NVIDIA Corporation NVDA
NVDA, up 600% since 2016, faces in-house silicon from every mega-customer, collapsing crypto and autonomous narratives, and a benchmark-less Turing launch — shares revert to $200.
Thesis
Citron argues NVDA — the only >$100bn market-cap stock up more than 600% since 2016 — is pricing perfection just as execution risks compound. Q2 FY19 was the smallest beat in three years (<1% revenue, <5% EPS), ending a 13-quarter streak of crushing expectations. Mega-customers Baidu, Alibaba, Tencent, Google, Facebook and Amazon are all funding in-house AI silicon or TPU alternatives; EBAY publicly migrated from NVDA GPUs to Google Cloud TPUs (40-day ML task cut to 4 days), and Tesla's Elon Musk declared NVDA 'no longer necessary' with an in-house chip 10x faster. Jensen Huang's crypto tailwind reversed from 'important driver' to 'immaterial' in months, Graphcore is gaining mindshare, and the new RTX 2080 launched without a single traditional benchmark. With short interest at an all-time low and analysts still hiking targets, Citron sees shares retracing to $200.
SCQA
NVDA is the sole >$100bn market-cap stock up more than 600% since 2016, riding AI, datacenter and gaming momentum while analysts refuse to cut estimates even as Q2 beats narrowed sharply.
Every top customer — Google, Amazon, Tesla, Facebook, Alibaba, EBAY — is funding in-house AI silicon or TPUs; Musk called NVDA 'no longer necessary', crypto collapsed from 'key driver' to 'immaterial', and Graphcore plus Turing execution risks loom.
Short NVDA or take profits; discount analyst knee-jerk price-target hikes on AI/datacenter, wait for real Turing benchmarks, and recognize short interest at all-time lows signals consensus complacency.
Stock retraces from ~$245 to $200, roughly 18% downside; Citron is so confident it pledges $100k to the Humane Society if NVDA closes below $200 in April.
The three reasons
- 1
NVDA's biggest customers (Google, Amazon, Tesla, EBAY) are building in-house AI chips to replace GPUs
- 2
Q2 FY19 was the smallest beat in 3 years; crypto and autonomous tailwinds are unwinding
- 3
RTX 2080 launched without a single traditional benchmark — bull case rests on unproven upgrade cycle
Primary demands
- Sell or short NVDA
- Discount sell-side's knee-jerk price-target hikes on AI/datacenter narrative
- Wait for real traditional benchmarks on Turing/RTX 2080 before underwriting the upgrade cycle
KPIs cited
Pattern membership
Where this document fits across the library's 12 rhetorical / structural patterns.
Precedents cited
- Citron's NFLX short call ($390 → ~$329)
- NVDA has dropped 85% from highs twice before
Notable slides (5)
Notes
Classic Citron Word-memo format with Citron Research header/date on every page — not a designed slide deck. Memorable title metaphor ('dog that caught the car'). Heavy on CEO-quote contradictions (Jensen on crypto: 'important driver' → 'immaterial'; 'GPU fantastic at cryptography') and Musk quoting NVDA as 'no longer necessary'. EBAY before/after (NVDA GPUs → Google Cloud TPUs) is the strongest narrative device. Closing gimmick: Citron pledges $100k donation to Humane Society if NVDA closes below $200 in April — a trademark Citron rhetorical flourish. Document is primarily text with screenshots/charts pasted in (Bloomberg short-interest terminal shot, Reddit GeForce reactions collage, dog photo in closing). No formal valuation framework shown — $200 target asserted rather than built up.