In a July 16, 2025 address, SoftBank CEO Masayoshi Son announced a bold internal initiative: the deployment of 1 billion AI agents across SoftBank Group by year-end. Each employee is expected to operate around 1,000 agents, effectively automating tasks such as email replies, scheduling, system checks, and customer support.
✅ Why This Matters
- Mass Automation: These agents will autonomously execute common knowledge work—akin to transforming each employee into a multi-armed deity of multitasking.
- Cost Efficiency: Estimated cost is only ¥40 ($0.27) per agent per month, totaling about $3.2 billion annually for the full deployment.
- Productivity Boost: Son expects agent productivity to be at least four times higher than human-level performance
- Strategic AI Backbone: This effort ties into SoftBank’s broader AI vision—including its $3 billion/year deployment of OpenAI’s Cristal Intelligence and the $500 billion Stargate infrastructure plan.
📊 Key Features & Scale Expectations
| Feature | Description |
|---|---|
| Agent Count Target | 1 billion AI agents within SoftBank Group |
| Per Employee Deployment | Roughly 1,000 agents each |
| Annual Cost | ~$3.2 billion (¥40/month/agent) |
| Agent Capabilities | Email response, system checks, negotiations, etc. |
| Agent OS & Coordination | New soft network to support interacting agents |
🔭 Broader Context
- Enterprise Rollout via OpenAI: Launched alongside SB OpenAI Japan in early 2025, SoftBank is deploying “Cristal intelligence” across its enterprises—automating over 100 million workflows.
- Infrastructure Support: SoftBank and OpenAI’s joint venture Stargate is building the AI data center backbone required to scale billions of agents.
- Chip & Compute Strategy: Complemented by SoftBank’s acquisition of Ampere Computing for efficient Arm-based AI chips and its investment in Homegrown LLMs.
🔍 What’s Next?
- Agent Ecosystem Build-Out: SoftBank must develop agent-creation tools and an operating environment to manage agent interaction; execution remains early-stage with many technical challenges.
- Pilot Scaling Phases: Initial phase will focus on internal teams—workflows in support, research, IT, and administrative functions.
- Global Model: Success in Japan could scale globally, as SoftBank builds an AI agent infrastructure integrated with enterprise systems worldwide.
✅ Final Take
SoftBank’s plan to deploy 1 billion AI agents by the end of 2025 is one of the most ambitious AI deployments ever. At just a few cents per agent each month, the initiative signals a transformative shift toward automation, built on SoftBank’s investments in OpenAI’s Cristal intelligence, massive data centers via Stargate, and its own AI infrastructure. For employees and enterprises alike, this could redefine productivity.
