Hefei-Backed AI Unicorn HiDream.ai Raises $290M in Three Months, Signals Multimodal Shift

On July 23, HiDream.ai announced a 1.5 billion yuan Series C round, the third funding round the multimodal AI company has closed within three months. With over 2.1 billion yuan raised across the three rounds, HiDream.ai is now valued at over $1 billion, officially becoming a global AI unicorn.

The investor lineup reads like a who’s who of Chinese capital. The round was co-led by the Sichuan Revitalization Innovation Fund under the national social security fund, ICBC Capital, Hongyi Asset Management, and Dunhong Capital. Follow-on investors included Xiamen C&D Capital, Shanghai Film Group’s New Horizons Fund, Hubei Yangtze River Industrial Investment Group, Huace Film & TV, and multiple other institutional investors. Existing backers including Hefei Industrial Investment, Oriental Fortune, and Jinpu Investment continued to add capital.

This marks the first time the national social security fund has invested as an LP in a multimodal AI company, and the first time ICBC Capital has backed a multimodal AI enterprise. The rare convergence of capital with fundamentally different investment theses — national strategic funds, provincial state capital, film industry money, and top-tier market-driven VCs — all betting on the same startup signals a major shift in China’s AI investment landscape.

The funding comes at a pivotal inflection point for the AI industry. Large language models have dominated capital attention through code generation and text interaction, but recent waves of open-source model releases have eroded LLM valuation premiums. Meanwhile, AI vision and multimodal capabilities are experiencing an explosion in capital concentration, with over 30 billion yuan flowing into China’s AI vision sector in 2026 alone.

Technically, HiDream.ai has bet on a native full-modal architecture starting from visual pixels. Its UiT architecture unifies image pixels, text tokens, video voxels, audio, motion, and spatial relationships into a single Transformer for understanding, generation, and reasoning — eliminating independent text encoders and modality bridges.

This approach has been validated by third-party benchmarks. In May 2026, HiDream-O1-Image (8B open-source version) anonymously topped the text-to-image leaderboard on independent platform Artificial Analysis, ranking first among open-source models. Its commercial version, HiDream-O1-Image-1.5, later became the highest-ranked Chinese image model globally, placing in the top three.

On the commercial front, HiDream.ai has built a “1+1+3” system: the HiDream full-modal foundation model at the bottom, the HiHarness enterprise capability platform in the middle, and three vertical applications — marketing, film production, and social media content. Its vivago R1 multimodal creation agent, launched at WAIC, supports unlimited-length autonomous video generation with long-chain narrative planning, achieving an 85% commercial content success rate. The international version of vivago has covered over 100 countries and served more than 50 million users.

Hefei’s role in this story is significant. The city, which previously made landmark investments in ChangXin Memory Technologies and BOE Technology, has once again placed a major bet during the formative stage of a strategic technology sector. HiDream.ai now stands as a key test case for whether China’s native multimodal architecture approach can compete on the global stage.