STMicroelectronics N.V. -- Thematic Exposure
6 / 10
Thematic Score — Weight: 25%
The Starlink/SpaceX thesis is CONFIRMED, not speculative. Management has quantified the exposure
four ways: (i) explicit "main LEO customer" + "second-largest LEO customer" references each quarter
since Q3'25; (ii) pre-announced ">$3B cumulative LEO revenue 2026-2028" at the May 4, 2026 investor
call; (iii) PIC100 silicon photonics in high-volume production to AWS starting Q1'26; (iv) multi-year
multi-billion-dollar AWS commercial engagement. Third-party reporting: >90% RF share in Starlink
terminals, 7.5B ICs into >10,000 satellites. LEO revenue ~$175M (2021) → ~$600M (2025) → ~$1B (2026E).
But all this lives in RFOC -- the smallest segment (~12% of revenue). The other 57% (AM&S + P&D)
is commodity-cyclical analog and power that drags the multiple. Score capped at 6 because no segment
has >30% share (fails oligopoly hard gate at the portfolio level); half-point lift for RFOC's
Starlink monopoly economics inside the smallest segment.
Weight: 25%
Starlink RF Share
>90%
Per management; 10+ yrs SpaceX co-development
ICs Shipped to Date
7.5B+
Into >10,000 Starlink satellites & terminals
LEO Target 2026-2028
>$3B
Cumulative LEO revenue (mgmt pre-announced)
Oligopoly Gate
FAIL
No segment >30% share; RFOC monopoly economic but only ~12% of rev
Segment Exposure & Market Share -- FY2025
| Segment | FY25 Rev | % of Rev | Market Share | TAM | Theme Growth |
|---|---|---|---|---|---|
| AM&S Analog / MEMS / Sensors |
$5,085M | 43.1% | ~5-8% in analog; top-3 auto MEMS post-NXP deal; fragmented in analog |
Analog TAM ~$90B; MEMS ~$18B |
Industrial/auto sensors; NVIDIA Robotics partnership; AWS multi-billion engagement |
| P&D Power & Discrete, SiC |
$1,685M | 14.3% | ~30% SiC contractual target; top-5 globally in power MOSFET; 5-player oligopoly (IFX/STM/Wolfspeed/onsemi/ROHM) |
SiC power: $2.7B (2025) → $8.4B (2030); broader power ~$20B |
800V DC AI datacenter (NVIDIA); SiC recovery to 2024 lvls by 2027; Sanan JV ramps China |
| EMP STM32 MCU + digital ICs |
$3,580M | 30.3% | #1 GP MCU worldwide (5 yrs running per Omdia); ~20-23% share; targeting return to 23% by 2027 |
GP MCU ~$24B; broader MCU ~$28-30B |
High-perf MCUs in pluggable optics (DC pull-through); 18nm STM32 launch |
| RFOC -- THESIS RF + silicon photonics |
$1,436M | 12.2% | >90% RF in Starlink terminals (mgmt; 7.5B ICs to >10K satellites); leading-edge silicon photonics on 12-inch (unique globally) |
Pluggable optics DC: $7B (2024) → $24B (2030) @ 23% CAGR; silicon photonics ~$2.65B → $9.65B (2030); LEO chips: $3B+ cumulative '26-'28 |
The thesis lives here. $600M LEO in '25 → ~$1B in '26; $1B AI/DC revenue target '27 |
| Total | $11,799M | 100% | — | — | — |
Oligopoly Hard Gate -- Segment-by-Segment Verdict
AM&S (Analog/MEMS/Sensors)
FAILS HARD GATE
Fragmented -- TI, ADI, Infineon, Bosch (MEMS), Sony (imaging) all relevant; no single >15%
share in MEMS post-NXP deal. Price-TAKER: Q1'26 mgmt explicitly said "low single-digit"
pricing pressure across portfolio. Engaged programs (Apple sensors, ADAS imaging) sticky
but commodity analog highly contested.
P&D (Power & Discrete, SiC)
FAILS HARD GATE
Infineon (#1 ~14% auto semi), Wolfspeed, onsemi, ROHM all >10-15% in SiC; STM
contractually targeting 30% SiC. Main SiC customer (Tesla widely understood) already
qualified second source per Q1'25 ("normal process to have at least two source").
Price-TAKER: China SiC price war explicitly cited Q2'25.
EMP (STM32 MCU)
PASSES (GP MCU subcategory only)
Renesas, NXP, Microchip, Infineon all in top-5; STM #1 in GP MCUs but only ~20-23% even
in that subcategory. Top-5 ~55-81% MCU concentration. Sticky once designed-in (years of
code/toolchain lock-in via STM32Cube). Mid-pricing power: STM32 ecosystem (1.5M devs)
confers real pricing power; Q1'26 "selected price increase" environment.
