STMicroelectronics

IThe business

STMicroelectronics sells semiconductors to about 200,000 customers.⁵ Most of what it sells is not the headline kind. It makes analog chips that manage power, microcontrollers that run a small job inside a car or a thermostat, sensors that sense motion, and power transistors that switch large currents. It designs them and builds most of them in its own plants, mainly at Crolles and Rousset in France and Agrate and Catania in Italy, with other sites in Singapore, Malaysia, Malta, Morocco and China.⁵ That makes it what the industry calls an integrated device manufacturer. The plant is a big part of the story below, because a plant is a cost that does not shrink when sales do.

The company is Franco-Italian in origin and Swiss in management. It is registered in the Netherlands, run from Geneva, and listed in New York, Paris and Milan, with the Paris listing dating from its 1994 flotation.⁵ The New York line is the ordinary share itself, not an ADR, so one NYSE share is one ordinary share. Paris and New York agree: the Euronext close of €41.88 converts at the ECB rate to $46.79, against the NYSE close of $46.74.¹¹⁰¹¹ The French and Italian states own 27.5% through a holding company, a fact that shapes the board and appears in the Management section.⁵ The company reports in US dollars under US accounting rules. Many of its costs are in euros while its prices are in dollars, a point I return to in Risks.

In 2025 revenue was $11.8B, down 11%, after falling 23% in 2024. Operating income was $175M. Then the cycle turned. Revenue in the first, second and third quarters of 2026 was $3.10B, $3.49B and $3.76B, and the company guides to $4.15B for the fourth, which would make about $14.5B for the year, 22.8% more than 2025.²¹²

What it sells, and where it makes money

The company reports four segments. Their profits are very uneven, and that is the single most useful fact about it today.

Three segments earn their keep. Embedded Processing and RF and Optical made $412M between them at margins of 23% and 26%, and Analog and Sensors made $219M.² The fourth, Power and Discrete, lost $81M in the quarter and has lost money in each of the last five: $84M, $99M and $81M in this year’s three quarters and $124M and $67M in the two before.² Its revenue has fallen from $3.1B in 2023 to $1.7B in 2025.⁵ This is the silicon carbide business for electric cars and industry, and the company is building more of it, at Catania and in a joint venture with Sanan Optoelectronics in Chongqing, which started production in the second quarter of 2026.⁷ Much of the unallocated loss of $183M in the third quarter is also tied to the plant program.

Embedded Processing is the engine now. Its revenue rose 24.1%, helped by general-purpose microcontrollers, and its margin reached 23.1%.² RF and Optical rose 46.2%. The company links that to AI data centers, where it sells optical interconnect parts, and to satellites in low Earth orbit.² In June it raised its data center revenue ambition to about $1B in 2026, up from “nicely above $500M”, and said revenue could double in 2027 if the current dynamic continues.⁶ I treat that as management’s hope, not a forecast, and I give it a weight in the bull case below.

Who buys it

Apple was 17.7% of 2025 revenue, up from 14.5% in 2024 and 12.3% in 2023.⁵ The rest is spread across car makers and their suppliers (the 20-F names Bosch, Continental, Denso, Mobileye, Tesla and Vitesco), industrial firms, Samsung, HP and SpaceX.⁵ About 28% of 2025 revenue went through distributors, and in the third quarter of 2026 that rose to 32%, with distribution sales up 38.8% from a year before.⁵² That gives me pause. Distributors buy ahead of demand when times look good, and the company itself says inventory in distribution is still below its target.² So orders are real but they include a restocking effect I cannot size.

In February 2026 the company gave Amazon a warrant for up to 24.8M ordinary shares at $28.38, exercisable over seven years and vesting over five as Amazon reaches product revenue milestones.⁷⁵ The accounting treats the warrant as a price cut: the grant-date value of $331M is taken off revenue over time, $6M so far.⁷ I read it as a sign that a large buyer has been won at a cost. At the current price, if all of it vested, the warrant would be worth about $18 a share on 24.8M shares, roughly $0.45B, and I count it as dilution in my base and bull values C (calculated).⁷¹ The terms of the milestones are not in the filings I read.

