SoftBank Group CEO Masayoshi Son said global artificial intelligence will demand $5 trillion in yearly investment by 2040, adding that talk of a bubble is “absurd.” He delivered the forecast on Tuesday, arguing that spending on chips, data centers, power, and software must scale quickly to meet demand. The statement puts a bold marker on the next phase of tech spending and sets up a debate over how fast the sector should grow.
Son’s Case for Massive AI Spending
Son framed AI as an infrastructure shift on the scale of electricity or the internet. He said the next 15 years will require sustained funding to build capacity and enable new services. He dismissed fears of an investment bubble, saying the market has not yet met demand for compute and talent.
“The development of AI will require investment of $5 trillion each year by 2040, and any talk of a bubble forming around the technology is absurd,” said Masayoshi Son.
His view matches a surge in capital outlays by tech giants. Analysts project annual data center and AI chip spending in the hundreds of billions of dollars this decade. Power projects and network upgrades add to that bill. Son’s number suggests those budgets would scale many times over by the late 2030s.
Background: SoftBank’s AI Pivot
SoftBank has positioned itself as an AI-focused holding company. It controls Arm, the chip designer that went public in 2023 and supplies instruction sets for mobile and growing AI workloads. The firm’s Vision Fund investments have faced swings, yet Son has said the next wave of value will come from AI platforms and the compute that fuels them.
Global IT spending offers a reference point. Research firms estimated total worldwide IT outlays at roughly $5 trillion in 2024. Son’s projection suggests AI alone could match the size of the entire IT market within 15 years. That comparison explains both the optimism and the concern.
Where the Money Would Go
Industry discussions point to four main buckets for AI capital: semiconductor fabrication, data centers, electricity generation and grid upgrades, and software and services. The weakest links today are factory capacity for advanced chips and reliable power for large sites. New plants can cost $10 billion to $20 billion each, and take years to build. Utilities face long permit timelines and supply chain gaps for transformers and turbines.
- AI chips and fabs: advanced nodes, packaging, and supply diversification.
- Data centers: land, cooling, and high-speed networking.
- Power: generation, storage, and transmission build-outs.
- Software: models, safety tooling, and domain-specific apps.
Son’s number implies large, concurrent build-outs across each area. It also assumes steady advances in model performance and adoption across industries such as health, finance, and manufacturing.
Skeptics See Risks of Overbuild
Not everyone shares Son’s confidence. Some investors warn that model training cycles can cool, leaving unused capacity. Others point to high interest rates that raise the cost of megaprojects. Critics also highlight bottlenecks, including skilled labor shortages and local resistance to large facilities.
Environmental groups question the water and energy footprint of new sites. Regulators in the United States and Europe are reviewing data use, competition, and safety. Tighter rules could slow rollouts or increase compliance costs. These factors could make the path to $5 trillion per year uneven, even if demand stays solid.
Signals to Watch
Several indicators will test Son’s thesis. First, chip supply and pricing for top-tier accelerators will show whether demand stays ahead of supply. Second, utility plans for new generation and grid upgrades will signal if power constraints are easing. Third, unit economics for AI services must improve as inference, not training, drives usage at scale.
Corporate adoption is another test. Productivity gains, new revenue, and clear return on investment would support larger budgets. Weak returns would push buyers to delay upgrades or favor cheaper models.
Son’s forecast sets a high bar for the sector. It captures the scale of ambition and the build required to meet it. The next few years will reveal whether growth in chips, power, and real-world use can justify the pace of spending. For now, the debate over an AI bubble will track hard numbers: capacity shipped, power delivered, and cash flows from services built on top of that stack.
Deanna Ritchie is a managing editor at DevX. She has a degree in English Literature. She has written 2000+ articles on getting out of debt and mastering your finances. She has edited over 60,000 articles in her life. She has a passion for helping writers inspire others through their words. Deanna has also been an editor at Entrepreneur Magazine and ReadWrite.