RFOC (RF & Optical)
PASSES DECISIVELY (Starlink RF)
In Starlink RF: STM is dominant >90% per management. SpaceX cannot
meaningfully second-source on this timeline -- a decade of co-development, 7.5B ICs
shipped, BiCMOS+panel-level packaging IP, US export rules favor STM as Euro-domiciled.
Price-SETTER in Starlink RF (monopoly economics). Mixed in silicon photonics broadly
(Intel ~21.5%, Coherent ~25% transceivers, Marvell, Lumentum, Broadcom).
Gate Summary: Only RFOC clears the Oligopoly Hard Gate, and only the
satellite-RF sub-business inside RFOC has true monopoly economics. EMP is a near-pass on the
STM32 franchise (~20-23% share is below the 30% bar). AM&S and P&D are commodity-cyclical
with documented price-taking behavior in this very cycle. At the portfolio level, the oligopoly
hard gate FAILS -- which is one of the three quality gates that caps the composite score at 4.0/10.
Starlink / SpaceX Thesis Evidence -- Management Direct Quotes
| Quote (verbatim) | Transcript | Topic |
|---|---|---|
| "Our low-earth-orbit satellite business based mainly on our BiCMOS and panel-level packaging technologies strongly progressed during the quarter. We were selected to develop a power amplifier controller for direct-to-cell satellites based on our proprietary BCD technology by our main low earth orbit customer, and we continued to ramp shipments to our second largest customer." | FY2026Q1 (Apr 2026) |
Starlink direct + 2nd LEO customer (Kuiper-class) |
| "On May 4, we will host a dedicated call on ST's low-earth-orbit satellites, explaining how we are going to achieve our ambition of well above $3 billion cumulative revenues over the period '26 to '28 for this opportunity." | FY2026Q1 (Apr 2026) |
LEO TAM / quantification |
| "We confirm our data centers revenue expectation to be nicely above USD 500 million for 2026 and well above $1 billion for 2027." | FY2026Q1 (Apr 2026) |
AI infra target (PIC photonics + power + MCU) |
| "We expanded our strategic engagement with Amazon Web Services through a multiyear multibillion U.S. dollar commercial engagement to enable new high-performance compute infrastructure for cloud and AI data centers." | FY2026Q1 (Apr 2026) |
AWS -- Kuiper adjacent + AI hyperscaler |
| "We also announced the start of high-volume production for our silicon photonics-based photonics ICs 100 -- PIC100 platform used by hyperscalers for optical interconnect for data centers and AI clusters." | FY2026Q1 (Apr 2026) |
Silicon photonics PIC100 ramping |
| "The strong momentum in optical connectivity technologies for data centers also contributed to a significant rise in demand for our high-performance microcontroller used in pluggable optics." | FY2025Q4 (Jan 2026) |
EMP pull-through from datacenter optical |
| "The low-earth orbit satellite business based on our BiCMOS and panel level packaging technologies continued to progress during the quarter with shipments ramping to our second largest customer." | FY2025Q4 (Jan 2026) |
2nd LEO customer (Kuiper/OneWeb-class) ramping |
| "In low-earth orbit satellites, we are expanding our customer base, and we anticipate continued revenue growth as low earth orbit constellation projects expand globally and penetrate new applications such as direct-to-cell constellation." | FY2025Q4 (Jan 2026) |
Direct-to-cell = new TAM layer (T-Mobile/Starlink) |
| "Our CCP segment grew sequentially 23% and year-over-year 22%. Definitively, it is linked to the low or satellite business we have, and it is driven both by our first customer and then by the second one." | FY2025Q4 (Jan 2026) |
Quantifies satellite as the CCP driver |
| "In low earth orbit satellites, we have further strengthened our leadership position in the rapidly growing low broadband market by beginning shipment to a second global customer, leveraging our combination of BiCMOS technology for front-end modules and paddle level packaging for user terminals." | FY2025Q3 (Oct 2025) |
Confirms 2nd LEO customer started shipping Q3'25 |
| "During Q3, we have seen an increased demand for photonics IC prototypes to be launched in the next quarter and beyond in our 300-millimeter wafer fab." | FY2025Q3 (Oct 2025) |
Silicon photonics demand pre-PIC100 ramp |
| "Our high level of innovation and ability to execute with MEMS and optical sensing as well as with our analog technology portfolio... also position us for growth in low earth orbital satellites and data centers. For example... new technologies to enable higher performance optical interconnect in data centers and AI clusters... bring better performance to address the ongoing evolution of optical interconnect for customers like Amazon Web Services." | FY2025Q1 (Apr 2025) |
First explicit AWS callout on silicon photonics |
Third-party corroboration (not in transcript): "ST claims an RF market share
above 90% in 2025, supported by a decade-plus co-development with SpaceX that has delivered
>7.5 billion ICs into millions of Starlink terminals and over 10,000 satellites"
-- Yahoo Finance / Global Banking and Finance reporting on STM's May 4, 2026 LEO investor call.