The plant is the business

ST has 49,157 employees, of whom 30,546 work in manufacturing.⁵ Depreciation was $1.85B in 2025, 15.7% of revenue, and net plant spending took 15.6% of revenue that year and 22.0% on average in 2022 to 2024 C (calculated).⁵ Since 2025 it has run a program to move wafer capacity to 300mm silicon at Agrate and Crolles and to 200mm silicon carbide at Catania and Singapore, and to cut costs by a high triple-digit million dollar amount by the end of 2027.⁵ It cost $376M in charges in 2025 and $184M in the first nine months of 2026.⁵² It also spent $936M on NXP’s sensor business, bought in February 2026.⁷ Plant spending for 2026 is planned at $2.0B to $2.2B net of grants; nine months have taken $1.41B.¹²² The third-quarter release did not restate the plan, so the fourth quarter is unknown to me.

Why the cycle turned

Chip makers in this part of the market were hit twice. First, car and industrial customers filled their shelves in 2022 and 2023, then spent two years working the stock down. Revenue fell 23% in 2024 and 11% in 2025, and the company took on unused-capacity charges of $88M in the fourth quarter of 2025.⁵² Second, the company’s costs kept rising, because it was building plant for the demand it had expected. Both ended in 2026. Unused-capacity charges were $23M in the third quarter, down from $102M a year earlier.² The company says demand is up on strong bookings and improving visibility, and I have no data to argue otherwise.² I do note that the Street has not followed the stock up. Four analysts on Nasdaq’s feed expect $1.27 of earnings per share for 2026 and $2.77 for 2027, on an unstated basis.⁹ The same page shows a consensus target price of $72.50, which I give no weight E (estimate).¹⁷

IIThe moat

A durable business keeps others from taking its profits. For a chip maker the record shows whether it does. Over 2016 to 2025 ST’s operating margin averaged 15.9%, weighted by revenue, and ran from 1.5% in 2025 to 27.5% in 2022 C (calculated).^1314155 Its gross margin averaged 41.0%, with a low of 33.9% in 2025 and a high of 47.9% in 2023 C (calculated). Peers that earn more at the gross line show what protection looks like: Texas Instruments earned 64.2% and NXP 57.8% in the third quarter of 2026, against ST’s 37.3% C (calculated).³⁴²

That gap is the first thing to understand. ST’s products are less differentiated, or its plants are less efficient, or both, than the leaders in analog and automotive chips. The second reading is the more hopeful one: ST has spent two years with unused plant that other firms do not carry. In the fourth quarter of 2025 unused capacity charges alone cost 2.6 points of revenue, and in the first quarter of 2026 about 2.2 points.¹²² But TI’s 64% includes years of advantage from 300mm analog plants built before the last upturn. ST is building the same thing now, with grants from European governments to help.⁵

Evidence it is widening

The first sign is the price of access. Amazon was willing to take warrants worth $331M at grant for revenue milestones, and the data center ambition went from half a billion to a billion dollars in 2026 within five months.⁷⁶ New customers are choosing ST for parts that were not a large business two years ago. The second is cost. The company has said it will cut annual costs by a high triple-digit million dollar sum by the end of 2027, and selling, general and research costs fell from 34.3% to 29.6% of revenue between the first halves of 2025 and 2026.⁷ A narrower moat would not allow that and still gain share.

Evidence it is narrowing

Chinese competition is the main worry in the segments where ST is weakest. Power and Discrete revenue fell 31.5% in 2025 after a 20.6% fall in 2024, and its margin is now negative.⁵ In March 2025 ST signed a gallium nitride agreement with Innoscience, a Suzhou-based maker.⁵⁷ I read that as an admission that low-cost GaN from China is the pace to match, which is practical and also a concession. Pricing is another warning. The 20-F blames part of the 2025 revenue fall on lower selling prices and mix, as well as on volume.⁵ A business with strong pricing power would show less of that in a down year.

My judgement: ST has a good position in several niches and an average one in the whole. It is not a business where the record supports a high premium on today’s margins. The test is the next downturn. If operating margin stays above 10% through it, the protection is larger than I assume. If it falls to 5% again, the price needs a different argument.

