Theses›Industrial›Specialty silicon

Specialty silicon

Specialty silicon (≥28nm RF, power, sensors, automotive) doubles to $80B+ revenue by 2030

  • Shut
  • $197B by 2035

The call, a draftGrowth is winning: the 2024-25 mature-node glut is reversing in 2026 and the RF, power, analog and MEMS slice grows at about 12.8%, though the gains go to listed foundries and chipmakers. The window is structural.

D1D2D3D4R1R2R3C1C2C3foragainst D1D2D3D4R1R2R3C1C2C3foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the drivers.
  1. D1AI power chips pushed 8-inch utilisation to about 90% in 2026 from about 80% in 2025, with SMIC up about 10%, Vanguard up 4-8% and UMC confirming price rises for the second half.
  2. D2A new specialty fab costs $5-10B and takes 4-6 years, and capital flows to leading-edge, so capacity stays short while 5G, EV power, radar and MEMS demand keeps growing.
  3. D3EU Chips Act 2.0, published 3 Jun 2026, puts €120B to 2035 into joint procurement and buy-European rules aimed at the mature and specialty chips Europe consumes.
  4. D4GlobalFoundries' silicon photonics revenue is on track to roughly double in 2026 to about $400M and pass a $1B run-rate by the end of 2028.
  5. R1On the broad line, mature-node utilisation sits near 70%, prices fell about 5% and profit fell about 23% year on year, so the aggregate grows at about 7%.
  6. R2China heads from 21% to about 30% of world mature-node capacity, with Hua Hong, Nexchip and SMIC able to absorb much of the 28nm-and-older expansion.
  7. R3Leading-edge foundry revenue grows faster on AI demand, so specialty leads on units while leading-edge leads on revenue.
  8. C1The RF, power, analog and MEMS slice grows at 9% or more a year through 2027 while Western specialty-fab utilisation recovers.31 Dec 2027
  9. C2GlobalFoundries passes $10B of revenue with its specialty platforms growing 25% or more.31 Dec 2028
  10. C3Specialty silicon on 28nm and older nodes passes $80B of global revenue.31 Dec 2030
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today1234567predictionswatch 20272028202920302031today1234567predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1CONTESTED: Specialty silicon revenue exceeds leading-edge (≤7nm) foundry revenue in 2027.Dec 2027 · open
  2. 2Through 2027 the specialty slice grows ≥9%/yr and Western specialty-fab (GF / STMicro / X-Fab) utilisation recovers; the mature-node glut stays isolated to commodity Chinese ≤28nm logic, not the specialty slice.Dec 2027 · open
  3. 3Whether the glut spreads from commodity Chinese logic into specialty processes.Dec 2027 · open
  4. 4GlobalFoundries achieves >$10B revenue with >25% growth from 22FDX-derived specialty platforms by 2028.Dec 2028 · open
  5. 5Whether GlobalFoundries passes $10B of revenue on its specialty platforms.Dec 2028 · open
  6. 6Specialty silicon (≥28nm) revenue exceeds $80B globally by 2030.Dec 2030 · open
  7. 7Whether specialty silicon passes $80B of global revenue.Dec 2030 · open
Expressionslisted, private, technologies
Private · 2 private companies.

Vanguard International Semiconductor · United Microelectronics Corporation UMC

31 more private companies tracked, not shown

Technologies
What changedthe record, newest first
  • 23 Jun 2026Mature-node foundries are climbing the value chain via specialty differentiation, not geometry shrink: UMC 22nm specialty hit a record 14% of revenue and GF silicon photonics is on track to double in 2026 and pass a $1B run-rate by 2028. The differentiated specialty layer is real and growing, and the incumbents are competing hard for it, which is why a new entrant needs a properly cornered module rather than generic specialty design work.
  • 3 Jun 2026The 2024-25 'mature-node glut' is reversing in 2026: AI power-IC demand and capacity discipline are lifting legacy-node prices and utilisation, and the specialty slice (RF, power, analog, MEMS) is growing low-double-digits, faster than commodity logic. The real oversupply risk sits in commodity Chinese 28nm logic, a different game than the design-win-locked specialty fabs.
Reading
Show the reading