LEO revenue trajectory: ~$175M (2021) → ~$600M (2025) → ~$1B (2026E).
Bull Case Materiality -- Is $1B AI by 2027 Enough?
FY2027E Consensus Revenue
~$14B
Bloomberg/MS path: $14.1B '26 + cycle growth
$1B AI as % of FY27E
~7%
Material but bounded
AI + LEO Combined
~15-18%
Of FY27E rev; ~100% of incremental growth above cycle baseline
RFOC FY25 Op Margin
18.5%
Highest-margin segment; Q4'25 hit 23.4%
Materiality verdict: YES, but with two caveats.
(1) The $1B AI number is partially funded by P&D (40%) and EMP/RFOC (60%) per management's own framing -- so it is not pure RFOC upside; some of it is just P&D and MCU getting datacenter share rotation.
(2) $1B AI + $1B LEO in '26 = $2B at ~20% segment margin = ~$400M op income contribution = roughly enough to take EPS from the current $0.53 non-GAAP trough toward consensus $1.05-$2.10 in '26-'27. So the math works on the path to consensus; it does NOT validate the share price ALREADY $66.86 vs $55 consensus PT. RFOC FY2025 op income $265M on $1,436M revenue = 18.5% segment margin -- already the highest-margin segment. Q4'25 RFOC margin hit 23.4% and Q1'26 14.9% (mix shift toward photonics ramp).
(1) The $1B AI number is partially funded by P&D (40%) and EMP/RFOC (60%) per management's own framing -- so it is not pure RFOC upside; some of it is just P&D and MCU getting datacenter share rotation.
(2) $1B AI + $1B LEO in '26 = $2B at ~20% segment margin = ~$400M op income contribution = roughly enough to take EPS from the current $0.53 non-GAAP trough toward consensus $1.05-$2.10 in '26-'27. So the math works on the path to consensus; it does NOT validate the share price ALREADY $66.86 vs $55 consensus PT. RFOC FY2025 op income $265M on $1,436M revenue = 18.5% segment margin -- already the highest-margin segment. Q4'25 RFOC margin hit 23.4% and Q1'26 14.9% (mix shift toward photonics ramp).
Three Rubric Questions per Segment
| Segment | Share + TAM | Moat | Thematic Exposure |
|---|---|---|---|
| AM&S | ~5-8% in fragmented $90B analog + ~10-15% in $18B MEMS post-NXP deal. No oligopoly. | Manufacturing scale (300mm Agrate ramping); Apple/auto engaged programs; sensor IP. Moat real but narrow. | Indirect. NVIDIA Robotics partnership (sensors+MCU+motor control) and AWS engagement (MEMS for DC power telemetry/security) tag along. Not the primary play. |
| P&D | ~30% SiC contractual target ($2.73B-sized 2025 SiC mkt), top-5 in $20B+ broader power. Tied for bottom of the Hard Gate. | SiC manufacturing (only player with 8-inch SiC at scale + Sanan China JV); but Infineon/Wolfspeed/onsemi all racing same curve. | Yes via 800V DC datacenter (NVIDIA partner) + SiC for direct-to-cell satellite power amps. Mgmt expects SiC recovery to 2024 levels only by 2027 -- slow. |
| EMP / STM32 | #1 in GP MCUs at ~20-23%; broader MCU $28-30B TAM. Top-5 concentration ~55-81%. | STM32Cube ecosystem (1.5M unique devs, >160K AI projects in 12 months, +30% YoY) is the genuine moat -- code-level lock-in across millions of designs. | Pull-through into pluggable optics (datacenter) and direct-to-cell satellites (LEO); 18nm next-gen MCU launch. Real but modest. |
| RFOC | >90% RF in Starlink terminals (mgmt); leading position in silicon photonics on 12-inch ($2.65B → $9.65B 2030 TAM); user-terminal panel-level packaging. | BiCMOS process IP + 7.5B-IC track record + decade of SpaceX co-development + only 300mm silicon-photonics fab globally. The only true moat in the company. | This IS the theme. Starlink, Kuiper (2nd customer), direct-to-cell, AWS AI interconnect, PIC100 for hyperscalers all live here. |
One-Paragraph Assessment -- Confirmed or Speculative?