IIIThe competitive advantage

The table compares ST with four rivals in the latest quarter each has reported. The quarters do not line up exactly: onsemi and Infineon had not yet reported September when I looked.

The price is the strange part. ST’s shares cost 29 times annualized operating income, against 24 for TI and 13 for NXP, which earn three to five times more of each dollar of revenue. A buyer in the first group is paying for a margin that has not arrived. The market sees the recovery and the gross margin guide of 39%, and prices a company some way along that path.

Where ST wins

Growth is the clearest win. Revenue rose 17.9% in the third quarter, against 28% at TI and 19% at NXP.²³⁴ Embedded Processing is growing 24% with a 23% margin, and RF and Optical 46% at 26%. Both lines are up against a weak year. The second win is position in Europe: ST has 300mm fabs in France and Italy and silicon carbide plants in Italy and Singapore, backed by two governments.⁵ That helps with supply security and with grants, though not with price. Third is cash. Net financial position was $2.1B at 26 September, with $5.3B of liquidity and $3.2B of debt.² The bonds due in 2027 were settled in the third quarter: ST paid $750M in cash and delivered about 4.2M treasury shares rather than issuing new ones.²

Where it loses

Margin first, as the table shows. Second is the fixed cost of the plant, which unused capacity exposes: charges were $88M in the fourth quarter of 2025 and $69M in the first quarter of 2026.² Third is concentration. Apple at 17.7% is the largest buyer, and its share has risen three years in a row.⁵ Fourth is China. The power segment competes with low-cost Chinese makers, and Sanan, ST’s partner in silicon carbide, is itself a Chinese company.⁷ Fifth is the plant program, which has cost $560M of charges since the start of 2025 and is meant to pay back from 2027.⁵² None of these positions is permanent, and the first test of the plan is whether margins reach 39% in the fourth quarter.

IVManagement

Jean-Marc Chery, born in 1960, has been president and chief executive since May 2018. He joined Thomson Semiconducteurs in 1986, which became ST, and ran technology and manufacturing before the top job.⁵ He was reappointed in 2021 and 2024, and his current term ends at the 2027 meeting. On 13 October the Supervisory Board said it would propose one more year, so the company has not named a successor.⁵⁸ The chief financial officer, Lorenzo Grandi, joined the Managing Board in May 2024 and now has the title president and CFO.⁵¹⁶ On 29 October the company said Remi El-Ouazzane will become chief technology officer and Marco Cassis chief strategic officer from January 2027.¹⁶

How they are paid, and what that pays them to do

The two board executives get a base salary, a cash bonus and shares that vest after three years. In 2025 Chery was paid $5.64M in all, of which $1.28M was base, $0.78M a bonus for 2024 earned at 60% of target, and $2.31M the value of shares that vested. His pay was $9.47M in 2024 and $7.30M in 2023. Grandi was paid $2.81M in 2025.²¹ The bonus can reach 210% of base for the chief executive and 150% for the CFO. The share grant is capped at 100,000 shares a year for Chery and 90,000 for Grandi, scored over three years on revenue growth, operating margin and a sustainability index that includes diversity and carbon ratings.²¹ The targets are not published, which I cannot judge.

What that pays them to do is grow revenue and lift margin. Both are fine goals. Neither measures the return on the plant the company is building, or the cash left after paying for it, and the company can hit revenue growth in 2027 simply by recovering from a bad 2025. That is the gap I would watch. The sustainability criterion is a small part of the score and I do not read it as a worry.

How much of their own money is in

Not enough for me to tell. The 20-F says nobody on either board or in senior management owns 1% of the company, and it sets a guideline that Managing Board members build a holding of 1.5 times salary.⁵ As a foreign private issuer ST files no Forms 4, and I found no itemized holdings for Chery. His unvested awards at the maximum are 300,000 shares, worth about $14M at the close, and Grandi’s are 155,000, about $7M C (calculated).²¹¹ That is real money to them and it is stock they were given. In April 2025 the Italian press reported that two Managing Board members had sold shares before earnings. The Supervisory Board answered that the sales were made by the plan administrator through an automatic process to meet Swiss tax rules, were legal and complied with policy.⁵ A securities class action filed in August 2024 and covering March 2023 to January 2025 alleges excessively positive statements on 2024 revenue. The court denied a motion to dismiss on 15 September 2025, and the company says it has strong defenses.⁷ I cannot size the exposure, and I cannot see anything in the filings that would settle it.

