The best AI stocks in 2026 — only 3 pass even 5-of-7 value-investing frameworks (and the obvious one isn't on the list)

Every list of 'best AI stocks' includes NVIDIA, MSFT, GOOG, META. But when you apply the 7-framework value-investing screen, only three of the AI cohort actually pass. Here is the analysis — including why NVIDIA doesn't make the cut.

Every "best AI stocks 2026" list looks the same: NVIDIA, Microsoft, Google, Meta, AMD, maybe Palantir. They're listed because the AI narrative is correct — these are the companies building the AI infrastructure of the next decade.

But here's the question almost no list answers: do these stocks pass the value-investing canon's quality + price tests? We ran the 8 most-cited "AI stocks" through invest-like's 7-framework consensus screen. The result might surprise you.

Out of the 8 names, only 3 pass even 5-of-7 frameworks. The other 5 fail on valuation — sometimes severely. NVIDIA, which is the headline of every AI list, fails 5 of the 7 frameworks today.

This isn't a critique of AI as a thesis. It's a reminder that price paid is half of any value-investing decision. Below is the full screening result, with each ticker's verdict, the framework that's the killer, and the underlying number.

The screen results

Three names pass the 5-of-7 threshold: Microsoft, Alphabet, Amazon. The rest of the AI cohort either fails on valuation (NVDA, META, AMD, PLTR) or on cyclicality (TSM).

The three that pass — and why

Microsoft (MSFT) — the only 7-of-7

Microsoft is the cleanest AI play that also passes every value test. The framework consensus is unanimous:

• Buffett: passes on ROIC 30%+, balance sheet pristine, owner-earnings yield ~4% (borderline but acceptable)
• Graham: passes the enterprising-investor variant (P/E ~32 justified by EPS growth)
• Fisher: gross margin 70%+, sustained R&D, growth-quality
• Lynch: PEG ~1.5 (above his ideal but acceptable given EPS growth ~20%)
• Greenblatt: top-decile combined rank
• Munger: ROIC sustained 25%+ for a decade, willing-to-pay-up territory
• Smith: ROCE > 35%, FCF/NI 95%+, treated as software not capital-intensive

The Boardroom debate on Microsoft is the cleanest unanimity in the entire AI cohort. All four legend personas (Buffett, Graham, Lynch, Greenblatt) say long. The skeptic challenges AI-capex margin compression but doesn't reverse the verdict.

The risk: if cloud margins compress as AI capex ramps further, Buffett's pillar fails first. Microsoft would drop to 6-of-7 immediately. Watch the next two quarterly reports.

Live page: /buffett/msft/.

Alphabet (GOOG) — 6-of-7

Alphabet fails only on Graham's strict defensive criteria (P/E ~26 above the 15 threshold). The other six frameworks pass.

• Munger: ROIC sustained > 25% for a decade. Easy pass.
• Smith: FCF/NI 90%+, no banking exposure, ROCE > 30%. Pass.
• Greenblatt: top-decile combined rank (high ROIC, reasonable earnings yield).
• Fisher: gross margin 56%, sustained R&D. Pass.
• Lynch: PEG ~1.2, growth at reasonable price. Pass.
• Buffett: borderline pass — owner-earnings yield ~4.3% (just clears the 4% threshold for stable tech businesses).

The Alphabet thesis is that the search moat survives AI substitution. If Search's gross margin compresses materially (the bear case: ChatGPT-style answers replace search queries), several frameworks fall.

Live page: /buffett/goog/.

Amazon (AMZN) — 5-of-7

Amazon passes five of the seven frameworks:

• Buffett: partial pass on owner-earnings + AWS-driven cash generation
• Lynch: passes on EPS growth + GARP
• Greenblatt: passes on top-decile rank
• Munger: passes on AWS-business ROIC (despite retail-business low ROIC)
• Fisher: passes on R&D intensity + cloud growth

Fails:

• Graham: P/E ~50 fails defensive criteria
• Smith: capital intensity of e-commerce + AWS data centres trips Smith's no-cyclical filter

The Amazon thesis is that AWS becomes increasingly the dominant business and the e-commerce retail margin reverts toward Costco-like (low gross margin, high velocity). That's plausible; if AWS gross margin compresses on AI competition, Buffett and Lynch falter too.

