No. 1 globally by 2024 revenue, CIC

The world's R&D, decoded.

Patsnap is the AI-native platform that helps companies innovate faster by reading patents, scientific literature and legal records at global scale, then turning that knowledge into decisions.

In plain English Google organizes the internet for consumers. Patsnap organizes the world's inventions for companies, then uses AI to help R&D teams decide what to build, what to avoid and how to protect it.

Explore the story Jump to financials

FY25 revenue US$124.1m 18,000+ customers 60+ countries 110% KA net dollar retention

01 / The Problem

Why Patsnap matters

R&D is a US$3 trillion decision problem. Most teams still fly partially blind.

Global R&D spending reached about US$3.0 trillion in 2024 and is projected to reach US$4.5 trillion by 2030. Yet roughly 40% may be squandered on duplicative work because companies cannot fully see what has already been invented.

What changes with Patsnap

Months reviewing patents and papers manually

→

Minutes to search, synthesize and cite the relevant evidence

Fragmented tools across R&D, IP, legal and strategy teams

→

One shared intelligence platform across the innovation lifecycle

IP risk discovered after money has already been spent

→

Early visibility into freedom-to-operate, novelty and competitive conflicts

Simple investor frame

Patsnap moves from selling software seats to automating knowledge work.

The old market was IP search software. The new market is AI-assisted R&D execution: novelty searches, FTO analysis, drug-compound discovery, patent drafting and competitive intelligence.

40%

Potential duplicate R&D spend

CIC estimate cited in the roadshow draft.

90%

Biomed projects may fail to commercialize

A high-stakes sector where misses are expensive.

Strategy & scouting

Track competitors, white space and emerging technology areas.

Ideation & feasibility

Pressure-test ideas against patents, papers and market signals.

Experimentation

Use domain data for compound, material and technical solution discovery.

IP protection

Run novelty, prior-art and freedom-to-operate workflows.

Commercialization

Identify partners, licensing paths and competitive risks.

Patsnap's Directly Serviceable Market -- Segment Build

Innovation intelligence spend across IP and R&D services categories · illustrative

IP Software (legacy) IP Services R&D Services Total

TAM Expansion · Orders of Magnitude

Legacy

IP Mgmt
Software

Patent Law
Firm Services

IP Legal
Services

Legal Process
Outsourcing

IP Services
(Commercial)

CRO Discovery
Services

Tech Scouting
& Consulting

R&D Comp.
Intelligence

Directly Serviceable
Market

IP Software

IP Services

R&D Services

Total

Source: CIC. Segment build illustrates Patsnap's directly serviceable opportunity across IP software, IP services and R&D services labor categories addressable through AI-enabled applications and AI-native R&D agents. Bars are sized for illustrative comparison.

02 / Market Expansion

Why now

AI expands Patsnap's market from IP departments to the entire R&D value chain.

The category is shifting from databases people search to agents that complete high-value research tasks. That lets Patsnap capture a larger share of software, services and R&D workflow spend.

From

IP departments

Patent search, prior art and legal workflows.

→

Full R&D organizations

Strategy, discovery, development, IP and commercialization.

From

Software seats

Analysts pay to search and export reports.

→

Work budgets

AI agents replace outsourced analysis and manual expert labor.

The category point This is not just a better patent database. It is an operating layer for innovation work, similar to how Bloomberg became more than financial data by embedding workflows on top of proprietary information.

Innovation intelligence TAM

CIC estimate, 2024 to 2030E

21.8% CAGR

2024

US$5.3B

2030E

US$17.3B

US$3.0T

Global R&D base, 2024

US$4.5T

Global R&D base, 2030E

03 / Why It Is Defensible

The platform

Three layers compound into a moat.

The data foundation makes the AI better. The AI workflow layer makes the data usable. The agent layer turns both into repeatable, monetizable work.

Layer 3 / Agent layer

AI-native R&D agents

13 productized agents across Eureka IP, Engineering, Life Sciences and Materials. They complete workflows such as novelty search, FTO, patent drafting, lead-compound discovery and competitive Q&A.

