Nvidia hits back at short seller as leaked memo goes public

Nvidia circulated a seven-page memo rebutting a Substack claim that it is using “vendor financing” to inflate AI chip sales, saying investments in customers are small relative to overall sales and that customers pay for chips in about 53 days rather than multi-year vendor-finance terms. Noted short sellers Jim Chanos and Michael Burry counter that Nvidia may be supporting loss-making buyers, using debt and off‑balance-sheet structures and aggressive revenue recognition to fuel orders, raising questions about whether hyperscalers and startups are over-ordering GPUs ahead of real demand.

Analysis

Market structure: Nvidia remains the primary beneficiary — sustained lead in transformer inference/FP16 chips gives NVDA pricing power and order visibility into TSMC capacity for at least 12–18 months. Winners also include hyperscalers (AMZN, MSFT, GOOG) who extract margin from AI services; losers are smaller AI infra vendors and unprofitable start‑ups reliant on vendor equity/credit that face credit stress if capex tightens. Short‑term supply remains tight; a meaningful inventory build (>=2 quarters of sales) would flip dynamics fast and force discounting.

Risk assessment: Tail risks include an SEC probe into related‑party financing or restatements, sudden customer bankruptcies, or US export controls tightening to China — each could shave 20–40% off NVDA forward multiples within months. Immediate volatility (days) is likely around leaked memos and short reports; earnings and 10‑Q disclosures in 30–90 days are key catalysts. Hidden dependencies: NVDA’s revenue tied to a small number of large customers and TSMC node capacity; off‑balance‑sheet financing or equity stakes that materially exceed 3–5% of customer cashflows are red flags.

Trade implications: For bullish conviction, prefer defined‑risk 3–6 month call spreads on NVDA to capture upside into earnings while capping capital; for bearish conviction, use put spreads or buy CDS on small cap AI builders rather than naked short NVDA. Pair trades: long cloud operators (MSFT, AMZN) vs short capital‑intensive AI infra names; volatility trades: buy 1–3 month NVDA straddles ahead of quarterly filings if implied vol < historical realized +15bps.

Contrarian angles: Consensus undervalues the durability of AI compute demand — transformer workloads scale non‑linearly with model size and could sustain high GPU utilization for years — but overestimates short‑term elasticity: a 10–15% GPU price rise may only trim demand 5–10% this year. The market may be overpricing accounting risk; if filings show vendor financing <5% of receivables and DSOs remain ~53 days, downside is limited and a squeeze could occur. Historical parallels: telecom vendor‑financing collapses were driven by long‑dated receivables; Nvidia’s current cash collection profile is materially shorter, altering the expected outcome.

Key Decisions for Investors

• Establish a 1–2% portfolio long in NVDA via a 3‑month call debit spread (buy ATM call, sell 10% OTM) to capture upside into next earnings while limiting capital at risk; target hold 30–75 days and trim on +15% move.
• If you hold NVDA outright, buy 3‑month 12–15% OTM puts or implement a collar (sell 20% OTM calls) sized to protect 50–75% of position value; deploy if NVDA implied vol rises >30% vs 90‑day average or stock drops >10% intraday.
• Initiate a 1–2% pair trade: long MSFT (cloud exposure) and short 0.5–1% NVDA (hardware concentration) for 3–6 months to hedge hardware‑specific accounting/financing risk while keeping cloud secular upside.
• Reduce/new exposure to small‑cap AI infra companies by 50% and reallocate to TSMC (TSM) or large cloud capex beneficiaries (AMZN, MSFT); if any customer financing disclosures exceed 5% of vendor receivables or DSOs rise >10 days QoQ, increase hedges.
• Monitor next 30–60 days for three catalysts: NVDA’s detailed revenue/receivable disclosures in earnings/Q filings, any SEC/DOJ inquiries, and customer solvency signals (missed payments or public funding shortfalls); act to add protection or opportunistic longs if disclosures confirm vendor financing <5% of sales.