Why is SoftBank stock sliding today?

SoftBank shares fell 3.2% to ¥6,367 after Arm dropped nearly 6% in the U.S. following an HSBC downgrade to Hold from Buy, citing that much of Arm’s AI upside is already priced in and that constrained TSMC production limits shipments of AI server CPUs. The selloff also follows skepticism toward Masayoshi Son’s Softbank World 2026 claims (e.g., $5T annual AI infrastructure spend by 2040 and AI revenue at 20% of global GDP), alongside an overhang from the OpenAI IPO delay—SoftBank’s cumulative OpenAI investment is expected to exceed $60B by end-2026, with liquidity timing now pushed toward 2027.

Analysis

This is a de-rating event more than an earnings event. ARM has been treated as the purest liquid proxy for AI unit growth; once investors conclude the near-term growth curve is bottlenecked by foundry capacity and already embedded in the multiple, the stock stops trading on TAM and starts trading on delivery cadence. That matters most for SoftBank, because the holding-company discount widens when the market loses confidence in the timing of monetization and the asset stack becomes less liquid on the path to realization.

The second-order winner is likely TSM, not because demand disappears, but because constrained AI capacity preserves pricing power and utilization at advanced nodes. The loser set is broader than ARM: any AI-duration names priced on far-dated optionality, especially those with limited near-term cash conversion, can see multiple compression as the market rotates toward businesses with visible shipments and FCF. If the market starts to question whether AI infrastructure spend is being pulled forward faster than it can be manufactured, the first casualties are the highest beta enablers rather than the actual capex beneficiaries.

Catalysts are mostly 1-3 months: ARM/SoftBank commentary, any update on OpenAI IPO timing, and evidence that TSM can actually relax the bottleneck. The contrarian view is that the selloff may still be incomplete because the consensus is treating the IPO delay as a timing issue, when it is really a balance-sheet and liquidity issue for SoftBank’s capital recycling model. What would falsify the bearish view: a near-term ARM guide-up with accelerating royalties, or a concrete OpenAI liquidity event that restores the monetization narrative before the market assigns a larger holding-company discount.

Key Decisions for Investors

• Short SFTBY/SOBKY on any bounce over the next 1-3 weeks; target underperformance versus Japanese tech benchmarks as the market reprices delayed liquidity and a wider holding-company discount. Falsifier: a credible OpenAI IPO/secondary timeline announced before quarter-end.
• Pair trade: long TSM / short ARM for the next earnings cycle. Thesis is that foundry scarcity monetizes better than royalty-duration once AI capacity is the bottleneck; risk/reward favors TSM if advanced-node pricing stays tight. Falsifier: ARM reports upside from design wins without capacity constraints binding.
• Buy 1-2 month ARM put spreads 5-10% below spot after any intraday relief rally. This is a multiple-compression trade, not a fundamentals blowup, so defined-risk options are preferable to outright shorting. Falsifier: ARM re-accelerates with guide-up evidence on AI server CPU shipments.
• Do not chase the AI-beta basket indiscriminately; rotate from ARM-like duration exposures into supply-chain bottlenecks and cash-generative semis. If the market starts rewarding visibility over story, the relative winner set shifts toward TSM over the next 3-6 months.