APOLLO11 launches EdgeLLM, a fixed-price on-premise generative AI platform for Japan's compliance-bound industries.
The AMW Read
A new on-premise AI infrastructure entrant packaging open-weight models (Qwen, DeepSeek) into a fixed-price, compliance-driven product for Japan's regulated sectors — an incremental new player within an already-recognized deployment pattern, with impact confined to a single sub-segment.
APOLLO11 launches EdgeLLM, a fixed-price on-premise generative AI platform for Japan's compliance-bound industries.
APOLLO11, a Nagoya-based AI infrastructure provider founded in 2011, began selling EdgeLLM on August 31, an on-premise generative AI platform that runs inside a customer's internal network with zero external data transmission, including support for air-gapped deployments. Pricing starts at ¥9.8 million (~$66,000) for the base "Vision Pack S" tier (about five concurrent users), with "Agent Pack M" and "High-Context Pack" tiers at ¥19.8 million (~$132,000) each. Standard rollout takes five to six weeks, billing is fixed-cost with no usage-based charges, and each node runs on roughly 240W of power. The platform serves open-weight models such as Qwen and DeepSeek through an OpenAI-compatible API, so customers can port existing LangChain integrations with minimal rework.
The target buyers are law firms, hospitals, and government agencies in Japan that are barred by attorney confidentiality rules, personal-data-protection law, or internal security policy from sending data to external cloud APIs — the same constraint APOLLO11 says is driving unauthorized "shadow AI" use inside these organizations. The pitch depends on open-weight models having closed enough of the gap with proprietary frontier systems that a compliance-first, fixed-price on-premise deployment is now commercially viable for organizations that previously couldn't justify building their own LLM infrastructure.
For vendors selling AI into regulated Japanese verticals, the more durable signal here is the packaging, not the headline price: fixed hardware SKUs, a standard-LAN-only architecture positioned to avoid triggering cross-border transfer rules under Japan's personal-data law, and quarterly model refreshes to track open-weight releases. APOLLO11 says it is targeting ten deployments across legal, healthcare, and government by the end of 2026 — a small number that will test whether compliance-driven on-premise AI can scale past pilot volume in a market still mostly served by cloud APIs.
#AIInfrastructure #OnPremiseAI #OpenWeightModels #Japan #EnterpriseAI #DataPrivacy