Nvidia’s CFO admits the $100 billion OpenAI megadeal ‘still’ isn’t signed—two months after it helped fuel an AI rally

Nvidia clarified that its headline OpenAI arrangement—announced as up to 10 gigawatts of capacity and up to $100 billion of staged investment—remains at the letter-of-intent stage and lacks a definitive agreement, casting doubt on earlier market interpretations that implied multi-hundred-billion-dollar revenue upside (analysts had cited as much as ~$500bn). Nvidia's recent 10-Q highlights material execution risks: non-cancelable ahead-of-time component purchases, potential excess inventory, and multi-year constraints on data-center power, permitting customer delays or scale-backs; management says current 2025–26 demand guidance (~Blackwell and Vera Rubin systems) excludes the prospective OpenAI tranches and OpenAI currently buys via cloud partners. CFO Colette Kress emphasized the longstanding OpenAI relationship and Nvidia's platform moat, but investors should reprice the probability and timing of the megadeal given regulatory, supply-chain, energy and financing frictions.

Analysis

Market structure: The LOI clarification reduces the certainty of a multi-hundred‑billion revenue stream for NVDA and shifts near‑term winners to cloud intermediaries (MSFT, ORCL) who can continue to supply OpenAI without a direct buy; NVDA retains pricing power via CUDA/software, but hardware demand timing is now the key variable. Supply/demand: advance non‑cancelable orders, HBM constraints and multi‑year power buildouts make upside lumpy — expect 6–18 month phasing risk and potential short‑term inventory overhang if customers delay deployments. Cross‑asset: higher GPU capex implies increased demand for copper/transformers (benefit FCX, AEM) and power‑grid capex that can modestly lift utility equities and municipal issuance; credit spreads for data‑center developers could widen if project financing falters.

Risk assessment: Tail risks include definitive‑deal collapse, major power permitting failures, regulatory export/antitrust actions, or a gray‑market GPU dump that forces inventory impairments — each could shave 20–40% off NVDA near‑term EPS in a severe scenario. Time horizons: immediate (days) = elevated headline volatility; short (1–6 months) = order cadence and 10‑Q language driving guidance revisions; long (12–36 months) = actual buildouts and software moat determining market share. Hidden dependencies: customer financing, utility permitting, and third‑party HBM supply are single points of failure; catalysts are a signed definitive agreement, cloud vendor purchase disclosures, or utility permit milestones.

Trade implications: Reduce directional NVDA exposure and hedge event risk: favor cloud platforms (MSFT, ORCL) and infrastructure suppliers (transformer/equipment makers, copper miners) that capture buildout spend even if direct buys stall. Options: implement asymmetric hedges — buy 3‑month NVDA put spreads (ATM to 15% OTM) sized to cover 30–50% of beta exposure, sell near‑dated covered calls to monetize premium if holding. Pair trade: go long 2% portfolio weight MSFT and short 1.25% NVDA to express confidence in cloud capture of demand while limiting single‑name hardware exposure.

Contrarian angles: Consensus understates execution friction of power and financing — market is pricing near‑certainty of the LOI; if NVDA fails to sign within 90 days, downside is underpriced. Conversely, if a definitive deal is announced, NVDA likely re‑rates materially; plan contingent buys: add to NVDA if shares drop >15% on LOI uncertainty or buy on definitive‑deal announcement within 30–90 days. Historical parallel: crypto‑era GPU oversupply shows inventory and gray‑market risk can compress margins quickly, so size positions to absorb a 20–30% drawdown.

Key Decisions for Investors

• Trim NVDA exposure to target 3–5% portfolio weight (from typical 6–8%) within 10 trading days and establish downside protection: buy 3‑month NVDA put spread (buy ATM, sell 15% OTM) sized to cover 40% of residual NVDA delta exposure.
• Establish 2–3% long positions in MSFT and ORCL (split 60/40) to capture Cloud/partner routing benefits; add if either reports cloud procurement growth >10% QoQ or announces OpenAI‑related capacity deals.
• Allocate 1–1.5% to infrastructure beneficiaries: enter long positions in copper miner FCX and utility/transformer suppliers (e.g., AEM) with 6–18 month horizon to play power/buildout capex; add on any NVDA‑led guidance shortfall.
• Implement a pair trade: long 2% MSFT vs short 1.25% NVDA to express relative conviction in cloud capture; rebalance if NVDA outperforms by >12% in 30 days or if a definitive NVDA‑OpenAI agreement is signed.
• Set specific triggers: if NVDA fails to sign a definitive agreement within 90 days, increase hedges to cover 75% of NVDA exposure; if a definitive agreement is announced, selectively add 2–4% to NVDA on confirmation of funding tranches and power permits.