Pledge to invest $100 billion in OpenAI was ‘never a commitment,’ says Nvidia’s Huang

Nvidia CEO Jensen Huang clarified in Taipei that the company’s previously announced plan — a letter of intent to invest as much as $100 billion in OpenAI to build at least 10 gigawatts of data-center capacity — was never a binding commitment and that Nvidia will evaluate funding rounds "one at a time." The comments follow a Wall Street Journal report that the plan has stalled amid internal doubts and Huang’s private concerns about OpenAI’s business discipline and competition; he said Nvidia’s actual contribution to the current round will not approach $100 billion. The exchange underscores investor scrutiny of circular AI deals (vendors investing in major customers) and signals more cautious, incremental capital deployment from Nvidia while it continues other strategic investments such as an additional $2 billion into CoreWeave.

Analysis

Market structure: The headline weakens the narrative that Nvidia (NVDA) irrevocably bankrolls OpenAI; winners remain NVDA (channel demand, incumbency) and hyperscalers/cloud GPU specialists (AMZN, MSFT, GOOGL) who supply complementary services, while smaller cloud/infra players face capital and power-availability squeezes. The nonbinding $100B reduces immediate dilution risk for NVDA but increases uncertainty in projected multi-year GPU demand tied to a 10 GW buildout — expect meaningful capex phasing over 1–3 years and near-term demand concentration in 2–4 quarters.

Risk assessment: Tail risks include antitrust/competition probes (US/EU) if preferential supply emerges, project execution risk (failure to deploy first GW within 6–12 months), and a demand cliff if OpenAI pivots suppliers; probability moderate but impact high. Near-term (days–weeks) volatility will be driven by headlines and NVDA earnings; medium-term (3–12 months) by GPU shipment cadence and utility power approvals; long-term (1–3 years) by actual data-center builds and model-training cycles.

Trade implications: Preferred tactical exposure is directional NVDA with risk control: small overweight position implemented via 6–9 month call spreads to cap premium, and a relative-value pair long NVDA / short AMD to express NVDA’s superior AI moat while hedging cyclical semiconductor risk (delta-neutral 0.8:1). Rotate modestly into cloud/software (MSFT, AMZN) for secular AI demand capture and underweight legacy CPU/server hardware; watch IV spikes >30% as entry triggers for options hedges.

Contrarian angles: Consensus underestimates that a nonbinding LOI actually reduces execution risk/dilution — the market could be overpricing governance concerns if NVDA provides staged capital under strict milestones. Historical parallels (large-strategic vendor investments in early cloud era) show phased funding often preserves demand without immediate balance-sheet strain; downside is regulatory scrutiny that could force structural remedies within 6–18 months.

Key Decisions for Investors

• Establish a 1–3% portfolio long position in NVDA via a 6–9 month bull call spread (buy ATM call, sell 20% OTM) to capture continued AI hardware upside while capping premium; scale in over 2–6 weeks on headline volatility.
• Implement a pair trade: long NVDA and short AMD (ticker AMD) sized to be approximately delta-neutral (NVDA:AMD 0.8:1 by delta) as a 3–6 month relative-value play on NVDA’s software/hardware ecosystem advantages; trim if NVDA/AMD spread narrows <10%.
• Overweight cloud software leaders MSFT and AMZN by +2–4% relative to benchmark to capture demand for managed AI stacks; target rebalancing at next earnings (within 60–90 days) or if cloud growth guidance misses by >200 bps.
• If NVDA implied volatility rises >30% (relative to its 12-month median), buy 3–6 month protective puts (10–15% OTM) on existing NVDA exposure or convert call spreads to collars; this hedges against a regulatory or execution shock within the next 3–12 months.