Humanoid Robot Patent Landscape. Patent Landscape & Market Signals.

The humanoid robot patent landscape has gone from a research footnote to one of the fastest-moving fields in all of intellectual property. GreyB counts more than 19,500 filings since 2006 and a record 2,592 patent families in 2025 alone — so the real question is no longer whether to watch the field, but how to read it for a decision. This page maps the trend, the geography, the owners and the open space from the public record.

What a Humanoid Patent Landscape Shows in 2026

A humanoid robot patent landscape turns a scattered record of filings into three answers a strategy team can act on: is the field accelerating or cooling, which organisations already hold the ground, and where is the ground still open. Each is a different cut of the same public data — the filing trend, the assignee ranking, and the white-space gap — and each drives a different decision, from where to file next to whom to license from or design around.

The backdrop makes the exercise urgent. Patent-analytics firm GreyB counts 19,501 humanoid-robotics patents filed since 2006, resolving to 15,022 patent families, of which 15,931 records remain active. Nearly half of that activity has landed in the last five years. For a technology that was a lab curiosity a decade ago, that is an unusually steep adoption curve — and it means a snapshot taken today will look materially different in twelve months.

This page reads the public record the way we build it for clients: trend first, then geography, then ownership, then the open space. Every figure below is attributed to its source so you can verify it, and the method behind it is the same competitive-intelligence workflow we apply to any fast-moving field.

How Fast Humanoid Robot Patenting Is Growing

The growth is not a talking point; it is in the filing counts. GreyB records 1,087 patent families in 2023 rising to 2,592 in 2025 — more than a doubling in two years, and the highest annual total on record. The International Federation of Robotics (IFR) frames the same surge from the money side: general-purpose robotics funding grew roughly fivefold between 2022 and 2024 to pass US$1 billion a year, with patent filings climbing at about a 40% compound annual growth rate since 2022.

Two cautions matter when reading any recent-year number. First, patents publish 12–18 months after filing, so the most recent year always understates reality and will be revised upward. Second, counts are not value: a doubling in volume tells you attention is flooding in, not that each new filing is defensible. That is why a landscape pairs the volume trend with a quality read, covered below.

For a strategy team, the practical signal is timing. A field compounding at 40% a year closes its obvious white space quickly; the gaps that are claimable in 2026 may be occupied by 2028. The filing trend is the clock on the wall for anyone deciding whether to file now or wait.

Where the Filings Concentrate: China’s Scale, America’s Quality

Humanoid filings are not evenly spread. China dominates by raw volume — GreyB puts it at 9,975 patents, roughly half the global total and nearly four times the United States. The rest of the field trails well behind.

Scale is not the whole story. LexisNexis Intellectual Property Solutions, scoring more than 26,000 humanoid-related patent families with its Patent Asset Index, finds China holding 73% of active morphology-related families but 63% of portfolio strength, while the United States holds only about 5% of families yet roughly 11% of portfolio strength and the highest average patent quality across interaction, morphology and control domains. In plain terms: China out-files, the US out-concentrates value per patent. A landscape that counts only documents would miss that split entirely.

Who Owns the Humanoid Robot Patent Landscape

The assignee ranking is where the humanoid robot patent landscape becomes a target list — the parties to license from, design around, or benchmark against. GreyB’s global top ten mixes long-standing robotics and electronics incumbents with a wave of Chinese start-ups:

Two features stand out. The leader’s 521 filings do not come close to controlling the field — the top ten together hold a minority of all records, so this is a contested landscape, not a locked one. And the newcomers matter: LexisNexis finds China-based start-ups occupying six of the top ten dedicated-humanoid positions, with Fourier, AgiBot and LimX Dynamics ranking highest. More than 30% of Chinese filing activity, GreyB notes, originates from universities — Beijing Institute of Technology (159), Harbin Institute of Technology (127), Zhejiang University (121) and Tsinghua University (81) — a pipeline of licensable, early-stage IP that rarely appears in a company-only view.

The Technology Clusters: Actuators, Hands, Locomotion and Control

Splitting the field by technology is what turns a count into a map. Humanoid IP concentrates in a handful of hard problems, and the ownership looks different in each. On the hardware side, PatSnap’s leaderboard of 14,310 hardware records is led by Harmonic Drive Systems with 584 filings (4.1%) — a joint-and-actuator supplier rather than a robot integrator — followed by UBTech (317) and Boston Dynamics (237). Hardware applications alone climbed from 756 in 2016 to 2,836 in 2025.

The component clusters that absorb the most filing effort are consistent across providers:

• Actuators and joint modules — harmonic drives, electric and tendon-driven joints; the densest and most supplier-controlled cluster.
• Dexterous hands and manipulation — multi-finger grippers and tactile sensing, where filing is accelerating fastest and claims are still relatively open.
• Bipedal locomotion and balance — gait, balance control and fall recovery, long dominated by Honda, Boston Dynamics and Japanese incumbents.
• Control, planning and embodied AI — the software layer where LexisNexis records Chinese inventors raising activity roughly fivefold in a decade, and where general-robotics players such as Alphabet and Nvidia appear.
• Morphology and sensing — head, neck and whole-body design, now 73% China-held by active family count.

Mapping a portfolio onto these clusters is what tells a client whether their freedom to operate is at risk in actuators (crowded, supplier-owned) or still achievable in manipulation (growing, but not yet saturated).

