Theses›Semiconductor manufacturing›Chiplets

Chiplets

Chiplet-based products grow from ~$3B (2021) to ≥$20B by 2027 at 40% CAGR, then capture ≥8% of total semiconductor revenue by 2030.

  • Bull
  • Closing
  • $54B by 2035

The callThe drivers win for one slice: the photonic I/O chiplet, where the EDA majors hold no library. Everything else in chiplets is priced, absorbed or moving to the packagers.

D1D2R1R2R3C1foragainst D1D2R1R2R3C1foragainst
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 packages become assemblies of compute, memory, I/O and optics on different processes, as NVIDIA's Vera Rubin already is, so the limit moves to how fast the pieces talk.
  2. D2Copper die-to-die links work inside a package but not between packages and racks at AI power, and the EDA majors have no photonic library.
  3. R1Value is moving down-stack to packaging, where no startup has entered the top six, and the die-to-die IP slice grows only 9.6% a year.
  4. R2Licensing economics: every chiplet vendor wants premium pricing and system designers want commodity, so the open marketplace never closes and value stays with TSMC, AMD, Intel and the EDA majors.
  5. R3Cross-vendor interop progress cuts against this view, because a strong merchant marketplace relieves the die-to-die scarcity that is the only obvious moat for a newcomer.
  6. C1Two shipped products mix chiplets from two independent vendors over UCIe, making cross-vendor chiplets a merchant market; the March 2026 Intel and Cadence interop demo was the step before.end-2027
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today123456predictionswatch 20272028202920302031today123456predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1Chiplet market reaches $20B revenue by 2027.Dec 2027 · open
  2. 2AI accelerator architectures shipped in 2027 are >60% chiplet-based.Dec 2027 · open
  3. 3UCIe reaches multi-vendor mix-and-match in production by 2027.Dec 2027 · open
  4. 4Two or more products ship with chiplets from different IP vendors over UCIe, and over 60% of leading-edge accelerators ship as multi-die stacks.Dec 2027 · open
  5. 5Chiplets capture ≥8% of semiconductor revenue by 2030.Dec 2030 · open
  6. 6Chiplets reach 8% of semiconductor revenue.Dec 2030 · open
  • ·TYLsemi as the merchant-chiplet test, and the known-good-die gap feeding the second-wave test theme.
Expressionslisted, private, technologies
Private · 9 private companies, 5 of them in the US.

Kandou · Cadence Design Systems · Alphawave Semi · Qualcomm · Eliyan · Blue Cheetah Analog Design · Baya Systems · Silicon Box · Chip Interfaces

4 more private companies tracked, not shown

Technologies
Who has already bet on itprivate rounds, from public sources

$465M raised by 6 companies, $441M of it in the last three years. Wave under way: Large rounds in the last three years; the category is funded and getting crowded.

Backed by a top-tier venture firm: Blue Cheetah Analog Design. Backed by a strategic investor: Eliyan, Silicon Box, Baya Systems.

What changedthe record, newest first
  • 7 Aug 2026A team that sold its last interconnect company to a large acquirer has raised again at the same layer, this time selling the integration work around the chip rather than the chip. When the hard part of a standard is everything the standard does not cover, that gap becomes the business.
Reading
Show the reading