Softbank Group shares fall amid reports of steep OpenAI spending

SoftBank shares fell as much as 5% after reports showed OpenAI’s cash burn tripled year-on-year to $3.7 billion in Q1 2026, even as revenue rose to $5.7 billion. The Financial Times also reported OpenAI’s 2025 loss widened nearly eightfold to $39 billion on about $13 billion of revenue, while uncertainty over valuation and stalled financing weighed on sentiment. The article adds pressure to SoftBank’s heavily concentrated OpenAI exposure, despite the startup’s confidential IPO filing.

Analysis

The market is starting to price AI not as a single secular growth trade but as a capital-allocation stack where margin structure matters as much as model quality. That is a second-order negative for the highest-multiple “AI proxy” names: if frontier model economics are proving far worse than assumed, investors will demand clearer monetization paths from infrastructure and application layers, compressing the premium on anything perceived as a levered beta to AI enthusiasm.

The biggest beneficiary is likely not the obvious model leaders but vendors with usage-based, enterprise-embedded monetization and strong gross margin retention. That argues for relative outperformance in application software with sticky workflow penetration versus infrastructure beneficiaries whose growth depends on an ever-rising AI capex treadmill; the latter can still grow, but the market may stop paying for growth that is increasingly financed through dilution or balance-sheet stress. On the flipside, any company whose investment case depends on a clean IPO exit or mark-to-market uplift in private AI stakes now faces a longer duration mismatch: public market repricing can hit the exit window before revenue catches up.

Near term, the main catalyst is not fundamentals alone but financing conditions. If leverage against private AI holdings remains unavailable, forced de-risking can show up as secondary sales, delayed capital raises, or more conservative guidance from strategic investors over the next 1-2 quarters. That creates a squeeze risk for crowded AI longs and a setup for volatility selling or put structures, especially if the next leg of news flow confirms that revenue growth is still outrun by compute and inference costs.

The contrarian angle is that the selloff may be more about valuation and balance-sheet optics than a true break in AI adoption. If enterprise spending on AI keeps accelerating, the market could rotate from “pick-and-shovel” winners to companies with immediate ROI, and the current drawdown in the ecosystem may be a better entry point for quality names than for speculative one-product AI stories. The key is separating adoption durability from financing fragility; those are now diverging variables rather than one trade.

Key Decisions for Investors

• Long APP / short SMCI for 4-8 weeks: favor application-layer monetization over hardware exposure if AI capex discipline tightens; target a 10-15% relative spread with a stop if AI infrastructure re-accelerates on new hyperscaler guidance.
• Buy 1-2 month put spreads on high-beta AI proxy baskets or individual crowded names: use the next relief rally to express downside on financing-risk compression; structure for limited premium outlay with 2:1+ payoff if sentiment rolls over again.
• Rotate into enterprise software names with demonstrated AI attach rates and recurring revenue for a 3-6 month horizon: prefer businesses where AI improves retention and pricing rather than requiring external funding to grow.
• Avoid adding to private-market-linked AI holdings until IPO/financing windows clear: the risk-reward is asymmetric to the downside if margin loans, secondaries, or listing plans are delayed by 1-2 quarters.
• If holding profitable AI winners, trim 20-30% into strength and re-enter only on confirmation of margin stabilization: the market is likely to reward cash generation over headline growth until cost curves visibly improve.