SoftBank shares plunge as Google's Gemini sparks fears over OpenAI's competitiveness

SoftBank Group shares plunged to a 2½-month low, falling as much as 11% and closing down 9.95%, subtracting 338 points from the Nikkei amid concerns Google’s new Gemini 3 model undermines OpenAI — one of SoftBank’s largest investments. The hit comes despite SoftBank reporting ¥2.5 trillion (≈$16.6bn) net profit for Jul–Sep and Vision Fund gains of ¥3.5 trillion (including ¥2.16 trillion from OpenAI); investors are re-evaluating SoftBank’s exposure after it reportedly approved the final tranche of a planned $30bn investment in OpenAI (potentially reduced to $20bn) while OpenAI pursues aggressive capital spending plans exceeding $1 trillion over five years.

Analysis

Market structure: Leadership in model provision (GOOGL/GOOG) is likely to capture incremental cloud inference spend and ad/monetization upside while capital-hungry model operators (private OpenAI-style stakes) face margin compression; expect a 5–15% reallocation of high-margin inference revenue to hyperscalers over 12–24 months, boosting cloud pricing power and GPU demand (NVDA positive). Supply/demand: GPU capacity shortages may persist near-term, supporting NVDA ASPs and reorder cycles, but end-customer price sensitivity will pressure provider gross margins and raise capex intensity.

Risk assessment: Tail events include aggressive antitrust remedies (structural or API access limits) and a disorderly private-funding shortfall for OpenAI forcing asset writedowns; both are low probability but could wipe >20–30% from implied private valuations within 3–12 months. Near-term (days–weeks) expect volatility spikes and CVaR deterioration in Japan equities; medium-term (3–12 months) look for NAV repricing of private holdings; long-term (1–3 years) winners consolidate around integrated cloud+model stacks.

Trade implications: Favor conviction long positions in GOOGL (cloud + ad capture) and tactical NVDA exposure to ride constrained supply; hedge balance-sheet risk by shorting SoftBank (9984.T / SFTBY) or buying puts sized to cap downside. Use options to express view: 1–3 month put spreads on SoftBank to limit cash outlay and 2–4 month call spreads on NVDA/GOOGL to exploit falling IV after initial shock.

Contrarian angles: Consensus underweights SoftBank’s remaining public stakes and option-like upside of its Vision Fund winners — a >20% dislocation from implied NAV could be a buying opportunity on a 6–12 month horizon. Conversely, rapid dominance by GOOGL could invite stronger regulatory intervention than currently priced; position size accordingly and use pairs (long GOOGL, short high-exposure private-backers) to neutralize macro beta.

Key Decisions for Investors

• Establish a 2–3% long position in GOOGL (Class A) with a 12-month target +20–30% and a hard stop at -12%; add on any pullback to within 5% of the 50-day MA.
• Initiate a 1–2% exposure to NVDA via 90–120 day 5–10% OTM call spreads (limit max debit) to capture upside from continued GPU tightness; close on 30–50% realized spread return or NVDA guide-downs.
• Open a 2% short sized position in SoftBank (9984.T or SFTBY) and hedge with $-neutral pair by allocating an equal-dollar long in GOOGL; increase short by +1% if SoftBank confirms a final OpenAI tranche or falls another 10% within 30 days.
• Buy a 3-month put spread on SoftBank (e.g., -10% / -25% strikes) sized to cap downside to a 0.5–1% portfolio risk while maintaining optionality to profit from private-valuation revisions; unwind if market-implied NAV gap narrows below 15%.
• Rotate 3–5% from small-cap AI services into mega-cap cloud/semis (GOOGL, NVDA) over 30 days; reassess after NVDA earnings or any DoJ/FTC filings within 60–90 days for regulatory shock risk.