Get essays like this every week

Subscribe
Company Analysis

The Magnificent Seven Are Not the Same Bet

Wall Street groups them together. Your portfolio treats them as interchangeable. They aren't. One sells the raw materials. One sells the cloud. One sells ads. The differences determine which ones win if AI delivers and which ones don't.

Seven Companies, One Label, Seven Different Businesses

Media coverage trained investors to think Apple, Microsoft, NVIDIA, Amazon, Alphabet, Meta, Tesla move together. "Magnificent Seven rose today." Label implies they move together and benefit from same forces.

They don't. AI exposures range 8% to 88% of revenue. Value chain positions differ. Risks differ. Owning all seven because they're AI companies is like owning both oil driller and gas station because both are energy.

I've included Broadcom (honorary eighth). Not officially Mag-7, but major AI infrastructure company. Leaving it out creates blind spot.

AI Revenue as a Percentage of Total Revenue
Estimated share of each company's revenue directly tied to AI products, infrastructure, or AI-driven systems
NVIDIA
88%
Meta
40%
Broadcom
35%
Microsoft
32%
Alphabet
28%
Amazon
22%
Apple
8%
Tesla
10%

NVIDIA gets 88 cents per dollar from AI. Apple gets 8 cents. Treating them as same bet hides real portfolio decisions.

Where Each Company Sits

AI economy has layers. Understanding which layer a company occupies tells you what needs to go right for the stock to work.

Position in the AI Value Chain
From silicon manufacturing to consumer-facing applications
Company Layer What They Sell Forward P/E
NVIDIA Infrastructure GPUs and networking for AI training and inference 35x
Broadcom Infrastructure Custom AI chips and data center networking 30x
Microsoft Platform Azure AI cloud, Copilot enterprise tools, OpenAI partnership 32x
Amazon Platform AWS AI services, Bedrock, custom Trainium chips 38x
Alphabet Platform + Application Google Cloud AI, AI-powered search, TPU chips 22x
Meta Application AI recommendation engine for advertising, Llama models 24x
Apple Distribution Devices that run AI models (Apple Intelligence) 30x
Tesla Application Full Self-Driving, Optimus robotics, energy storage 95x

Infrastructure companies (NVIDIA, Broadcom) win if AI spending grows, regardless of which applications succeed. Platform companies (Microsoft, Amazon, Alphabet) win if enterprises adopt AI through their clouds. Application companies (Meta, Tesla) win only if their specific products deliver.

Further down the stack, bet becomes more specific. NVIDIA wins in almost every AI scenario. Tesla wins only if autonomous driving and robotics work at scale. Categorically different risk profiles hiding behind same label.

What You're Actually Buying

NVIDIA The Arms Dealer
88%
AI Revenue
78%
Gross Margin
35x
Forward P/E

Sells to everyone, depends on no one application. Pure AI infrastructure play. Risk: priced for 40%+ annual growth. If AI capex slows, multiple compresses fast. CUDA lock-in is moat. Custom silicon slowly chipping it.

Microsoft The Platform Tax
32%
AI Revenue
45%
Net Margin
32x
Forward P/E

Most diversified AI bet. Azure growing 60%+ annually. Copilot embedding AI into Office. OpenAI partnership gives exclusive access. Risk: $80B annual capex must convert to revenue. If enterprise adoption slows, Microsoft spending more than earning on AI.

Alphabet The Undervalued Hybrid
28%
AI Revenue
28%
Net Margin
22x
Forward P/E

Cheapest at 22x earnings. Owns infrastructure (TPUs, Cloud) and applications (Search, YouTube). TPU development gives structural cost advantage over NVIDIA-dependent competitors. Risk: AI search could cannibalize ad revenue. This is existential question for Alphabet.

Amazon The Cloud Landlord
22%
AI Revenue
8%
Net Margin
38x
Forward P/E

Largest cloud by share, but margins thin because retail is low-margin. Trainium custom chips could reduce NVIDIA dependency. Bedrock gaining enterprise customers. Risk: AWS AI growth must accelerate to justify 38x earnings on 8% net margins.