Eureka IP Eureka Engineering Eureka LS Eureka Materials

Layer 2 / AI workflow layer

LLM-agnostic middle layer

Patsnap treats third-party LLMs as interchangeable. The defensible layer is proprietary retrieval, citation control, workflow orchestration, fine-tuning corpus and domain-specific reasoning.

Citation-backed Model-agnostic Domain tuned

Layer 1 / Data layer

Innovation knowledge graph

205M+ patents, 212M+ scientific literature records and 1.6B+ legal data points across 174 jurisdictions, organized into a computable knowledge graph and refreshed daily.

2B+ entries 2,200+ fields 174 jurisdictions

Investor translation A general AI model can answer broad questions. Patsnap can answer high-stakes R&D questions because it owns the domain data, the workflow layer and the expert-ready output format.

04 / Differentiation

Why Patsnap wins

Not a chatbot. Not a search tool. A vertical AI platform for high-stakes R&D.

Data is the moat

Our profound data foundation -- the innovation knowledge graph.

No. 1 in number of patent and journal records globally, according to CIC. Built over nearly two decades, Patsnap's data foundation consists of curated, domain-specific technical information that functions as a computable technical knowledge graph rather than a traditional database.

205M+

Patent records

212M+

Scientific literature

1.6B+

Legal records

174

Jurisdictions

Specialized domain corpora

Chemical structures

280M+

Biological sequences

1B+

Clinical trials

1.2M+

Corporate tech trees

3M+

The data factory -- six-stage pipeline

01 -- Ingest

Multi-source collection

174 patent offices, 20K+ journals, regulatory filings and litigation databases.

02 -- Normalize

Standardize formats

2,200+ data fields normalized across jurisdictions and languages.

03 -- Translate

Multilingual alignment

12 languages with domain-tuned translation models.

04 -- Structure

Knowledge graph

Entities, relationships, claims, citations and legal status.

05 -- Enrich

Domain augmentation

Chemical structures, bio sequences, clinical data and corporate genealogies.

06 -- Serve

Daily refresh

Real-time API, retrieval-grade indexing and citation guarantee.

Example -- biomolecule curation at nano-scale detail

Data curation with deep vertical know-how

For biomolecules and chemistry-heavy literature, Patsnap does not just OCR a PDF. It parses figures, tables, sequences, reactions, entities and relationships into structured fields that agents can reason over.

Application

X-embedding problem-based

Problem clues engine

Comp / bio-sequence search

Image retrieval

RAG

Knowledge
extraction

Companies
Data

Business entity NER
Business entity link

Tech issue
Tech solutions
TRIZ / innovation matrix

Biomedical

Drug / disease / target
Preclinical drug SAR
Bio sequence / mutation
Antibody / antigen

Chemical / Materials

Chem/material reactions
SPUM
Catalyst / solvent fields

Data
analysis

Multi-report formats:
PDF, PPT, Word, Excel, image, web...

Layout analysis

Table analysis

Chemical reaction recognition

OCR

Image reclass

Biomolecule structure recognition

Public formula LaTeX

Sequence parsing

Table reconstruction

Proprietary document parsing model

Example: turning one biomolecule paper into computable data

Raw scientific content

Paper / patent

Figures, claims, tables and assay text.

Biomolecule

Sequence, mutation, antibody and antigen mentions.

Experiment

Target, disease, cell line, dose and result.

Knowledge graph

Linked to patents, literature, companies and legal status.

Extract -> normalize -> link -> serve to AI agents

Extracted field	Structured output example
Biomolecule type	Antibody fragment / engineered protein
Sequence features	Heavy-chain CDRs, light-chain CDRs, mutation sites, modification tags
Target biology	Target protein, disease indication, pathway and species context
Assay evidence	Binding affinity, inhibition, expression system, dosage and experimental conditions
IP context	Claim scope, priority family, cited prior art and freedom-to-operate signals
Agent-ready use	Novelty search, sequence similarity, lead optimization, partner scouting and patent drafting

Why it matters

General AI models start from broad internet-scale knowledge. Patsnap starts from a curated R&D knowledge graph, so its agents can reason over patents, science, legal status and domain-specific technical data with citations.

