China dominates the generative AI patent landscape with six times more inventions than the US, creating potential headwinds for open-source and decentralized AI initiatives.
The World Intellectual Property Organization just dropped a patent landscape report that puts hard numbers on something the tech world already suspected: generative AI isn’t just growing fast, it’s growing at a pace that makes “fast” feel inadequate.
GenAI patent families surged from roughly 14,000 in 2023 to over 37,800 by 2025. That’s nearly triple the volume in just two years. The patent activity from 2024 and 2025 alone exceeds the entire preceding decade combined.
China is running away with it #
Between 2014 and 2023, China racked up more than 38,000 GenAI inventions. The United States, for all its Silicon Valley swagger, managed roughly one-sixth of that number.
The top patent holders read like a who’s who of Chinese corporate and institutional power. SoftBank, Tencent, Ping An Insurance Group, Baidu, and the Chinese Academy of Sciences all sit near the top, with six of the ten largest GenAI patent owners being Chinese firms.
The US and Japan are filing more aggressively too, but the trend remains clear: most GenAI patent activity stays concentrated domestically in countries like China.
LLMs have dethroned GANs #
The WIPO data also tracks a meaningful architectural shift within generative AI itself. Large language models have overtaken generative adversarial networks as the most patented technology category in the space.
Image and video outputs still dominate in terms of application, but the underlying engine has changed. Researchers and corporations have moved systematically away from experimental architectures like GANs and variational autoencoders toward mainstream, deployable applications across industries.
The share of GenAI patents within the broader AI patent ecosystem grew from 4.2% in 2017 to an anticipated 8.7% in 2025.
The 2017 transformer architecture paper, which introduced the attention mechanism that underpins virtually every modern LLM, appears to be the inflection point. The commercial boom triggered by large language models after 2022 is the obvious catalyst for the subsequent spike in patent filings.
What this means for crypto and decentralized AI #
The crypto industry has placed enormous bets on decentralized computing networks, open-source AI models, and token-incentivized machine learning. Projects building on these premises now face a patent landscape that is rapidly consolidating around a handful of massive corporations and state-backed institutions.
When SoftBank and Tencent own thousands of GenAI patents, the barrier to entry for smaller, community-driven projects doesn’t just rise. It potentially becomes a legal minefield. Decentralized AI protocols that rely on open-source model architectures could find themselves navigating patent claims they didn’t anticipate, particularly as these technologies move from research into commercial deployment.
This creates an asymmetry that’s worth watching. Large enterprises can absorb patent litigation costs as a routine business expense. A decentralized autonomous organization with a treasury of governance tokens generally cannot.
There’s also a geopolitical dimension. If China controls the majority of GenAI intellectual property and most decentralized AI projects are built by Western or globally distributed teams, the regulatory and legal friction between these ecosystems could become a material factor in project viability.
WIPO has a follow-up Technology SPARK update scheduled for 2026 that will include data through 2025.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our