Nvidia (NVDA)

NVDA is Nvidia Corporation, the dominant designer of GPUs for AI training and inference and the primary supplier of accelerated computing infrastructure to hyperscale cloud platforms. Revenue mix in 2026: roughly 85% Data Center, 10% Gaming, and 5% Professional Visualization plus Auto. The Data Center business (Hopper, Blackwell, and now Rubin generations) drives essentially all incremental earnings. Gross margin runs above 70% at the company level, the highest of any semiconductor company at scale.

Why it matters: NVDA is the cleanest pure-play on the AI capex cycle. Roughly 80% of AI training infrastructure spending flows to Nvidia GPUs, with the four hyperscalers (MSFT, GOOGL, AMZN, META) plus Oracle and CoreWeave accounting for over half of revenue. NVDA market cap volatility is the largest single swing factor in QQQ and a meaningful swing factor in SPY because index concentration has lifted NVDA to roughly 7% of the S&P 500 in 2026.

NVDA trades near all-time highs in April 2026 after roughly 8x appreciation from the late-2022 ChatGPT-launch base. Blackwell and the upcoming Rubin product cycles support a multi-year revenue trajectory at scale, and the four hyperscalers' combined 2026 capex of approximately $300 billion implies an NVDA-addressable market significantly larger than current run-rate revenue.

The bear case is concentration risk (top five customers are roughly 50% of revenue), competitive risk (AMD MI400, Google TPU v7, Amazon Trainium 3, plus custom hyperscaler ASICs), and US export controls to China that have already removed roughly $5-$10 billion of annualized China revenue. The bull case is that AI capex growth absorbs all of these risks for at least another 18-24 months and that inference workloads (rather than training) become the next leg of growth.

NVDA has compounded approximately 70%+ annualized since 2015, the strongest large-cap return profile in modern history. Major drawdowns: -85% in dot-com (2000-2002), -65% in 2018 (crypto-mining wipeout plus GPU oversupply), -65% in 2022. The three drivers are AI training capex (hyperscaler spending), datacenter GPU pricing (Blackwell/Rubin ASP), and competitive share. Inference revenue growth (a smaller but lower-volatility business) is the multiple-stabilizer.

First, hyperscaler capex guidance. Combined MSFT + GOOGL + AMZN + META + ORCL capex above $400 billion for 2026 is the bull threshold; below $300 billion would trigger meaningful re-rating.

Second, Blackwell and Rubin shipment cadence and gross margin per unit. Any sign of price erosion would compress the high-70s gross margin meaningfully.

Third, China revenue trajectory under export controls. The H20-and-successor product line is the regulatory swing factor; any further restriction of the AI Diffusion Rule could remove additional billions.