#分享美股交易赢英伟达股票 NVIDIA (NVDA) Deep Dive Analysis: An AI Money Printer Earning 600 Million Yuan Daily, Is It Still a Buy?

1. Who is NVIDIA?
NVIDIA (NASDAQ: NVDA) is the absolute leader in global AI chips. The company was founded by Jensen Huang in 1993, initially known for gaming graphics cards (GPU). After the deep learning revolution in 2012, GPUs became the core infrastructure for AI training, sparking NVIDIA’s epic growth.
Currently, the company's business is divided into two main segments:
Compute & Networking — accounting for 91% of revenue, including Data Center AI accelerators (H100/H200/B200/B300), InfiniBand/Spectrum-X network switches, NVLink interconnects, AI software ecosystem (CUDA), autonomous driving platforms;
Graphics — accounting for 9% of revenue, including GeForce gaming GPUs and RTX workstation GPUs.
In one sentence: Almost every round of large AI model training worldwide runs on NVIDIA’s chips. Its moat isn’t just hardware itself but a combination of CUDA software ecosystem + chip generation leadership + system-level integration capabilities.
2. Financial Panorama: A Money Printer
Latest quarterly data is explosively impressive
Key details:
What does a 71.5% net profit margin mean? For every $100 earned, $71.5 is pure profit. Apple’s net margin is about 25%, Microsoft around 35%, NVIDIA is 2-3 times higher.
Net profit growth (+211%) far exceeds revenue growth (+85%) — a peak scale effect. Operating expenses grew only 52%, while gross profit increased by 129%.
Blackwell B300 full deployment — Data Center revenue $75.2B, up 92% year-over-year, with Blackwell now comprising most system shipments.
Customer diversification — mega cloud clients account for about 50% of DC revenue, with the other 50% from AI cloud, industrial enterprises, sovereign clients, etc., reducing concentration risk.
3. Balance Sheet: Fortress Level
Points worth noting:
Current assets $151B, current liabilities only $44B — liquidity ratio of 3.4x, almost no short-term debt pressure.
Marketable equity securities surged from $12.9B to $30.2B, non-marketable securities from $22.3B to $43.4B — NVIDIA is heavily investing in AI ecosystem companies. Quarterly purchase of non-marketable securities was $18.6B, contributing $15.9B in "other income" on the income statement.
Operating cash flow in 2026Q1: $50.3B — generating over $50 billion in cash in just one quarter, truly a money printer.
4. Valuation: Expensive or Not?
Valuation frameworks — various perspectives
Perspective 1: PEG valuation method
PEG = PE ÷ earnings growth rate. Peter Lynch believed PEG < 1 indicates undervaluation. NVIDIA’s latest quarterly net profit grew 211%. Even with a conservative assumption that future growth slows to 50%, forward PE ~26x / 50 = PEG ≈ 0.52, still undervalued. Market doubts about sustainability of growth create valuation discounts.
Perspective 2: DCF thinking
NVIDIA’s Q1 operating cash flow is $50.3B, annualized over $200B. Even assuming zero growth over the next 5 years, just maintaining current profitability, the cumulative cash flow over 5 years can cover the current market cap ($5.1T / $200B ≈ 25 years). For a company growing at 85%, this payback period indicates extremely strong growth potential.
Perspective 3: Comparable valuation
Compared to other tech giants: Apple PE ~33x (growth ~5%), Microsoft PE ~35x (growth ~15%), Tesla PE >100x. NVIDIA PE is 32x but with much higher growth — from a PEG perspective, NVIDIA is the cheapest among the Mag 7.
Perspective 4: How much PE has compressed? In 2024, NVIDIA’s PE once soared to 60-80x, but the current 32x PE is after EPS has multiplied several times. The stock price has pulled back about 10% from a high of $236, but EPS growth far outpaces the stock price increase. This is a classic case of "profits chasing valuation."
5. Growth Drivers
NVIDIA’s growth isn’t driven by price hikes but by technological leaps leading to volume and price increases:
Blackwell platform B300 has fully ramped up, accounting for most system shipments. B300’s performance is several times higher than H100, with higher ASPs, and gross margins have improved rather than declined.
Vera Rubin’s next-generation architecture will launch in late 2026, maintaining a "one generation per year" rhythm, with at least 2-3 years of technical lead.
Market for AI inference is exploding; concerns that inference demand is less than training are unfounded, as GPT-5 level models require exponential inference compute. The inference market may surpass training.
Enterprise AI penetration: 50% of DC revenue now comes from non-huge cloud clients, with AI expanding from cloud providers to enterprises, industrial sectors, and sovereign states. This is long-term infrastructure building.
Network business: InfiniBand + Spectrum-X Ethernet + NVLink form bundled sales, increasing customer value. Network business is also growing rapidly.
Software ecosystem: 15 years of CUDA accumulation, over 5 million developers, with no competitors able to replicate in 5 years. This is a true economic moat.
6. Risks: The Unseen Grey Rhino
Export controls
Hopper has stopped shipments to China (last year’s 🇨🇳$4.6B). If restrictions expand to Blackwell or more countries, impact will be significant.
Customer self-developed chips
Google TPU, AWS Trainium, Microsoft Maia, etc., are advancing. But self-developed chips mainly replace low-end inference scenarios; high-end still depends on NVIDIA.
AI Capex cycle
Microsoft, Google, Amazon’s AI Capex is a major revenue source. If AI investments underperform, Capex could sharply decline.
Slowing growth
85% growth isn’t sustainable. Due to base effects, growth will inevitably slow, but whether it’s a gentle 40-50% decline or a cliff drop is key.
Competition (AMD, etc.)
AMD MI300X has some competitiveness, but CUDA ecosystem lock-in is too strong. Difficult to shake dominance in the short term.
High inventory risk
Inventory of $25.8B is substantial, but turnover days are healthy. Last fiscal year’s $4.5B impairment provision for H2 serves as a warning.
Geopolitical risks
Chip manufacturing depends on TSMC, geopolitical risks always exist. NVIDIA is actively promoting supply chain diversification.
7. Overall Judgment
Strong fundamentals + reasonable valuation → Long-term buy rating
8. Investment Strategy Suggestions
Rating: Long-term buy
Buy logic:
PEG < 0.7, severely undervalued among large tech stocks — market overpricing growth slowdown risk
Gross margin 75% and net margin 72% prove this isn’t a cyclical stock but a platform company with extreme pricing power
After Blackwell, Vera Rubin (late 2026) with a clear tech roadmap
AI inference, enterprise AI, sovereign AI still in early stages, growth potential far from exhausted
$19B+ quarterly buyback indicates management believes the stock is undervalued
Positioning (for reference, not investment advice):
Current (~$210): establish initial position (1/3), PE 32x is reasonably low
If it dips to $180-190 (PE ~27x): add to 2/3 — the lowest PE zone since 2025
If it drops below $150 (PE ~23x): go all-in — unless AI narrative completely collapses, this valuation is giving money away
Take profit when PE exceeds 50x (stock price ~$330+), consider partial profit-taking
People who shouldn’t buy:
Investors unable to withstand 20-30% drawdowns
Income-focused investors expecting dividends
Skeptical about long-term AI prospects
Short-term traders / swing operators (high volatility)$NVDA