Patent Strategy
Semiconductor Patent Strategy
Circuit design, process technology, cross-licensing programs, and SEP licensing for the wireless standards stack.
FAQ
What types of patents are most important in the semiconductor industry?
The semiconductor industry relies on multiple overlapping patent categories: CIRCUIT DESIGN PATENTS: patents on circuit architectures; specific cell designs; novel logic structures; memory cell designs (DRAM; NAND Flash; SRAM); processor microarchitecture; digital signal processing circuits; analog circuit topologies (PLLs; ADCs; DACs); SCOPE: generally claim the functional structure (circuit elements + their interconnections); typically covers the design itself; FABRICATION PROCESS PATENTS: patents on the manufacturing process; the most valuable secrets in the industry; gate dielectric materials (high-k dielectrics for CMOS scaling); transistor structures (FinFET; GAAFET); photolithography methods (EUV lithography process optimization); deposition techniques (ALD; CVD; MOCVD); etching processes (atomic layer etching); chemical mechanical planarization; SCOPE: cover the sequence of process steps; specific temperatures, pressures, and chemistries; equipment configurations; these patents often expire before the process is fully deployed at scale; LAYOUT PATTERNS: specific physical layout configurations; memory array layouts; standard cell libraries; these may be protected by patents AND by the Semiconductor Chip Protection Act (SCPA) mask work registration; PACKAGING AND INTERCONNECT: advanced packaging patents (3D stacking; chiplet architectures; through-silicon vias (TSVs); HBM memory configurations); interposer designs; bump architectures; gaining importance as Moore's law slows; STANDARDS-ESSENTIAL PATENTS (SEPs): patents covering wireless standards (Wi-Fi; 4G LTE; 5G NR; Bluetooth; USB; PCIe; Ethernet); extremely valuable because they are essential to implementing the standard; subject to FRAND licensing commitments; key holders: Qualcomm; Ericsson; Nokia; InterDigital; Apple; Samsung; Huawei; EDA/DESIGN TOOL PATENTS: patents on electronic design automation tools; simulation methods; formal verification techniques; synthesis algorithms.
How does cross-licensing work in the semiconductor industry?
Cross-licensing is the dominant patent strategy among major semiconductor companies: WHAT CROSS-LICENSING IS: two or more companies agree to license each other's patents, typically under a blanket agreement covering all patents owned by each party at the time; each party can practice the other's patents without paying per-patent royalties or risking infringement suits; ECONOMIC RATIONALE: in semiconductors, every major company needs access to thousands of patents owned by competitors; litigation over all relevant patents would be prohibitively expensive and would threaten the entire industry's ability to function; cross-licensing converts patent litigation risk into a negotiated business arrangement; NET BALANCING PAYMENT: if one party's portfolio is more valuable than the other's, a net balancing payment (cash or royalty) is agreed; the party with the weaker portfolio pays the stronger; IBM'S CROSS-LICENSING PROGRAM: IBM has one of the largest semiconductor patent portfolios; IBM cross-licenses with virtually every major semiconductor company; IBM received over $1 billion/year in patent licensing revenue in the 1990s-2000s; QUALCOMM'S CDMA/WCDMA/LTE LICENSING MODEL: Qualcomm licenses its SEPs and baseband chip patents separately from cross-licensing; charges royalties on the entire device price (controversial; FTC and foreign antitrust regulators challenged); CHIPS ACT IMPLICATIONS: as chip supply chains become strategic, governments are increasingly interested in how patent licensing affects chip availability and cost; ARM's LICENSING MODEL: ARM licenses its processor architecture IP (instruction set + microarchitecture) to chip designers; foundational to the mobile and IoT chip ecosystem; fabless companies pay per-unit royalties or upfront license fees; NEWCOMER PROBLEM: companies without large patent portfolios (startups; new entrants) are at a disadvantage in cross-licensing — they have nothing to offer in exchange; must build portfolios or pay cash.
What are the patent strategies for fabless semiconductor companies?