Dimension	General LLMs	Traditional tools	Patsnap
Data	Open web; weak patent and jurisdiction coverage.	Fragmented databases and keyword search.	Domain-curated knowledge graph across patents, literature, legal status and specialized scientific corpora.
Reasoning	Good general language capability, but poor domain grounding and weak citations.	Rules-based analysis that still requires manual expert synthesis.	LLM-agnostic workflow engine with retrieval, citations, evaluation and domain-specific reasoning.
Output	Generic chat response that experts must verify and rewrite.	Search results and static reports.	Decision-ready outputs: novelty searches, FTO reports, lead-compound analysis and patent drafts.

Novelty search benchmark, May 2026

85%

Patsnap hit rate across 500 expert-validated novelty-search queries.

Patsnap

85%

Perplexity

39%

GPT-4o

17%

Gemini

14%

In high-stakes R&D, missing the wrong patent can cost millions. This is why domain-specific AI matters.

University research institution

92%

Reduction in time spent on compound searches, plus a 10x acceleration in commercialization workflows.

Global industrial company

40x

Efficiency improvement in identifying competitive technologies.

International law firm

94-95%

Reduction in patent review time, turning manual expert work into scalable workflows.

05 / Customers

Commercial proof

The world's biggest R&D spenders already rely on Patsnap.

The business is global, diversified and increasingly recurring: more than 18,000 cumulative customers across 17 industries and over 60 countries.

18,000+

Cumulative customers

Across 17 industries.

60+

Countries

Global innovation footprint.

110%

KA net dollar retention

Existing key accounts spend more over time.

>70%

Revenue from existing customers

Subscription-anchored revenue base.

Large enterprise motion

Deep relationships with global R&D leaders in semiconductors, life sciences, energy, chemicals and consumer electronics.

Cross-sell runway

AI apps create entry points; AI agents expand wallet share by automating higher-value workflows.

Founder-led execution

Founded in 2007, with 318 R&D staff and 44% of R&D staff carrying relevant domain background.

06 / Financials

Financial highlights

Growing revenue, expanding margins and real operating cash flow.

Patsnap is not just an AI concept. It is a scaled SaaS business with improving monetization, a rising agent mix, positive adjusted profit and growing cash generation.

Revenue by product type

US$m, 2023-2025

AI apps AI agents

16012080400

US$92.7m

2023

US$105.8m

2024

US$124.1m

2025

What matters AI-native R&D agents grew from 8.1% to 14.9% of revenue mix in 24 months, increasing the weight of the highest-value product layer.

US$17.4m

Adjusted net profit, 2025

From US$(2.1)m in 2023 and US$5.2m in 2024. Adjusted net margin reached 14.1%.

US$33.1m

Operating cash flow, 2025

From US$4.2m in 2023, supported by an asset-light software model and subscription prepayments.

75.1%

OPEX as % of revenue, 2025

Down from 94.8% in 2023, showing operating leverage as revenue scales.

US$60.3m

Greater China

48.6% of FY25 revenue, +7.2% YoY.

US$42.3m

Americas & EMEA

34.1% of FY25 revenue, +24.6% YoY.

US$21.5m

APAC ex-China & Japan

17.3% of FY25 revenue, +55.2% YoY.

07 / Growth

Growth strategy

Five ways Patsnap keeps compounding after IPO.

The next phase is about deepening the agent layer, monetizing consumption, expanding wallet share and extending global reach without losing discipline.

Agentic evolution

Embed predictive AI deeper into daily R&D workflows and verticals such as pharma, materials, semiconductors and communications.

Product evolution

Turn Eureka into a host environment with APIs, orchestration and hybrid pricing.

Targeted customers

Deepen wallet share in existing accounts and expand into clean energy, semiconductors and consumer electronics.

Disciplined M&A

Acquire niche datasets, AI capabilities and domain talent that strengthen the moat.

Globalization

Strengthen North America, Europe and APAC while entering South Korea, Australia and the Middle East.

The investment case

No. 1 AI-native innovation intelligence platform. Proprietary data. Vertical AI. Agent monetization. Cash-generative scale.

18 years of data-building, 18,000+ cumulative customers, 110% key-account net dollar retention, US$124.1m FY25 revenue, 14.1% adjusted net margin and US$33.1m operating cash flow.

AI-native R&D intelligence HKEX / SGX Project Kinabalu

IPO Roadshow html