Live page: /buffett/amzn/.

The five that don't make the cut

NVIDIA (NVDA) — 2-of-7

NVIDIA's framework consensus is shocking on first look — only 2 of 7 frameworks pass. But it's correct.

NVIDIA fails:

• Buffett: owner-earnings yield 1.6%, well below the 5% floor. P/E 65.
• Graham: P/E 65 fails by 4x. P/B 30+ fails by 20x.
• Munger: ROIC current is excellent (60%+) but the sustained requirement (5+ years) trips because ROIC was much lower 5 years ago. Munger requires sustained.
• Smith: cyclical semiconductor industry classification trips Smith's hard exclusion.
• Fisher: passes (gross margin 76%+, R&D intensity high) but the consensus screen requires 5-of-7.

NVIDIA passes:

• Lynch: partial — growth justifies the multiple in his framework
• Greenblatt: top-decile on ROIC rank, so cleared via mechanical rank

The Boardroom debate on NVIDIA is published in full at /blog/boardroom-nvda-debate-transcript/. The bear case is the cyclical concern; the bull case is "this time is different on autonomy/AI cycle structure."

Meta (META) — 3-of-7

Meta is a quality compounder (ROIC 25%+, FCF/NI 100%+) but fails:

• Buffett: owner-earnings yield ~3% — too low
• Graham: P/E 30+ — fails
• Smith: pre-2023 era Smith ruled out social media on regulatory durability concerns; current Smith framework still flags it
• Munger: voting structure plus AI capex creates capital-allocation concerns

Passes Lynch (growth), Fisher (R&D), Greenblatt (rank).

AMD (AMD) — 2-of-7

AMD is the volatile cyclical play in the AI cohort. Passes Lynch (growth) and Greenblatt (on rank during cycle peaks). Fails everything else because cyclical semiconductor margins make the durability and management pillars unstable.

Palantir (PLTR) — 1-of-7

Palantir's reported earnings barely exist (P/E ~250+), making most frameworks impossible to pass. It's a story stock — long if you believe the AI-government / commercial moat, but no value-investing framework can underwrite it at current prices.

TSMC (TSM) — 4-of-7

The Taiwan-based foundry that all the AI chips are made at. TSMC has extreme quality (ROIC 28%+, gross margin 55%+) but fails on:

• Smith: cyclical commodity classification
• Graham: P/E ~30 marginal
• Buffett: geopolitical-risk discount reduces the durability score

Passes Munger (sustained ROIC), Greenblatt (top-decile rank), Fisher (gross margin), partial Lynch.

The most controversial of the 8 — TSMC's geopolitical Taiwan-strait risk is the kind of qualitative concern the screen can't quantify directly. Read the Boardroom debate for the nuance.

What this means for an AI-focused portfolio

If you're building an AI-focused portfolio and you want to lean on value-investing principles for risk management:

1. Microsoft, Alphabet, Amazon are the three names that pass even 5-of-7 frameworks. Position-size them as high-conviction.
2. NVIDIA, Meta, TSMC are the high-quality / high-price names. Position-size them smaller, or wait for a price reset.
3. AMD, Palantir are speculative on current numbers. Don't expect value-investing frameworks to support them; if you own them, do so on a different framework (growth investing, momentum, narrative).

The 7-framework screen isn't anti-AI. It's anti-price-at-any-cost. There's a difference.

Verify any of this yourself

Every ticker named has a live framework breakdown on its /buffett/[ticker]/ page. The 5-pillar radar chart visualises the per-pillar scoring; the per-framework verdict is listed below; the numbers behind every claim are visible.

For the Boardroom-style debate on any of these stocks, open /boardroom/[ticker]/. The four investor AIs will weigh in side-by-side with a skeptic challenging the bull case.

Disclosure

Educational tool. The screen results are mechanical outputs of published rules applied to current Financial Modeling Prep fundamentals data. Past framework consensus does not predict future returns. None of the stocks named is a recommendation. Tickers may have been added to or removed from the screen since this article was written; the live pages reflect current scoring.

Author: Zaid Ghazal, founder of invest-like, indie SaaS, Kiel, Germany.