SoftBank's Son, Sam Altman see AI demand surge, eye more compute

SoftBank Group Corp. founder Masayoshi Son and OpenAI CEO Sam Altman see an insatiable global demand for AI, making it imperative to keep building more computing capacity.

In a teleconference at the SoftBank World event, they emphasized the need to scale infrastructure.

Altman noted that lowering the cost of AI drives higher usage. Son shared plans to deploy one billion AI agents this year and develop an operating system for them.

The two also talked about robotics, with Altman highlighting the potential for robots to build other robots.

Son mentioned their joint venture to promote AI in Japan for industries like automotive and retail.

🔗 Source: Bloomberg

🧠 Food for thought

1️⃣ AI’s insatiable demand faces physical infrastructure constraints

The optimistic vision shared by Son and Altman overlooks significant physical challenges that could limit AI expansion.

Power demands from AI data centers in the U.S. alone are projected to grow from 4 gigawatts in 2024 to 123 gigawatts by 2035, representing an enormous strain on existing electrical infrastructure1.

AI facilities require substantially more energy per square foot than traditional data centers, with some advanced facilities potentially needing up to 2,000 megawatts—equivalent to powering small cities1.

A survey of power and data center executives confirms this challenge, with 79% identifying grid stress as the primary obstacle for AI infrastructure development1.

Beyond electricity, thermal management has become a critical bottleneck, as excess heat from densely packed AI chips can lead to system failures, requiring sophisticated cooling solutions that often underperform in real-world environments2.

These infrastructure limitations explain why Son and Altman are focusing so heavily on massive projects like the $500 billion Stargate initiative to build new data centers. The current infrastructure cannot support their scaling ambitions.

2️⃣ Economic data supports Altman’s counterintuitive demand elasticity claim

Altman’s assertion that making AI “10 times cheaper” leads to “30 times” more usage reflects a well-documented economic pattern in technology adoption.