Table of Contents
The headline numbers
SemiWiki's tally of the 2025 semiconductor design IP market, as reported on 2026-09-05, puts the market at $9.45 billion, up 12.7% year over year. Two forecasts sit alongside it: MarketsandMarkets puts the market at $18.64 billion by 2032, implying a 10.2% CAGR for 2026–2032.
The split of that $9.45B across twelve sub-segments is the part worth reading:
| Segment | Share of design IP revenue |
|---|---|
| CPU IP | 43% |
| Wired interface IP | 26% |
| GPU + ISP | 6% |
| DSP, system, security, standard cell libraries | ~3% each |
CPU IP and wired interface IP together absorb close to 70% of industry budget. Processor IP on its own accounts for 52.2% of the total.
Royalties are outgrowing licences
The single most informative line is the revenue-mix change: licensing revenue grew 10.7%, royalty revenue grew 16.1%.
That gap is a leading indicator, and here is why. Licensing revenue is recognised when a design is signed. Royalty revenue is recognised per unit shipped. When royalties grow faster than licences, it means designs signed in prior years have finished the long path through tape-out, qualification and production ramp, and are now shipping in volume.
For anyone tracking where RISC-V actually stands, this is the metric to watch rather than core-shipment press releases. Design wins are cheap to announce. Royalty-bearing volume is not.
Arm's position, and the change in its role
The reported figures for Arm: +26.4% growth, 49.4% of the total design IP market, and 93.3% of the CPU IP market.
The coverage argues that Arm's share is not the story — its role change is. Arm has moved from being a neutral IP supplier to shipping its own compute hardware, which puts it in a different relationship with the cloud operators and system architects who were previously its customers. When your IP supplier becomes your competitor, supply-chain planning changes.
Where that leaves an open standard
We are not going to claim a RISC-V share number, because the source does not publish one. What the data does support is a narrow, defensible statement about buying criteria:
- Multi-source supply. A single-vendor CPU IP market at 93.3% is a concentration risk that procurement teams already model. An open ISA with multiple independent IP vendors is the structural alternative, and it is the argument that does not depend on anything political.
- Cost structure. An open ISA carries no per-core royalty on the instruction set itself. The licence and royalty economics reported above are for commercial IP products regardless of ISA, so this is about who captures the royalty, not whether one exists.
- Long-term availability. Custom silicon programmes in automotive, industrial and infrastructure run on decade-plus horizons. An ISA governed by a standards body with many implementers is a different availability proposition than one governed by a single vendor's roadmap.
The honest counterpoint is in the same data: CPU IP is 43% of the market and one vendor holds 93.3% of it. That is not a market that turns quickly.
What the numbers do not tell you
Be careful what you conclude from this:
- No RISC-V revenue or share is broken out. Any figure you see attributing a percentage of CPU IP to RISC-V does not come from this source.
- The 93.3% CPU IP share is a share of commercial CPU IP licensing revenue. It is not a share of cores shipped, and it says nothing about internally developed cores that never enter the licensing market.
- Royalty growth is measured across all IP categories, not per ISA. It does not indicate which architecture's designs are ramping.
- The 2032 forecast is a third-party projection, not an observed trajectory. The 10.2% CAGR is below the 12.7% observed in 2025, i.e. the forecast models deceleration.
What is not disclosed
None of the following appears in the source retrieved, and nothing here is estimated:
- Absolute dollar figures per sub-segment (only percentage shares were published)
- RISC-V-specific revenue, licence count, unit volume or royalty base
- Geographic breakdown
- Vendor-level figures for any non-Arm IP supplier
- The underlying survey methodology, respondent set, and whether the figures are calendar-2025 or fiscal-2025
- Any unit-shipment number by architecture
Sources
- 半导体产业纵横 (Semi IC Views), "95亿美元设计IP市场：Arm的固守与RISC-V的进击", 2026-09-05, citing SemiWiki and MarketsandMarkets — https://www.toutiao.com/article/7681883051084808746/
- Referenced primary data: SemiWiki 2025 semiconductor design IP market report
- Referenced forecast: MarketsandMarkets, semiconductor design IP market to 2032
Verification notes
- All figures in this draft are quoted from the 2026-09-05 trade-press article, which attributes them to SemiWiki. The SemiWiki report itself was not retrieved directly for this draft; if these numbers are going into a customer-facing document, confirm them against the SemiWiki original.
- The MarketsandMarkets figures ($18.64B by 2032, 10.2% CAGR for 2026–2032) are a third-party forecast, not observed data.
- The interpretation of royalties-outgrowing-licences as a volume-ramp leading indicator is this draft's analysis, clearly separated from the reported figures. The royalty/licence growth split itself is reported.
- No political or supply-chain-autonomy framing has been added. The framing used throughout is multi-source supply, cost structure and long-term availability.
- Not disclosed and not estimated: every item in the "What is not disclosed" section above.