Meta The AI Ad Engine
40%
AI Revenue
35%
Net Margin
24x
Forward P/E

AI runs entire revenue model. Recommendation algorithm decides what 3B users see, determining ad value. Llama open-source builds developer loyalty. Risk: all revenue from advertising. Downturn hits harder than other Mag-7. Reality Labs burns $15B annually.

Apple The Distribution Layer
8%
AI Revenue
26%
Net Margin
30x
Forward P/E

Least AI-exposed Mag-7 stock. Apple Intelligence is feature, not revenue driver. Bull case: on-device AI drives iPhone upgrade cycle. Bear case: mature hardware company trading at premium because used to be growth story. If AI doesn't drive upgrades, 30x multiple comes down.

Tesla The Wildcard
10%
AI Revenue
13%
Net Margin
95x
Forward P/E

Only Mag-7 where AI thesis is future revenue, not current. At 95x forward earnings, market pricing autonomous driving, robotaxis, Optimus robots. If ship, cheap. If not, car company at 95x. Highest risk, highest reward by wide margin.

How to Choose Which Ones to Own

Rather than buying all seven because they share a label, pick based on what kind of AI bet you want to make.

01
If You Believe AI Spending Grows Regardless
NVIDIA and Broadcom. Infrastructure always wins if industry grows. Don't need any particular application succeeding. Need aggregate spending continuing. Risk: entirely about duration of capex cycle.
02
If You Believe Enterprise AI Adoption Accelerates
Microsoft and Amazon. Cloud platforms capture enterprise AI spending. Every company deploying AI models pays MSFT or AMZN for compute. Risk: AI revenue must outgrow capex. Both spending aggressively.
03
If You Want Best Risk-Adjusted AI Exposure
Alphabet. Cheapest forward multiple (22x), owns infrastructure (TPUs) and applications (Search, YouTube), cost advantage from custom silicon. Existential risk (AI cannibalizing search) real but priced. If not materializing, rerated higher.
04
If You Want Maximum Asymmetry
Tesla. Position should be small because downside severe (car company at 95x). But if Full Self-Driving reaches Level 4 and robotaxis scale, upside measured in multiples, not percentages. Venture bet inside public stock.
7 ≠ 1
Seven companies. Seven different AI bets. Seven different risk profiles. Stop owning them as a group. Start owning them for what they actually are.

How I Built This

Revenue breakdowns and valuations from most recent earnings reports and SEC filings. AI revenue percentages are estimates based on segment disclosures and analyst consensus.

AI Revenue Percentages
Company filings + analyst estimates
NVIDIA 88% is Data Center segment share. Microsoft 32% uses Intelligent Cloud proxy. Meta 40% reflects entire ad engine running on AI recommendation systems. Apple 8% is estimated from Apple Intelligence-driven upgrades. Definitions debatable. Narrow (only products labeled AI) gives lower. Broad (anything using ML) gives higher. I use middle: revenue meaningfully declining if AI capabilities removed.
Forward P/E Ratios
Consensus estimates, February 2026
Forward P/E uses next 12 months estimated EPS from analyst consensus. Change daily with stock prices and estimate revisions. Relative ordering (Tesla highest, Alphabet lowest) stable over year.
Value Chain Classification
Author's framework
Infrastructure/platform/application/distribution layers are my classification, not industry standard. Alphabet spans two layers. Others debatable. Point isn't exact label but insight: different companies face different risks by location. Disagree with classification? Framework still works with your labels.
Jesse Walker
Jesse Walker has been an individual investor for 30 years. Before that, he was a poker professional, which is where he learned that the best decision and the best outcome aren't always the same thing. He writes about investing through AI uncertainty.

More where this came from

Data-driven AI investment analysis, delivered weekly. No spam.

Nothing on this site constitutes investment advice. All content is for informational purposes only. Full terms.