Who really controls it

ST Holding owns 27.5% of the shares, with 70.0% held by the public and 2.5% in treasury at the end of 2025.⁵ Bpifrance and the Italian ministry own ST Holding between them and have a shareholders’ agreement that gives each the right to nominate the same number of Supervisory Board members; any vote of ST Holding needs both to agree.⁵ BlackRock is the only other holder the filing names, at 5.2%.⁵ Two governments with different priorities on jobs and plant sites are the largest owners. The 20-F says that may conflict with other shareholders’ interests and that it can limit the company’s ability to issue shares or buy them back, since a holder crossing 30% of votes must bid for the rest under Dutch law.⁵ I do not see a bad faith case. But this is a company where Rome and Paris have a say in where fabs go, and an outside buyer pays for that.

What they did with capital over a full cycle

Over 2016 to 2025 ST earned $16.3B and generated $27.9B of operating cash. It spent $20.3B on plant and intangibles, net of grants, which is 17.5% of revenue, took out stock pay of $1.6B, and was left with $5.9B of owner free cash flow C (calculated).^1314155 It paid $2.3B of dividends and bought back $2.6B of stock. Together they took $4.9B of the $5.9B, leaving little for the plant program that followed.

My view is simple. Capital allocation has two parts. The big part, plant, has the weaker record: the company spent $10.3B in 2022 to 2024 and earned $175M of operating income in 2025 C (calculated).⁵ Cycles explain much of that, and the plan to move to 300mm may well pay off. The smaller part, the financing and the buybacks, is good. It issued convertible debt near the top and repurchased near the bottom, and the dilution I describe later is mostly the Amazon warrant. A holder with a ten-year horizon is therefore betting on the plant, not on the financing.

VFinancials

Ten years of ST contain one full cycle and the start of a second. Revenue rose 20% in 2017, 25% in 2021 and 26% in 2022, then fell 23% in 2024 and 11% in 2025 C (calculated).^1314155 One year says little. I use averages across the years, and I use the 2016 figures as restated in the 2019 20-F where they differ.

What an owner could really take out

Net income runs ahead of cash. In 2016 to 2025 ST reported $16.3B of net income and produced $27.9B of operating cash, but spent $20.3B on plant and intangibles, and paid its people $1.6B in stock that a pure cash view would not count. That leaves $5.9B, or 36 cents for each dollar of earnings C (calculated). In 2024 owner free cash flow was $0.01B and in 2025 $0.02B, against net income of $1.56B and $0.17B.⁵

Two items inside that number need a note. First, stock pay: ST pays about $0.2B a year in shares, 1.6% of revenue. I count it as a cash cost in my model, as I do on every page, since share grants dilute someone and the company spends cash buying shares back to offset them.⁵ Second, grants. ST received $320M, $441M and $258M of capital grants in 2023 to 2025, and it held $300M of advances for plant it has not yet built at 26 September.⁵² I count the grants received as a reduction in plant spending, as the company does, but they depend on governments continuing to pay.

What does it take to stand still? Depreciation and amortization ran $1.85B in 2025 and will be about $2.0B in 2026 E (estimate). Net plant spending averaged 17.5% of revenue over the decade C (calculated). I put the cost of keeping the plant at its current capability at $1.9B to $2.2B a year today, and I cannot narrow that honestly E (estimate). Against $14.5B of 2026 revenue and 9.8% operating margin, that is why free cash flow in the first nine months was $0.6B even without the NXP payment, and −$0.3B with it C (calculated).² At the third-quarter run rate, operating income of $1.47B a year taxed at 16%, plus $2.0B of depreciation, less $2.1B of plant, leaves about $1.1B before working capital, or 2.6% of the market value E (estimate).

