📰 Key Summary

SoftBank Group founder and chairman Masayoshi Son stated at the SoftBank World 2026 conference in Tokyo on July 14, 2026, that global AI infrastructure will require $5 trillion in annual investment by 2040. He pointed out that the expansion of data centers, power supply, and humanoid robots will support AI’s shift from a human-centered work model to a new type of labor structure. Son expects AI to account for 20% of global GDP by then, with AI agents and humanoid robots becoming the key forces shaping the future workforce. See the original article link for full details.

💬 JudyAI Lab Perspective

The number Son dropped at SoftBank World 2026 is pretty blunt: AI infrastructure needs to burn $5 trillion a year by 2040 — that’s not “investment” anymore, that’s rebuilding the entire global power and computing foundation from scratch.

For AI builders, the real signal isn’t that huge $5 trillion figure — it’s that he tied three things together: data centers, power supply, and humanoid robots. That tells you industry consensus has already shifted from a “model capability race” to a “physical-layer bottleneck race.” When AI is supposed to make up 20% of global GDP, the driving force won’t be smarter models — it’ll be whether agents and robots can actually take over real-world labor tasks. For people building AI products, here’s the wake-up call: guardrails aren’t just API rate limits anymore. Power, hardware, deployment scenarios — stuff that used to be “the infra team’s problem” — are now variables you have to bake into product design from day one.

Next time you’re sizing up an AI product’s long-term positioning, try asking one extra question: can the compute and power assumptions this product relies on actually hold up until agents truly deploy at scale?

📅 Source Info

🔗 Further Reading