Nvidia Slams Michael Burry's Claims of an AI Bubble and Faulty Accounting

NVIDIA pushed back on accounting accusations from investor Michael Burry, saying his calculations misinterpret tax-related RSU entries and that stock repurchases since 2018 total $91 billion. The company defended its use of stock-based compensation, saying grants were not unusually large when made, argued real-world GPU depreciation aligns with a roughly four-to-six-year useful life, and noted strategic investments are a small share of revenue with portfolio firms serving outside customers—mitigating claims of circular cash flows or distorted profitability.

Analysis

Market structure: NVIDIA (NVDA) benefits most — the company’s detailed rebuttal reduces immediate headline risk and preserves its pricing power in high-end datacenter GPUs where ASPs and gross margins remain above peers. Competitors (AMD) and smaller AI chip entrants lose relative negotiating leverage because NVDA’s rebuttal reinforces incumbent advantages; buybacks of $91bn since 2018 signal continued capital return capacity that supports EPS even if stock-based comp noise persists.

Risk assessment: Tail risks include an SEC probe or accounting restatement that reduces reported EPS by >10% or forces buyback accounting adjustments within 90–180 days, which would trigger sharp re-rating and volatility. Hidden dependencies: long 4–6 year effective GPU lifecycles mean capex cadence and secondary-market GPU sales can distort revenue/cost recognition and inflate near-term cash ROIs; monitor stock-comp as % of revenue — >5–8% is a red flag.

Trade implications: Near-term (days–weeks) expect muted negative reaction; short-term (1–3 months) volatility around earnings/filings. Favor hardware/infra exposure (NVDA, LRCX, AMAT) and hedge narrative risk with options; avoid or short smaller AI software names where stock-based comp and buybacks are a larger share of shareholder returns (e.g., C3.ai) over 3–6 months.

Contrarian angles: Consensus underestimates enforcement risk—if regulators press accounting changes, multiple compression could be 15–30% for narrative-driven AI names while NVDA’s moat (software stack + ecosystem) limits downside to low double-digits. Mispricing opportunity: crowding in longs without hedges; volatility premium on NVDA options may spike 20–40% on any formal inquiry, creating tactical option-entry windows.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (ticker NVDA) for a 6–12 month horizon, targeting +15–30% upside if datacenter demand persists; size small due to crowding and set a stop-loss at -15% or on any SEC probe announcement.
• Buy 3-month NVDA puts 5% OTM sized to 25–30% of the long position (hedge ratio) to cap downside ahead of next earnings; if implied vol is >annualized 60%, switch to a 3-month ATM put to control premium.
• Implement a pair trade: long NVDA (or NVDA supplier LRCX) vs short C3.ai (ticker AI) — size short at ~50% of long notional for 3–6 months to exploit accounting-scrutiny dispersion; cover if AI reports stock-comp <5% of revenue or posts positive FCF.
• Increase overweight to semiconductor capital equipment (Lam Research LRCX, Applied Materials AMAT) by +1–2% of portfolio over 3–6 months anticipating sustained GPU-driven capex; trim high-valuation AI application/software names by 2–4% where stock-based comp >7% of revenue or buybacks exceed 50% of free cash flow.