The balance sheet

At 26 September ST held cash, deposits and securities of $5.3B against financial debt of $3.2B, a net position of $2.1B, or $1.8B after capital-grant advances.² The debt includes two new convertible bonds of $0.75B each (2031 and 2033), and most of the rest is loans from the European Investment Bank. S&P rated it BBB+ with a negative outlook and Moody’s Baa1 with a stable one at the end of July.⁷ Pensions and similar obligations were $386M, and there is $407M of minority interest, mostly the Sanan silicon carbide joint venture.² I take all of those off the net cash, along with a $182M dividend already declared, and count $0.83B outside the business, or $0.93 a share C (calculated).

VIValuation

First, how this goes badly wrong

Before any numbers, I asked what would most likely ruin this investment, and worked back. Three things can.

• The recovery was borrowed from the future. Distributors bought 38.8% more in the third quarter than a year earlier and the company itself says their stock is still below target, so some of 2026 is refilling pipes that may empty again.² After the last peak, revenue fell 23% in 2024 and 11% in 2025, and operating margin went from 26.7% to 1.5% C (calculated).⁵ A second fall from a higher fixed-cost base, with the new 300mm plants running half empty, would show up as unused-capacity charges again.
• The plant bill never gets paid for. ST plans about $2B a year of plant spending, and depreciation is $2B. Power and Discrete has lost $67M to $124M a quarter for five quarters.² If Chinese makers keep prices low in silicon carbide and mature nodes, the margin I need cannot be earned, and owner cash stays near zero as it was in 2024 and 2025 C (calculated).
• The AI story is small. Data centers are about $1B of $14.5B, 7% of revenue.⁶² The share reached $79.91 in June, three weeks after the data center ambition was raised, and fell 42% by the end of October. If it does not double, the part of the price that rests on it goes, and the car and industrial base has to carry the rest.

None of these needs a surprise. They set my bear case and the reason I use a margin that cycles. Outside the model sit two more: a world with higher tariffs, which the company says its outlook does not include,² and the owners in Paris and Rome, who can ask for plants where I would not build them.

What the price implies

At $46.74 the market values ST at $41.9B on 896.0M shares.¹² Book equity is $17.9B, so the price is 2.3 times book C (calculated).² It is 3.1 times revenue for the twelve months to September and 16.9 times the mean 2027 estimate of four analysts, who sit at $2.77 a share against my base of $2.68.⁹ The share has fallen 42% from $79.91 on 22 June, and risen 4.9 times from $9.45 ten years ago C (calculated).¹ My $2.68 and the Street’s $2.77 are close. The disagreement is further out.

A growth rate from a recovering year means little, so I ask the question another way. My base cash through 2029 is $6.9B, or 17% of today’s market value, discounted to $5.4B. Add what I count outside the business and the price has $35.6B left to pay for. That needs $3.9B of owner cash a year from 2030, growing 3% for good C (calculated). The best year in the last ten was $1.55B, in 2023, and the ten-year average $0.59B. So the price assumes ST becomes a company that produces two to three times its best year, every year, and it asks for it before the plant is built.

Put in margin terms, the price needs operating margin to settle near 25% on my base revenue, against my 22% and the ten-year average of 15.9% C (calculated). ST reached 25% or more in 2022 and 2023 only. Texas Instruments earns 47% and NXP 32%, so the target is not absurd for the sector. It is high for this company.

What I can’t judge

I cannot tell you the margin ST will earn at $20B of revenue. Nobody can say how much of Power and Discrete’s loss is start-up cost and how much is price, or whether data center revenue doubles. I also cannot see the terms of the Amazon warrants, the targets that score management’s shares, or what Rome and Paris would do with a weak year. Here is the range, and read my $40 with it in mind.

The discount rate

I use 11.0%. The capital asset pricing model starts with the 4.94% ten-year Treasury yield at the Q4 2026 close and Damodaran’s 4.20% implied equity risk premium of 1 October.²²²³ My ten-year beta for STM against the S&P 500, on month-end returns, is 1.75, which gives 12.3%. The five-year beta of 1.96 gives 13.2%, and the three-year 2.23 gives 14.3% and is mostly 2026’s run C (calculated).¹²⁴ I take 11.0%, a beta of about 1.4. ST’s earnings are steadier than a memory maker’s, so I sit below the 12.0% on my Micron page, and above the 10.5% on my Broadcom page because ST’s margins are weaker. I also use a rate lower than the raw betas because the downturn is already in my cash flows. It would be easy to argue for a point more. A point more takes the base to about $36, and at 10.0% it is about $47, which is the price.

