LibLib AI hits $2B valuation with $300M Series B extension as ARR reaches $300M
What's the deal? LibLib, the Chinese AI application maker behind image tool LiblibAI, has raised US$194.1M Series B extension, its parent EvokenDealroom has a profile for this one. Try Dealroom → announced in June 2026. The round values the company at $2 billion, one of the largest single rounds in China's AI application layer to date.
The numbers: LibLib's annual recurring revenue reached $300 million as of May 2026, with group revenue in that month up more than 3,000% year on year. The valuation marks a sharp step-up from its previous round.
What's the endgame? LibLib does not build its own models. Instead, it packages third-party large models into its own products, adding service fees and premium features on top of compute costs. Founder Chen Mian compares models to a camera and his products to the photographer who gets professional results from it.
The products: The company runs three lines. LiblibAI serves Chinese creators, with 30 million users and 500,000 original LoRA fine-tuned models. Overseas design tool LovartDealroom has a profile for this one. Try Dealroom → passed $80 million in ARR five months after launch, while short-drama platform LibTVDealroom has a profile for this one. Try Dealroom → saw revenue jump 13-fold in its first two months.
Why now? AI investment logic shifted in the first half of 2026, according to an ITjuzi report, as capital moved from model-parameter bets toward applications that can be deployed and monetised. LibLib is among the biggest beneficiaries of that turn.
What could go wrong? Critics argue LibLib is little more than a token middleman, and some suspect it resells API access — a claim Chen denies. Without its own model, its pricing power depends on upstream suppliers such as ByteDance's Volcano EngineDealroom has a profile for this one. Try Dealroom →, which could adjust API prices or launch competing apps. Chen has said application-layer margins should stay below 30% in the near term.
Founder view: Chen, a former ByteDance executive who led global commercialisation for CapCutDealroom has a profile for this one. Try Dealroom →, left in 2023 rather than compete head-on with model builders. "I know what happens when you go up against giants," he said. The company nearly failed in early 2024, once down to US$550 in the bank, before turning to a lighter, revenue-first model.
The signal: LibLib's rise is a test of whether the AI application layer can build lasting value on top of models it does not own. Its speed to revenue has bought it a lead — but whether that becomes a durable moat, or simply captures a passing window of model dividends, remains unproven.
Read more: 36Kr
Image credit: Generated with Gemini
