SoftBank’s Son says AI will need $5 trillion per year by 2040, dismisses bubble talk

SoftBank CEO Masayoshi Son said AI will require about $5 trillion of investment per year through 2040 (≈800 trillion yen), dismissing “bubble” concerns as “absurd.” He also expects OpenAI-related spending to bring SoftBank’s cumulative investment to over $60 billion by end-2026, while estimating AI data centers may need ~3 terawatts of power by 2040—about 1.8x current global consumption. Son’s bullish outlook on AI demand and infrastructure buildout supports risk-on sentiment toward AI platform spend despite capex-driven skepticism.

Analysis

This is less a product thesis than a capital-allocation signal: SoftBank is positioning itself as a levered upstream toll collector on AI infrastructure, and that matters because the first price action should be in funding-sensitive assets, not in end-demand names. SFTBY/SOBKY can benefit if the market concludes the company has a credible path to monetize its private AI marks and recycle balance sheet capacity into the next round of scarce assets; the real optionality is in NAV re-rating, not near-term earnings.

Second-order winners are the picks-and-shovels exposed to power, data-center density, and compute scarcity, while the losers are businesses whose AI story depends on proving monetization at the edge. AAPL is not an obvious direct beneficiary: if agent-centric workflows become real, value shifts upstream to cloud and model owners, and the handset only keeps a premium multiple if Apple can show durable on-device AI differentiation. BABA is a cheaper but messier China proxy for the same theme, with cloud optionality offset by macro and policy discount.

The key risk is that this narrative requires continuous financing, not just enthusiasm. Over the next 1-3 months, watch for OpenAI funding cadence, any SoftBank asset-mark revisions, and whether AI capex actually translates into revenue line items; over 6-18 months, the bottleneck becomes power and cost of capital, which can starve marginal projects. Consensus is probably underestimating how fragile the private-market valuation chain is if growth decelerates even modestly.

Key Decisions for Investors

• Buy SFTBY/SOBKY on a 1-2 week pullback after the event-driven move; target 1-3 month upside from continued AI-funding headlines, and exit if the stock retraces the full post-speech pop without new OpenAI or asset-sale catalysts.
• Pair trade: long SFTBY vs short AAPL for 3-6 months to express 'AI capital formation' over 'AI endpoint monetization'; the trade is invalidated if Apple delivers two straight quarters of credible AI-driven services acceleration or handset upgrade re-acceleration.
• Treat BABA as a watchlist expression of China AI optionality rather than an outright buy; initiate only if Alibaba Cloud growth or domestic AI spending data start to inflect, otherwise the policy discount likely overwhelms the theme.
• No standalone chase in TSTS or broad AI software names from this item alone; the message is about financing and infrastructure scarcity, so any equity expression should be concentrated where balance-sheet optionality is highest.