Son Bets €75B on French AI Data Centers, ¥1,000T NAV Goal

A profile in the Financial Times captures something the announcements have been signalling for a year: Masayoshi Son has bet SoftBank on AI infrastructure at a scale no peer is attempting, and the numbers are worth reading slowly. At May's Choose France summit hosted by Emmanuel Macron, SoftBank committed up to €75 billion to build 5 GW of AI data center capacity in France, with a €45 billion first phase for 3.1 GW in the Hauts-de-France region and an industrial cluster in Dunkirk co-developed with Schneider Electric, as CNBC reported.

Behind the French headline is the frame Son laid out at SoftBank's shareholders' meeting. He set a ¥1,000 trillion net asset value target by 2042, roughly a 14-fold expansion over sixteen years, spread across four pillars he named as AI models, robotics, semiconductors, and AI infrastructure. He is 68, dismissed talk of retirement, and told shareholders that bubble fears were "an insult to AI." SoftBank already owns about 13% of OpenAI and is publicly committed to the Stargate build-out.

Why this matters if you are pricing European AI capacity, GPU allocation, or sovereign AI politics: Son has effectively become the single largest private counterparty across that whole stack. When Macron talks about French AI sovereignty, he is now talking about a Japanese balance sheet. When OpenAI talks about compute, one of its largest shareholders is also building the roof. The concentration cuts both ways. Lenders have already shown hesitation; Yahoo Finance reported that a $6 billion margin loan against the OpenAI stake hit a stalemate, and SoftBank's shares still trade at a material discount to reported net asset value, which is the market's way of saying it is not sure any of this compounds the way Son claims.

The honest caveat is that a lot of this is announcement, not execution. The reporting does not give you the financing split between SoftBank equity and project debt on the €45 billion first phase, or where 5 GW of French power will actually come from, or whether the ¥1,000 trillion target embeds a specific OpenAI valuation. Take the specifics as reported, not settled.

What is worth watching is who benefits if he is even directionally right. Schneider Electric and the French industrial supply chain around Dunkirk get anchor orders. Macron gets an EU AI sovereignty story that could pull follow-on capital from European pension and sovereign funds. OpenAI gets a committed non-US infrastructure partner it can point to in every hyperscaler negotiation from here on. Son's bet is that AI infrastructure is undersupplied for a decade; the interesting part is that he is the one putting the money down first.