buy now" when the analytically correct answer is "buy below $130."

Incentive (theirs, not mine). TI management is paid in part on FCF/share targets. During the capex cycle, those targets are formally adjusted, but the cultural pressure to defend the dividend and avoid cutting the buyback to zero remains. This is an incentive to financial-engineer through the cycle — drawing on debt or trimming opex — rather than letting per-share metrics breathe naturally. I should weight reported FCF/share with this awareness.

10-Year Outlook

Will TXN be the same fundamental business in 10 years? Almost certainly yes. Analog and embedded processing have been the same business model — broad catalog of long-life parts sold into industrial, automotive, communications and consumer electronics — for 40 years. Every electronic system needs power management, signal conditioning and a small digital brain. The form factors change; the function does not.

Will the customer base be larger? Yes, with high confidence. Industrial automation, vehicle electrification, energy infrastructure (grid modernization, solar/storage), and edge AI all increase analog and microcontroller content per system. Semiconductor content per car has risen from roughly $300 in 2010 to $1,500 today and is plausibly $2,500 by 2035 as ADAS, electrification and software-defined vehicles mature. Even with cyclicality, the secular line slopes up.

Will profit per customer be higher? Probably modestly, with two opposing forces. Tailwind: mix shift toward higher-value automotive and industrial sockets (ASPs higher, life cycles longer, qualification barriers higher). Headwind: Chinese low-end competition pressuring blended ASPs in personal electronics and commodity power. Net likely small positive in nominal dollars; gross margin per part roughly stable; FCF per dollar of capex meaningfully higher once the current cycle is fully absorbed (2027-2030).

Will the moat be wider? Marginal yes. The 300mm cost advantage will fully crystallize as Sherman ramps, deepening the cost moat by another 5-10 percentage points relative to 200mm peers. The catalog accumulates rather than depletes. Geopolitical tailwinds (CHIPS Act, supply-chain re-shoring) explicitly favor U.S.-headquartered, U.S.-fabricated analog players, and TI is the largest of those.

Single biggest threat? Not Chinese competition (slow, mostly low-end). Not technology obsolescence (analog is structurally durable). The biggest threat is capital misallocation — that the elevated capex cycle ends with permanent over-capacity if the demand recovery is shallower than the build-out assumed. This is a finite, observable risk, monitorable through quarterly utilization disclosures.

The second biggest threat is geopolitical: a Taiwan disruption hits the entire industry, and while TI is more insulated than fabless competitors, no global semis franchise is immune.

My 10-year confidence in the fundamental business is high; my 10-year confidence in the next-three-year per-share earnings trajectory is medium.

CONFIDENCE: high

Position Guidance

• Recommendation: Buy on weakness
• Conviction: Medium
• Target buy price: $130 (about 0.83x base IV of $157.28; ~46% above scorer IV-low of $88.98 to leave room for cyclical drawdown)
• Target trim price: $215 (about 91% of bull-case IV of $236.29; above this the asymmetry inverts)
• Position sizing: 2-4% starter position at first entry below $135. Add to 5-7% only if (a) shares trade below $115 with no dividend cut, or (b) first signs of 300mm utilization recovery (gross margin expansion of 200+ bps year-over-year). Cap at 8% even on conviction; this is a single-name semiconductor cyclical, not a Coca-Cola.
• Disqualifiers: trim if dividend is cut or if management abandons FCF/share framework; sell entirely if a major capital-allocation deviation (large premium-priced acquisition, debt-funded buyback above $200) occurs.
• Patience requirement: willing to hold three to five years through cycle without cycle-trough mark-to-market reaction.