SoftBank plans to list AI firm Roze in the US at up to $100 bln valuation- FT

SoftBank is reportedly planning to create and list a U.S.-based AI and robotics company, Roze, with a target valuation of up to $100 billion. The planned business would focus on data centers and could come to market as early as this year, underscoring Masayoshi Son’s aggressive push into AI and related infrastructure. The article is speculative and not finalized, but it signals another major AI-related capital markets event.

Analysis

The market is starting to re-rate the AI stack from “model winners” to “infrastructure toll collectors,” and this potential Roze listing would extend that trade into a new, more speculative leg: private-capital-to-public-market arbitrage. If SoftBank can float a dedicated AI/robotics/data-center vehicle near a $100B mark, it creates a visible comping event for every adjacent platform asset and could pull forward capital into power, land, fiber, and liquid cooling suppliers before revenue is fully proven. The second-order effect is that scarcity value may shift from software margins to physical bottlenecks — grid access, permitting, and high-density deployment capacity become the scarce inputs.

For SoftBank, the strategic signal matters more than the listing itself: this is effectively a financing machine wrapped around an AI narrative, which can lower the group’s cost of capital if the IPO is well received, but also concentrates execution risk into a single theme. The main vulnerability is that public markets may not pay up for a “build-first, monetize-later” asset if capex intensity rises faster than order visibility; that tension is likely to show up over the next 3-9 months as the market stress-tests assumptions around utilization, power availability, and customer concentration.

The broader winners are likely to be picks-and-shovels beneficiaries rather than the marquee AI names: electrical equipment, thermal management, and datacenter REITs can see order book acceleration if Roze catalyzes a wave of follow-on projects. The losers are late-stage private AI firms that rely on scarcity valuations; a successful listing could widen the gap between assets with hard infrastructure exposure and those with pure software/AI narrative. The contrarian take is that the IPO could become a top signal for froth if it prices aggressively while the underlying economics remain opaque — that would be a warning that the AI capex cycle is nearing a financing peak rather than a durable growth phase.

Key Decisions for Investors

• Short-term: buy call spreads on SMCI into any IPO-roadshow confirmation; if Roze validates a new AI capex leg, compute and cooling names can re-rate quickly over 1-3 months, but cap the premium because sentiment can reverse on any funding hiccup.
• Pair trade: long VRT / short a basket of high-multiple unprofitable AI software proxies over the next 2-4 months; the market is likely to reward tangible data-center exposure and punish narrative-only names if capital starts reallocating toward infrastructure.
• Add to EQIX and DLR on pullbacks with a 6-12 month horizon; a successful AI infrastructure IPO should improve leasing demand and pricing power, though the trade should be sized for rate sensitivity and multiple compression risk.
• Watch 9984 as a sentiment barometer rather than a pure fundamental trade; if the stock rallies on IPO optionality, fade strength after the first pop unless management provides clear de-leveraging mechanics, since the market may be underestimating execution and funding dilution.
• If Roze files within the next quarter, consider a tactical long on VST or NRG as a secondary beneficiary of higher datacenter power demand; risk/reward improves only if the market begins pricing incremental electricity load growth rather than treating this as another hype cycle.