Three cases

I keep the model simple. For each year I set revenue and operating margin, which is after stock pay, charges and purchase accounting. Depreciation is 16.5% of opening plant and plant spending is a stated number, net of grants. Tax is 16%, close to the 16% of the nine months to September 2026 (income tax $106M on $666M before tax).² Working capital is 20% of added revenue, about what ST carries now. The fourth quarter of 2026 is the company’s guide, $4.15B and 39% gross margin, with my estimate of a 14% operating margin. Each year’s cash arrives at year end. Net cash and other items outside the business are $0.83B. Per-share values divide by 896.0M shares and add dilution from the Amazon warrant by the treasury method: all 24.8M vest in the base and bull cases, and only the 0.6M already vested in the bear.⁷

The base is generous on one point and cautious on another. Generous: it has margin reach 22% in 2029 from 9.8% today, on a fixed-cost base that has barely covered itself. Cautious: it has no data center upside beyond $2B and no help from a weaker euro. The Q4 outlook assumes $1.14 to the euro and the market rate on 30 October was $1.1172.²¹¹ A weaker euro helps ST, since many costs are in euros and prices are in dollars, and I do not count it.

The bear is not mild. Its 2027 revenue is 7% below the fourth-quarter run rate, and its margin never gets past 11%. Only 21% of its value is cash due through 2029, so most of it rests on the long run. A stress case with no weight is worse still: 2027 revenue of $14.0B, 2028 of $11.5B, a loss in 2028 and an 8% margin after. It gives $10, about 79% below the price. That is close to 2024 and 2025 repeated and extended, and I think it unlikely.

Where this could be in 5 and 10 years

I also asked what the same model says for 2031 and 2036. The table gives my value on each date: the present value then of the owner cash that comes after, per today’s share count and in that year’s dollars. The third column adds the cash paid out up to 2036, including the $0.93 a share I count outside the business today. The last column is the yearly return from $46.74, from price growth plus that cash.

For the base to hold at five years, revenue must reach about $20B by 2029 and margin about 22%, with plant spending near $2.4B a year. By ten years it needs revenue of $26.5B and owner cash of $4.2B a year, nearly three times the best year in my record. The bear and bull differ on two numbers: settled margin, 11% against 27%, and growth, 1% to 3% against 7% to 10% a year. A third number, plant spending, matters more than it looks, since each $0.3B a year is worth $1.4 a share in the base.

The most likely way the base is wrong is on margin. If ST settles at 18% instead of 22%, the base is $32, and at the ten-year average of 16% it is about $27. The base value rises from $40 today to $53 in 2031 and $64 in 2036 only if each year’s cash arrives on schedule. These are estimates of value, not forecasts of the share price, and the market price can sit above or below value for years.

How good, and how good a price

These are two separate questions. The business: a decent position in analog, microcontrollers and sensors, a weak one in power, 49,000 people and two governments as its largest owners, and a ten-year record of 36 cents of owner cash for each dollar of earnings. It is not a great business by that record. The price: above my base by 17%, and well above the line where my bear case is covered. A fair business at a full price, on my numbers. The market is paying for the version in which margins reach where TI’s did before and the plant is full.

Other uses of the same dollar

The ten-year Treasury pays 4.94%. My base here returns 9.0% a year at this price, about four points above bonds, against the 11.0% I require, and my bear −3.9% E (estimate).²² The closest comparison on this site is Broadcom, which sells into the same AI demand with far higher margins: its base returns 7.5% a year at a price 25% above base. TSMC is 7.8% and Micron 7.2%, with the price 26% above base. Those pages use their own prices and rates. ST’s 9.0% looks better than all three, but with the weakest margin and the least certain plant, so I would not call it safer. Broadcom’s and TSMC’s earnings are steadier. ST offers more only if the margin recovery is real.

