Masayoshi Son: Annual AI Investment Demand to Reach $5 Trillion by 2040
N.R. Finch
SoftBank's Masayoshi Son told the company's annual meeting that AI will require $5 trillion a year by 2040 and dismissed bubble talk as "absurd" — but offered no methodology behind the figure.
$5 trillion a year — where does the number come from?
Son said AI development will need $5 trillion annually (roughly ¥800 trillion) by 2040.
He predicted AI revenue will reach 20% of global GDP by then, making ¥800 trillion "a rounding error."
He disclosed no methodology — no model breakdown, no third-party source.
This means → the figure reads more as a vision statement than a verifiable financial forecast.
How much power would AI consume?
Son forecast that AI data centers will require 3 terawatts of generation capacity by 2040 — a terawatt is a unit of large-scale power output, equal to one billion kilowatts.
In plain terms = that is 1.8 times current global power generation — AI alone would need nearly two "entire worlds" of electricity.
His proposed path: natural gas in the near term, then a shift to nuclear fusion.
This reflects a tension: fusion has no commercial track record, yet the back half of the energy plan depends on it.
What does an "agent-centric world" actually mean?
Son's 2040 vision: 100 trillion AI agents — autonomous programs that decide, act, and communicate with each other — will replace humans as the central actors.
His words: "The era in which humans are the highest form of life on Earth will end — for better or worse, this will happen and cannot be stopped."
This means → he is positioning AI not as a tool but as an autonomous species — one of the most aggressive framings from any tech leader today.
How much has SoftBank actually committed?
SoftBank's cumulative investment in OpenAI is expected to exceed $60 billion by the end of 2026.
The group has also deployed tens of billions more into data-center financing and robotics.
In plain terms = Son is not just talking numbers — SoftBank has put real capital behind the thesis, which is both a sign of conviction and a reason he needs the market to buy the story.
Bubble fears vs. Son's conviction — who is right?
Son called any AI-bubble discussion "absurd," arguing that AI revenue will ultimately cover the outlay.
The market's concern is specific: infrastructure capex is ballooning, and whether returns can cover costs remains unanswered.
Son's own track record cuts both ways — an early bet on Alibaba paid off spectacularly, but WeWork produced heavy losses.
This reflects the core divide in AI investing today: not "is AI useful?" but "at the current price of entry, can the money be made back?"
Content is for reference only, not financial advice.