Confirmed -- but narrow and concentrated.
Management has now quantified the exposure four ways:
(i) "main low earth orbit customer" + "second largest customer" (= Starlink and almost-certainly
Kuiper) explicitly named every quarter since Q3'25;
(ii) ">$3B cumulative LEO satellite revenue '26-'28" pre-announced for the May 4, 2026 dedicated
investor call;
(iii) ~$600M LEO in 2025 → ~$1B in 2026 trajectory per third-party reporting;
(iv) "RF share >90%" in Starlink terminals with 7.5B ICs already shipped into >10K satellites.
This is not narrative -- it is a single-customer monopoly franchise built on BiCMOS + panel-level packaging IP, with the second customer (Kuiper) now ramping and direct-to-cell as a third TAM-expansion layer. What's speculative is the AI/datacenter $1B-by-'27 number, which depends on PIC100 ramping into AWS plus winning hyperscaler share against Intel/Coherent/Marvell -- that competition is real and STM is not dominant there. What's confirmed beyond reasonable doubt is the Starlink franchise. The risk is concentration: if SpaceX's launch cadence slows (Starship delays) or if Starlink decides to bring RF in-house (Tesla-style), the franchise unwinds fast. The bull pillar is that SpaceX cannot meaningfully second-source BiCMOS RF on a 12-36 month horizon and has every reason not to disrupt a working partnership during the Kuiper-competition arms race.
This is not narrative -- it is a single-customer monopoly franchise built on BiCMOS + panel-level packaging IP, with the second customer (Kuiper) now ramping and direct-to-cell as a third TAM-expansion layer. What's speculative is the AI/datacenter $1B-by-'27 number, which depends on PIC100 ramping into AWS plus winning hyperscaler share against Intel/Coherent/Marvell -- that competition is real and STM is not dominant there. What's confirmed beyond reasonable doubt is the Starlink franchise. The risk is concentration: if SpaceX's launch cadence slows (Starship delays) or if Starlink decides to bring RF in-house (Tesla-style), the franchise unwinds fast. The bull pillar is that SpaceX cannot meaningfully second-source BiCMOS RF on a 12-36 month horizon and has every reason not to disrupt a working partnership during the Kuiper-competition arms race.
Score Rationale: 6/10
6/10 — The Oligopoly Hard Gate caps
this at 5 because no segment has >30% share -- and that cap is real for the broader portfolio.
The half-point lift to 6 reflects:
(a) RFOC's >90% Starlink RF share is genuine monopoly economics inside the
smallest segment, which the gate logic was designed to reward;
(b) STM32 is the global #1 GP MCU with code-level lock-in, an unusual asset for
a European semi;
(c) the AI/LEO secular growth is now showing up in segment revenue (RFOC +33.9%
YoY, EMP +31.3% YoY in Q1'26) -- not just slides.
Why not higher: AM&S and P&D (combined 57% of revenue) are commodity-cyclical price-takers actively losing pricing in the cycle just ending; SiC oversupply (P&D -31.5% in '25) was a textbook hard-gate failure; ADAS ASIC (Mobileye EyeQ5/6/7) is now moving to TSMC long-term; the entire AI thesis depends on hyperscaler designs that haven't closed yet against superior-positioned incumbents (Intel, Coherent, Marvell). The score is appropriate for a company with one excellent thematic asset embedded in a mostly-mediocre portfolio.
Why not higher: AM&S and P&D (combined 57% of revenue) are commodity-cyclical price-takers actively losing pricing in the cycle just ending; SiC oversupply (P&D -31.5% in '25) was a textbook hard-gate failure; ADAS ASIC (Mobileye EyeQ5/6/7) is now moving to TSMC long-term; the entire AI thesis depends on hyperscaler designs that haven't closed yet against superior-positioned incumbents (Intel, Coherent, Marvell). The score is appropriate for a company with one excellent thematic asset embedded in a mostly-mediocre portfolio.
Data sourced from Daloopa (STM series IDs embedded as citation links throughout). STMicroelectronics Q1 2025 through Q1 2026 earnings call transcripts. Yahoo Finance / Global Banking and Finance reporting on May 4, 2026 LEO investor day. STM PIC100 / AWS partnership disclosure (newsroom.st.com). DataCenterDynamics PIC100 coverage. SEC 6-K FY2025 (STM and Space). MarketsandMarkets silicon photonics and SiC market data. Omdia GP MCU share data.