Fabless companies (who design but do not manufacture chips) require specific patent strategies: FABLESS MODEL: design chips; outsource manufacturing to foundries (TSMC; Samsung Foundry; GlobalFoundries; SMIC); IP protection focuses on design; not manufacturing process; DESIGN IP PROTECTION: circuit architecture patents (what the chip does; how it is organized); algorithm patents (DSP algorithms; AI inference optimizations); interface and protocol implementations; memory subsystem designs; PROCESS IP: fabless companies typically do NOT own process patents (owned by the foundry); foundry provides process under a foundry agreement with process IP protections; fabless companies may own process-adjacent IP: specific design rule optimizations; device configurations achievable with the foundry's process; MASK WORK PROTECTION: the Semiconductor Chip Protection Act (SCPA) protects the specific physical layout (mask work) from direct copying; mask work registration at the US Copyright Office; 10-year protection; prevents direct layout copying but not independent design of same circuit; STANDARDS LICENSING EXPOSURE: fabless companies that implement wireless standards (Wi-Fi; 5G; Bluetooth) in their chips have SEP exposure; either need to cross-license or include SEP licensing pass-through from chip customers; ARM LICENSEES: companies that license ARM architecture pay per-chip royalties to ARM; these costs must be factored into chip pricing; STARTUP FABLESS STRATEGY: file patents on key design innovations before tapeout; provisional first; file before announcing the chip; focus on differentiating architecture rather than process technology; be aware of SEP licensing needs early — budget for them; ACQUISITION TARGETS: fabless companies with strong design IP portfolios are attractive acquisition targets; buyers value the portfolio + design team + customer relationships.
How do SEPs and FRAND licensing work for wireless semiconductor standards?
Standards-essential patents dominate licensing costs for wireless semiconductor implementations: THE WIRELESS STANDARDS ECOSYSTEM: 5G NR; 4G LTE; Wi-Fi 6/7 (IEEE 802.11ax/be); Bluetooth 5.x; USB 3.x/4; PCIe 5/6; NVMe; Ethernet 400G — each implemented through standards developed in standard-setting organizations (3GPP for cellular; IEEE for Wi-Fi/Ethernet/USB; JEDEC for memory); DECLARING ESSENTIAL PATENTS: companies that participate in SSOs disclose patents they believe may be essential to implementing the standard; these declared SEPs may or may not actually be essential (many declared SEPs are ultimately not technically essential); FRAND COMMITMENT: declaring essential status typically requires a FRAND licensing commitment; the company cannot refuse to license to anyone willing to pay FRAND rates; QUALCOMM'S LICENSING APPROACH (AND CONTROVERSY): Qualcomm charges royalties on the ENTIRE DEVICE PRICE (phone price) rather than on the chip price; critics: the patent's contribution should be valued at the chip level (the SSPPU), not the device level; Qualcomm: the patent enables the device's cellular functionality; FTC v. Qualcomm (N.D. Cal. 2019): found Qualcomm engaged in anticompetitive practices (reversed by 9th Circuit 2020); LICENSING COSTS FOR CHIP MAKERS: estimated total 5G SEP royalty stack: $10-$20 per handset from all SEP holders; individual SEP holder rates: Qualcomm (~3-5% of handset price); Ericsson; Nokia; InterDigital each ~0.5-2%; DESIGN-AROUND FOR SEPs: technically impossible for cellular standards (the standard mandates specific behavior that the SEP covers); for Wi-Fi: some room to implement non-standard optional features; STANDARD CONTRIBUTION STRATEGY: participate in SSO standard development; contribute patented technology to the standard; declare those patents as essential; collect FRAND royalties across the entire industry once the standard is adopted.
What are the patent risks for semiconductor startups and how should they manage them?
Semiconductor startups face specific patent risks that require proactive management: PRE-LAUNCH IP PROTECTION: file provisional patent applications before discussing the chip design with foundries, ODMs, or potential customers; under NDA is insufficient protection — file before disclosure; key innovations to patent: novel circuit architectures; algorithms implemented in silicon; new memory cell designs; interface protocol optimizations; energy efficiency techniques; PRE-TAPEOUT FTO ANALYSIS: before committing to tapeout ($1M-$5M+), conduct a freedom-to-operate analysis for key circuit elements; identify potentially blocking patents from major players (Qualcomm; ARM; TSMC; Intel; Samsung; Synopsys; Cadence); FTO before tapeout is far cheaper than a design change after; SEP LICENSING BUDGET: for chips implementing wireless standards, budget for SEP licensing early; estimate total stack cost per chip; factor into chip pricing; discuss with counsel whether to approach SEP holders before commercial launch or wait until sales volume justifies their attention; FOUNDRY AGREEMENTS: review foundry process technology license carefully; understand what design rule information is proprietary; confirm the foundry's IP indemnification scope; understand customer indemnification limits; ARM/MIPS/RISC-V ARCHITECTURE LICENSING: if using a licensed ISA (ARM; MIPS), understand the per-chip royalty burden; RISC-V (open ISA) eliminates architecture licensing fees but offers less ecosystem support; PATENT PORTFOLIO BUILDING: aim for 3-10 patents covering core differentiating technology by Series A; more by Series B; investors value patent protection; acquirers value patent portfolios; DEFENSIVE PUBLICATION: for innovations that are too numerous to patent (minor optimizations; implementation details), consider defensive publication (IP.com; technical disclosure) to create prior art and preserve freedom to operate.
Related Guides