The margin of safety

• Price$46.74Close, Q4 2026 [1]
• Base value$40My estimate E (estimate)
• Margin of safety below$20set from my bear case C (calculated)

I set the margin of safety from the bear case, as on my Micron page, because the cycle drives this business. Below $20, 51% under my base value, my $15 bear value is at least three quarters of the price paid. The price is 134% above that line. So the rule is not met here, by a wide margin, and I say so. At $46.74 a buyer who gets my bear case loses 68% of the money in this model, against a gain of 37% in my bull. Two things would change it. A downturn in which margin holds above 10% would lift my bear, and a price at or below $20 would meet the rule. The lowest close of 2026 was $27.12, and of 2025 $17.78.¹

VIIThe case for the price

What most people miss

Three things cut in the company’s favour and don’t make the headlines. First, ST has been a good trader of its own stock. From July 2024 to June 2026 it bought 20.5M shares for $551M, about $27 each, and it bought none in the first half of 2026 when the price ran up.⁷ In June it sold $1.5B of convertible bonds at conversion prices of $119.98 and $121.92, premiums of 55% and 57.5%, which implies the share was near $77 at the time.⁷ It sold dear and bought cheap. Second, the Amazon warrant works in the shareholders’ favour more than it seems: the 24.8M shares at $28.38 vest only as Amazon reaches product revenue milestones, so the dilution comes with the sales.⁷ Only 0.6M have vested.⁷ Third, the balance sheet kept its net cash through a $936M acquisition and a year of heavy plant spending: $2.1B at 26 September, up from $2.0B in June.²

The bull case

In plain words: demand is recovering from a low, ST has spare plant to fill, its costs are falling, and a new kind of buyer is arriving. Jean-Marc Chery said with the third-quarter results that demand rose further on strong bookings and improving visibility, that inventory in distribution is still below the company’s target, and that demand remains well above supply.² On 2 June the company raised its data center ambition to about $1B of revenue in 2026 and said it could double in 2027 with current engagements.⁶ The fourth-quarter guide is $4.15B of revenue and a 39.0% gross margin, against 37.3% in the third quarter.²

The one named investor I can verify is a filing, not a view. Situational Awareness LP, Leopold Aschenbrenner’s fund, held 7,802,700 STMicroelectronics shares at 30 June 2026, worth $584.3M or $74.89 each, according to its 13F-HR filed on 14 August 2026, which is 2.9% of the portfolio the filing reports.²⁵ That says nothing about what it holds now or why. A 13F does not show short positions or holdings outside the US, and the next one, for 30 September, is due in mid-November. I do not infer a view from it. Four analysts on Nasdaq’s feed give a mean target of $72.50, three rating it a buy and one a hold; I give that no weight.¹⁷

Against reality

Today’s numbers support the first half. If ST does what it says through 2029, my base has it generating $6.9B of owner cash by the end of 2029, 17% of the market value E (estimate). If revenue reaches $20B and margin settles near 25%, the price is possible. The base alone does not get there: it has 22%.

The bull case needs much more, and multiplying shows how the odds shrink. Say data center revenue doubles in 2027, 60%. Cars and industry recover as hoped, 60%. Chinese rivals do not cut prices hard in power chips, 55%. The new plants fill and cost cuts arrive, 60%. No downturn before 2030, 50%. Those are my guesses and they are not independent. Multiplied, they give 6% E (estimate). Allowing for overlap, perhaps 10% to 15%, so the 25% weight I give the bull in my weighted figure of $40 is generous. Each step looks likely. All five together do not.

What today’s price needs is shorter. Revenue near $20B by 2029, 60%. Margin reaching 25% when it has been above that for two years of ten, 30%. No downturn that takes it back to 15% before 2032, 55%. That multiplies to 10% E (estimate), and with overlap perhaps 15% to 20%. So the price asks for the less likely half of my range. A stock can still sit there, and has: the share reached $79.91 in June.¹

What a 10x would take. Ten times $46.74 is $467.40, about $0.4T of market value. NVIDIA, the largest company I cover, is about $5.0T. At 20 times earnings that needs $21B of net income a year, 50 times ST’s best year of $4.2B, and at a 25% net margin $84B of revenue, which is 6% of the whole $1.51T chip market WSTS forecast for 2026.²⁶ By 2036 that means 26% a year for ten years from a market value of $41.9B C (calculated). The stock rose 4.9 times in ten years, from $9.45, and 4.5 times from its 2025 low of $17.78 to its June high, so a fivefold move from a low has happened.¹ A tenfold move from here has no precedent in this company’s record, and I would not build a case on it.

