NVIDIA at $148 was the cleanest test of value-investing dogma since FANG in 2017. The price-to-earnings ratio of 65 says "speculative bubble". The 70%+ gross margin, the AI moat, and the 31% ROIC say "wonderful business, hold forever". Same stock, opposite reads from the canon.
This post applies the published rules of four investors to the same numbers, as of 19 May 2026, NVDA close $147.94. None of these investors said any of this about NVIDIA. The readings below are our AI-assisted paraphrase of what their published rules imply. Read them as an illustration, and open the NVIDIA page for today's numbers.
The four readings
Buffett's rules: fail on price
A P/E of 65 prices in many years of near-perfect compounding at the current margin. The moat is real, but the price asks you to predict the next two product cycles. At 25 times earnings the test would look different; at 65 times the second half of "a wonderful company at a fair price" fails.
Quality passes (ROIC 31.4%, balance sheet pristine). Valuation fails (owner-earnings yield 1.6% against the 5% floor).
Where the rule comes from: the 1989 Berkshire letter on fair prices, the 1986 letter on owner earnings.
Graham's rules: fail outright
P/E above 65, P/B above 30, current ratio at 1.9: zero of five defensive criteria pass. His enterprising-investor test still asks for a real earnings yield, and NVDA's was 1.5%. The rules do not need a view on the AI cycle to reject it.
Where the rule comes from: The Intelligent Investor, chapter 14.
Lynch's rules: cautious yes
Earnings growth is doing the heavy lifting, not multiple expansion. EPS grew 232% in the prior year and was forecast at 47%. PEG comes out to roughly 1.4: above his 1.0 ideal, but not in the dangerous zone if the growth rate is real. The open question is whether "AI capex" is a 20-year cycle like PC capex, or a 2-year cycle like crypto-mining capex. Growth comfortably above the bar, debt-to-equity clean. His warning about stories everyone already knows applies.
Where the rule comes from: One Up on Wall Street, on PEG and knowing the story.
Greenblatt's rules: mixed
The Magic Formula does not care about narratives. Earnings yield of 1.5% is weak; ROIC of 31% is excellent. On the rank-based formula, NVDA's combined rank in our universe was roughly 480 out of 3,500: top decile on quality, bottom quintile on price. The formula re-ranks once a year, so it would simply re-test.
Where the rule comes from: The Little Book That Beats the Market.
The strongest counter-argument
Lynch's and Greenblatt's readings lean on current ROIC and current EPS growth. NVIDIA's gross margin was 76%, against a 30-year semiconductor median of about 35%. If margins mean-revert even halfway, EPS falls by 50% or more, and a PEG of 1.4 becomes a PEG of 3.5.
The reply from the bull side: the classic cyclical pattern comes from commodity semis such as DRAM, NAND and low-end logic. NVIDIA designs accelerators, does not fab them, and sells to customers competing with each other on capex. That is a different industry structure, but it is still an extrapolation.
Where the rules agree and split
For NVDA at $148:
- Quality of business: all four read it as wonderful
- Balance-sheet risk: all four read it as clean
- Cyclicality risk: split. Buffett's and Graham's rules treat it as cyclical. Lynch's and Greenblatt's readings can treat it as structurally different from past cycles
- Valuation: 3-1 against. Only Lynch's PEG test can justify the price
What to take from this: NVDA's business is the bull case; NVDA's price is the bear case; the rest is your own view on cyclicality. Different rules weight risks differently, and the useful part is seeing where they split.
Do this on a stock you own
Every stock page on invest-like checks the company against the published rules of seven investors, with the number beside each rule. The AI analysis writes the bull case and the bear case from the same numbers. Every stock page is free, with 3 AI analyses a month.
Disclosure
Written with AI assistance. The readings above are paraphrases of what published rules imply, not statements by the investors named. Read the original letters and books yourself: the Berkshire letters are free at berkshirehathaway.com. Numbers are as of 19 May 2026 and have changed since.