AIToday
AI Regulation & PolicyAI Business & IndustryASCII.jpPublished: Nov 21, 2026, 10:01 JST

On-premise AI LLM Link joins Japan's AI adoption subsidy list

Send on LINE
On-premise AI LLM Link joins Japan's AI adoption subsidy list

3 Key Points

  1. What happened

    coiai's LLM Link, which runs generative AI inside a company's own servers so internal documents never leave the premises, was registered as an eligible IT tool under the Digitalization and AI Adoption Subsidy, with coiai as a registered IT adoption support provider.

  2. Why it matters

    Small and midsize firms weighing on-premise AI against sending sensitive data to cloud services can now apply for the subsidy with coiai handling the paperwork, potentially easing the upfront cost barrier the body cites.

WHO IT HITSSmall and midsize business owners and IT managers handling confidential documents, drawings, or customer data gain a subsidized path to running AI in-house, while coiai takes on application, setup, training, and maintenance.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The registration addresses two barriers the announcement highlights for smaller Japanese firms: reluctance to send confidential material such as internal documents, drawings, and customer data to external cloud AI services, and the risk of "shadow AI" where employees input company information into outside tools on their own. LLM Link is positioned so that AI processing stays entirely within the company's environment, and its permission-linked RAG feature means answers draw only on documents the asking user is allowed to see.

The second barrier is cost. By becoming a registered tool under the Digitalization and AI Adoption Subsidy, LLM Link lets adopting companies pursue on-premise AI while using the subsidy, with coiai — itself a registered IT adoption support provider — handling the application through to deployment, user training, and remote maintenance. The announcement cautions that subsidy rates and caps depend on the application category and deployment details, and that contracts or payments made before a grant decision are not eligible.

On pricing, coiai lists a standard configuration: software at 715,000円 as a perpetual license for one company and one environment with unlimited accounts, setup and training at 770,000円 per installation, support at 143,000円 per year, and the AI server quoted separately with a standard estimate of about 1 million yen.

FAQ
What does LLM Link do?
It is an on-premise AI application that runs a large language model on a company's own servers or internal network, so internal documents and business data are not sent to external AI services. It also offers permission-linked RAG, meaning document search tied to each user's or group's access rights.
How much does it cost?
The software is 715,000円 (tax included, perpetual, one company and one environment, unlimited accounts). Setup, manual creation, and training are 770,000円 per installation, maintenance support is 143,000円 per year, and the AI server is quoted separately, with a standard configuration estimated at about 1 million yen.
Can any company get the subsidy?
The subsidy is for small and midsize businesses adopting registered IT tools, applied for jointly with a registered IT adoption support provider. Subsidy rates and caps vary by application category and content, and contracts or payments made before a grant decision are not covered.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleDell at New York Symposium: AI moat is your own data, not models