Introduction

On May 17, 2025, SoftBank announced its new “Edge AI Core” platform—an open reference architecture for sovereign, decentralized inference. Built in collaboration with ARM Japan, EdgeCortix, and Renesas, the initiative targets municipalities, telecoms, and industrial integrators deploying AI at scale without relying on hyperscale cloud providers.

“AI must be embedded where decisions are made—not 1,000 miles away,” said Junichi Miyakawa, President of SoftBank Corp.¹. The architecture integrates ARM-based AI SoCs, secure mesh networking, and real-time inferencing stacks optimized for smart infrastructure and autonomous robotics.

Edge AI Core also introduces Japan’s first domestically standardized LLM-serving API for local environments, with hardware designed for passively cooled operation in street cabinets and rail networks.

Why it matters now

• Decentralized AI aligns with privacy mandates and smart infrastructure scale-out.
• Japan and the EU are building local AI stacks to reduce cloud and U.S. vendor dependency.
• Telecoms see AI edge cores as critical to monetizing 5G and IoT deployments.

Call-out: The sovereign AI edge is no longer optional—it’s policy

Edge AI Core represents a strategic response to cloud bottlenecks, data sovereignty, and latency in urban automation.

Business implications

Regional governments, rail operators, and energy providers can now adopt a national-standardized edge AI stack with validated hardware-software integration. Vendors aligned with SoftBank’s model may gain regulatory preference in public-sector procurement.

Private 5G builders can use Edge AI Core to launch inference-native microcells for security, transport, and event detection. Security leaders should note that the design includes on-device audit trails and policy enforcement for model outputs.

Looking ahead

SoftBank will begin public deployments in Fukuoka and Osaka by late 2025. A developer SDK and white-box router kit will ship in Q3. Partners expect deployments in Singapore, Germany, and the UAE by early 2026.

Gartner projects that by 2030, over 40% of city-scale AI inference will run on sovereign, non-cloud architectures.

The upshot: SoftBank’s Edge AI Core is more than an architecture—it’s a geopolitical statement. In the future of disruption, AI won’t live in the cloud. It will live in the crosswalk, the elevator, the substation—and the edge will rule.

––––––––––––––––––––––––––––
¹ Junichi Miyakawa, SoftBank AI Strategy Briefing, May 17, 2025.