A freelance illustrator in Hangzhou fine-tunes a small style model, shares it to a community, and watches strangers rent it for their own posters. The line between toolmaker and social network is blurring, and investors just put real money on the side that blends both.
Key takeaways
- LiblibAI (哩布哩布AI) closed a US$130M (1.3亿美元) Series B in October 2025, led by Sequoia China (红杉中国) and CMC Capital (CMC资本) with a strategic investor; earlier backers Source Code, Shunwei, and Mingshi added more.
- Founded in late 2023, LiblibAI says it is China's largest multimodal model and creation community, with over 20 million AI creators.
- The platform aggregates image, video, 3D, and LoRA training, and in October 2025 shipped version 2.0 as an "AI professional creation studio."
- Backers framed the deal as a signal that Chinese AI investment is rotating from foundation models (大模型) to applications.
The round
LiblibAI announced the US$130M Series B in late October 2025. The lead investors were Sequoia China and CMC Capital alongside a strategic party; returning shareholders Source Code Capital (源码资本), Shunwei Capital (顺为资本), and Mingshi Ventures (明势创投) increased their stakes. Yuan Shi Capital (远识资本) served as the sole financial adviser. Several Chinese outlets called it the largest financing in China's AI-application track for the year, though that label is the outlets' framing rather than an independently audited ranking.
What LiblibAI actually is
Unlike a model lab, LiblibAI runs a creation community. It gathers open and closed image, video, and 3D generation models behind one workspace, and lets users train lightweight LoRA adaptations — small style or subject models — on top. The company says it has incubated more than 20 million AI creators across illustration, photography, e-commerce, posters, and IP work.
Version 2.0, released the same month as the raise, repositioned the product from a tool aggregator to an "AI professional creation studio," with stronger video generation, multi-model workflows, and professional effect templates — an end-to-end path from idea to finished clip.
Why investors rotated to applications
For two years Chinese AI capital chased foundation models. LiblibAI's backers argue the model layer is consolidating into a few giants, while the value is moving to platforms that turn models into usable creative work. Source Code's commentary, carried in coverage, pointed to platforms that actually land multimodal capability in creation, production, and consumption.
That thesis is unproven at scale. Community platforms live or die on creator retention, content moderation, and the cost of the compute they resell. A US$130M war chest helps, but does not settle those risks.
The global question
LiblibAI said the proceeds would accelerate its global expansion, building a multilingual, multi-creator content ecosystem. China's creation communities face a familiar tension: domestic scale is real, but overseas growth means competing with established image and video tools and navigating content and copyright rules in each market.
The community model travels awkwardly across borders. A LoRA trained on one market's aesthetic may not transfer, and local content moderation expectations differ sharply. LiblibAI's overseas plan will be a test of whether a China-built creator network can become a default workspace for non-Chinese creators, or remains a strong domestic niche.
What the deal says about China's AIGC money
The raise lands at a moment when Chinese AI funding is splitting. Frontier-model training is capital-intensive and dominated by well-funded incumbents, while application-layer communities with real user bases look like a better risk-adjusted bet to many investors. LiblibAI's size — 20 million creators and a 2.0 studio — is the asset that justified a nine-figure check.
Still, the economics of reselling model inference to a price-sensitive creator base are thin unless the community drives enough volume and retention to spread compute costs. The next test is not the headline raise but whether monthly active creators and paid workflows climb after the funding.
Competition is already crowded
LiblibAI is not alone. China has several creation communities and model-hosting platforms vying for the same illustrators, e-commerce studios, and IP teams. The differentiator is usually not the underlying models — most aggregate the same open and licensed ones — but the density of shared assets, the ease of LoRA training, and the size of the audience a creator can reach by publishing. A nine-figure raise buys time and compute subsidies, but it does not automatically buy the network effects that keep creators from wandering to a rival.
Honest limitations
Facts come from Chinese financial and tech press (Economic Information Daily, Tencent News) reporting the company's announcement; we did not see the term sheet. The "largest AI-application financing in China in 2025" label is the outlets' framing, not an independently audited ranking. User and creator counts are the company's self-reported figures. We did not verify revenue, valuation, or the identities of all follow-on investors beyond those named in coverage. We also did not confirm exact overseas user numbers, which were described only as a goal.
What readers can do now
- Illustrators and e-commerce creators should test LiblibAI's 2.0 studio as an alternative to stitching together separate image, video, and LoRA tools.
- Founders should note the shift in Chinese AI funding toward application-layer communities with real creator bases, not just model papers.
- Investors should watch whether LiblibAI converts the round into measurable overseas users rather than only domestic community growth.