Where I come down

The gross margin guide and the AI data center ramp moved me most: they are why my base sits where it does and why the bear is not lower. The 38.8% jump in distributor sales did not, because that is the part most likely to reverse. Margin holding above 10% through the next downturn, or power chips reaching break-even, would raise my values most. A price near $20 would meet my safety rule.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Tariffs and trade. The company says its outlook does not include any further change to global tariffs, and it lists trade policy first among its risks.² A third of revenue is sold in the Americas and EMEA and nearly two thirds in Asia Pacific, by shipping location.⁵
• China. Power chips face Chinese makers on price, ST’s silicon carbide joint venture is with a Chinese firm, and much of Asia Pacific demand is Chinese.⁷⁵ Export rules could cut either way. I size some cases in the next subsection.
• Currency. The company says a stronger euro lifts revenue and costs together and that a weaker one reverses both, with a lag in pricing.⁷ The euro was $1.1172 on 30 October against $1.14 assumed in the guide.¹¹²
• Debt and rating. S&P rates it BBB+ with a negative outlook and Moody’s Baa1 stable.⁷ Net cash is $2.1B, but $0.75B of convertible bonds fall due in 2031 and $0.75B in 2033, and committed credit lines of $1.2B are undrawn.⁷²
• The plan. The footprint program has cost $560M of charges since the start of 2025 and must also deliver the high triple-digit million savings by the end of 2027.⁵² A miss would show in gross margin first.
• Governments. The French and Italian states own 27.5% through one holding company, and the 20-F says their interests can differ from other holders’. They also gave $258M of grants in 2025, which I count as a cost reduction.⁵
• Legal. The securities class action over 2024 guidance survived a motion to dismiss. A patent case brought by Purdue University carries a $32M judgment under appeal.⁷
• Customers. Apple was 17.7% of 2025 revenue. A change in its design would land on the segments that now grow fastest.⁵

Geopolitical risk, and what it means for the value

ST builds in Europe and Asia and ships nearly two thirds of its sales to Asia Pacific. Its 300mm fabs are in Crolles and Agrate, its silicon carbide plants in Catania and Singapore, and its back-end plants in Malaysia, Malta, Morocco and Shenzhen. It also owns 49% of a silicon carbide venture with Sanan in Chongqing that makes devices only for ST. China held $671M of the $11.1B of net plant at the end of 2025, about 6% C (calculated).⁵ Two governments stand behind 27.5% of the shares.⁵

Washington decides the tariffs. The 20-F names the 100% chip tariff the US has threatened and the Section 232 inquiry opened in April 2025.⁵ The August 2025 EU and US statement says the US intends to keep the combined rate on EU semiconductors at 15%, but it is not binding.²⁹ The Americas took 16% of 2025 revenue by shipping location C (calculated).⁵ Beijing decides on metals and on Chinese buyers. It suspended its ban on gallium and germanium exports to the US until 27 November 2026, and licences are still needed.³⁰ In October 2025 its curbs on Nexperia chips stopped production at several carmakers.³¹ On cars, EU registrations rose 5.7% in the first half of 2026 and battery-electric cars took 20.7%, the segment silicon carbide is built for.²⁸

The odds below are my judgement E (estimate), drawn from the record: both sides have used controls twice in two years and paused them, a tariff cap exists but is not law, and the Nexperia stop happened once. Each row changes only that item in my base. My base carries today’s tariffs and no break, as does the company’s outlook.²

No single case takes the base near the $20 margin-of-safety price. The tariff and China cases together take it to about $30, which is still above my bear value of $15. The $46.74 price already sits above my base, so the market is not charging for these risks on top of my values. If it charges for some of them, it must expect more than I do elsewhere. Quiet, slow items cost less than the odds suggest: the states’ plant choices and a car-demand dip cost about $1 each.