White Space: Where the Field Is Still Open

The most valuable output of a landscape is not what is claimed but what is not. Even in a field compounding at 40% a year, the density is uneven. Actuators and bipedal locomotion are crowded and heavily held by incumbents and suppliers; by contrast, dexterous manipulation, tactile and whole-body sensing, human-robot interaction safety, and the embodied-AI control layer show lower claim density relative to the commercial attention they are drawing.

Geography compounds the gap. Because 81% of hardware priority filings originate in China, a large share of humanoid IP is filed first — and sometimes only — in Chinese jurisdictions. For a company filing in the US, Europe or Japan, that can mean a crowded global picture hides genuinely open space in a specific office, or that a Chinese filing never enters the national phase where you operate. Reading white space per jurisdiction, not just globally, is where a landscape earns its fee.

White space is a hypothesis, not a conclusion. The deliverable is a ranked list of candidate gaps, each with the evidence behind it, so your own inventors can confirm whether a gap is open because it is hard, because it is uncommercial, or because no one has got to it yet.

What the Filing Counts Do Not Tell You

A count is a starting point, not an answer, and three distortions trip up anyone who reads the raw numbers alone. The first is the publication lag: because applications stay unpublished for 12 to 18 months, the most recent year or two always looks like a decline when in fact the filings simply have not surfaced yet. GreyB flags exactly this effect for 2026. Treating the latest data point as a downturn is the single most common misreading of a fast-growing field.

The second is the gap between volume and value. China files far more than anyone else, but the LexisNexis quality scoring shows the United States converting a much smaller share of filings into a disproportionate share of portfolio strength. A benchmarking exercise that ranks competitors by document count alone will systematically overstate the volume filers and understate the quality filers — the opposite of what a dealmaker needs to know.

The third is scope. ‘Humanoid’ is not a single patent class; the relevant art sits across robotics, actuation, machine vision, tactile sensing and machine learning. A search that leans on one classification code or one keyword set will miss whole clusters — which is why comparing two published studies that quote different totals (19,501 here, 26,000+ families there) usually reflects different scope definitions rather than a contradiction. Fixing the scope to the decision at hand is the first thing a credible study does, and the reason two honest analysts can report different headline numbers for the same field.

Reading the Landscape for Strategy, FTO and Deals

A landscape like this earns its cost when it changes a decision. The same analysis answers several at once. For R&D, it shows which clusters are crowded enough to require a design-around and which are open enough to file into. For freedom-to-operate, it surfaces the incumbents and suppliers — Harmonic Drive, UBTech, Boston Dynamics, Honda — whose claims a new product is most likely to read on. For corporate development, it benchmarks a target’s portfolio against the field and separates genuine invention from volume padding, using a quality measure rather than a raw count.

Because the field is contested rather than controlled, the strategic window is real but short. No single owner dominates, the top ten hold a minority, and universities are generating licensable early-stage IP — which means a well-timed filing or in-licence can still establish position before the 40% growth curve closes the gap.

How to Commission a Humanoid Robot Patent Landscape

The sharper the brief, the more decision-ready the output. A strong request names the technology clusters you care about (say, dexterous hands and tactile sensing rather than ‘humanoids’ broadly), the jurisdictions you file or sell in, the competitors and suppliers to benchmark, and the decision the study must support — file, license, design around, or acquire.

From that, we deliver a filing-trend analysis, assignee benchmarking, a technology-cluster map and a ranked white-space readout, each line traceable to the public record. If you are weighing a humanoid programme, a filing decision or a deal, start with the brief and we will scope the landscape to the questions you actually need answered.

What You Receive

• A filing-trend analysis — humanoid application and grant momentum by year and office, with the publication-lag caveat built in
• Top-assignee benchmarking — who owns the field (UBTech, Honda, Boston Dynamics, Fourier and the start-up wave), how fast each is growing, and where portfolios overlap
• A technology-cluster map splitting humanoids into actuators, dexterous hands, locomotion, control/embodied-AI and morphology
• A per-jurisdiction white-space readout — the clusters and offices where filing density is still low enough to claim position
• A primary-source evidence pack citing GreyB, LexisNexis IP, PatSnap and IFR behind every number

Data Sources & References

This analysis draws on primary patent and market data:

• GreyB — Humanoid Robotics Industry Patent Landscape (updated 2026) — 19,501 patents / 15,022 families since 2006; 1,087 families (2023) rising to 2,592 (2025); China 9,975 vs US 2,574; top assignees led by UBTech (521), Honda (372) and Toyota (214).
• LexisNexis Intellectual Property Solutions — What Patent Data Reveals in Humanoid Robotics — 26,000+ families scored on the Patent Asset Index: China 73% of active morphology families but the US holds the highest average quality; China start-ups take six of the top ten dedicated-humanoid positions.
• PatSnap — Humanoid Robot Hardware Patent Leaderboard 2026 — 14,310 hardware records led by Harmonic Drive Systems (584, 4.1%), UBTech (317) and Boston Dynamics (237); hardware applications up from 756 (2016) to 2,836 (2025).
• International Federation of Robotics — World Robotics 2025 / Humanoid position paper — ~7,000 full-size humanoids sold in 2025 (mostly R&D); general-purpose robotics funding up ~5x from 2022 to pass US$1bn/yr, with patent filings growing ~40% CAGR since 2022.

Request a Humanoid Robot Patent Landscape

Tell us the clusters, jurisdictions and decision you need the landscape to support, and we will scope it to those questions.

Related PerspireIP work: Competitive IP Intelligence · Patent White Space Analysis · humanoid white-space case study.

Frequently Asked Questions