SoftBank Group's market value pushed above ¥40 trillion (about $252 billion) on Monday as shares rose 4.6% to an all-time high in Tokyo, capping a roughly 40% rally since May 20 that has been driven almost entirely by the looming public-market debuts of two of its AI bets: OpenAI and power-infrastructure arm SB Energy.
The move makes SoftBank the single largest public beneficiary of the current AI capital cycle — and the cleanest liquid proxy for a trade that retail and institutional investors otherwise cannot access directly.
The two IPO catalysts
SB Energy — jointly owned by SoftBank and Ares Management — made a confidential US IPO filing on Wednesday, May 20, lighting the fuse. Reports then indicated OpenAI is preparing to file confidentially for its own US listing, working with Goldman Sachs and Morgan Stanley, with a public debut targeted for the fall. CNBC reported SoftBank added more than $61 billion in market cap across the two trading days through May 22 alone, with intraday gains approaching 20% earlier in the run before Monday's more measured 4.6% advance carried it to a record close.
Why SoftBank is the cleanest AI proxy
SoftBank has committed more than $60 billion to OpenAI to date, and its most recent quarter showed profit jumping roughly threefold on the back of that stake's mark-to-market gains. Because public-market participants can't buy OpenAI or Anthropic shares directly, SoftBank — which also controls Arm Holdings, the dominant licensor of CPU IP now central to AI inference roadmaps — has become the most liquid instrument for expressing the AI infrastructure thesis on a public exchange.
That is a double-edged setup. The bulk of the rally rests on paper gains pegged to private valuations that have not yet been tested by an actual offering. An OpenAI IPO would convert those marks into a market-clearing price for the first time — validating or repricing the entire portfolio in a single print.
What it signals for the infrastructure layer
The SB Energy filing is the detail builders and enterprise buyers should track most closely. It confirms SoftBank's pivot from purely funding model labs toward vertically integrating into the power and data-center backbone those models depend on. Energy availability — not chips alone — is now the binding constraint on frontier-scale compute, and a publicly traded SB Energy would give the market a dedicated vehicle to price AI power capacity.
For anyone provisioning large-scale training or inference, the read-through is concrete: the capital chasing AI is increasingly flowing into generation and grid infrastructure, not just GPUs. A successful pair of SoftBank-linked IPOs this fall would further normalize the multi-hundred-billion-dollar valuations underpinning the buildout — and tie the cost and availability of enterprise compute ever more tightly to public equity sentiment.
