Maase Signs Deal to Raise About $50 Million for AI Computing and Its Language Model

Maase has signed an agreement to sell shares for approximately $50 million in expected gross proceeds. The company intends to put most of the money toward its distributed AI computing centers and Lingyan language model, linking two existing projects to a financing that has yet to close.

The share sale is agreed, not completed

The private investment in public equity, or PIPE, covers 3,878,856 Class A ordinary shares at $12.89 apiece. Maase describes the buyer as a financial investment institution focused on AI infrastructure but does not name it. The shares are subject to a 36-month lock-up, restricting the investor from selling them during that period.

Maase expects the transaction to close in October 2026, subject to customary conditions. The $50 million is expected gross proceeds, not money already received or a stated amount available after transaction costs. The planned spending therefore depends first on the financing closing.

Building out the computing network

For its Star distributed computing centers, Maase expects to buy and deploy modular computing units housed in containers. It also plans upgrades to the platform that schedules computing capacity across the network, plus related green-energy and energy-storage work. The company says these investments would help it fulfill existing computing-services contracts and serve additional enterprise demand.

That spending would build on the Stars project Maase announced with partners in April. At the time, it described a five-phase rollout over 60 months and total planned investment of up to RMB 5 billion. Maase also cautioned that the project's implementation, investment and cooperation arrangements required further negotiations and definitive agreements. The new share sale is a financing plan for Maase, not confirmation that the full project investment has been committed.

The model has an enterprise path

On the model side, Maase plans to improve Lingyan's security and privacy-preserving computing capabilities. Its spending priorities also include versions enterprises can deploy privately, a consumer access portal, datasets and an AI security laboratory. Those efforts are intended to support AI token services and customized language-model projects. Maase has not specified what share of the proposed funding will go to any one of them.

There is already a named enterprise project behind that ambition. In July, Maase said its subsidiary Huazhi Future had agreed to develop a customized AI application for Zhongchuang Liankong using Lingyan Miaoyu. The agreement covers model customization, data work, systems integration and secure deployment. It provides an example of the tailored work Maase wants to support; it does not establish how much of this financing will be spent on that customer project.

Maase says the remainder of the proceeds would support market expansion, research hiring and delivery of existing AI computing and enterprise projects. It has not given a dollar split between its computing centers, Lingyan and those other uses. That leaves the balance between building capacity and developing services open.