Investment Map

7 Layers of Compute

Tokens flow downstream. Margins flow upstream. Here's the full supply chain, mapped with real financials, so you know where to put your money.

L1 Raw Materials Silicon, Memory, Fiber

Every AI chip starts as a silicon wafer. Every GPU needs HBM memory. Every data center needs fiber optic cables. This is the foundation of the stack.

CompanyProductAI RevenueAI %Op MarginGrowth
SK Hynix
000660
HBM memory (#1, 60% share)~$30B~40%+49%+50%
Micron
MU
HBM, DRAM, SSDs$20.75B56%47%+50%
Samsung
005930
HBM (#2), DRAM, NAND~$60B semi div~18%*50%++208% Q4
Shin-Etsu
4063
Silicon wafers (#1)N/AN/A29%+6%
Corning
GLW
Fiber optics for DCs$6.3B38%19%+35%
SUMCO
3436
Silicon wafers (#2)N/AN/A0.3%Declining
The play

HBM memory is the gold. SK Hynix prints money at 49% margins with 60% market share. Every H100 needs 80GB of HBM3e. Every Blackwell needs 192GB. Memory demand scales linearly with GPU shipments.

Avoid

SUMCO (0.3% margin, declining revenue). Being the #2 wafer maker with no pricing power is a bad place to be.

L2 Chip Fabrication Foundries

Someone has to print the transistors. TSMC does it for almost everyone. Samsung and Intel are trying to compete, and both are losing money doing it.

CompanyProductAI RevenueAI %Op MarginGrowth
TSMC
TSM
Chip fab (#1, 65% share)~$21B17-19%51%+32%
Samsung Foundry
005930
Chip fab (#2)N/AN/ANegativeDeclining
Intel Foundry
INTC
Chip fab (#3)$307M ext.1.7%-50%Losses 4x'd
The play

TSMC is a monopoly in everything but name. 65% foundry share, 51% operating margins, every AI chip on the planet runs through their fabs. AI revenue CAGR of 54-56% through 2029. Capex jumping to $52-56B in 2026.

Avoid

Samsung Foundry and Intel Foundry are both hemorrhaging cash trying to catch up. The gap is widening. TSMC's moat comes from yield rates: 80%+ at 3nm vs. Samsung at sub-60%.

L3 Chip Design & Networking Architects

This is where the highest margins live. NVIDIA designs the chips. Broadcom designs the custom ASICs. Arista builds the networking that connects GPU clusters. They don't make anything physical. They design it, license it to foundries, and collect rent.

CompanyProductAI RevenueAI %Op MarginGrowth
NVIDIA
NVDA
GPU design, networking, CUDA$197.3B91%~60%+65%
Broadcom
AVGO
Custom ASICs, networking$8.4B/q44%67% EBITDA+106% Q1
AMD
AMD
MI300/400 GPUs, EPYC CPUs$16.6B49%33%+35%
Arista
ANET
DC Ethernet switches~$6.1B68%43%+29%
Marvell
MRVL
Custom AI silicon, networking~$4.3B75%34%+27%
The play

NVIDIA is in a class of its own: $216B revenue, $120B net income, 91% from AI. But Broadcom is the stealth winner. 77% gross margins, 70% custom ASIC share. They design the chips (Google TPU) that compete with NVIDIA. AI revenue doubled in one quarter.

Why NVIDIA's margins matter: 75% gross margin on $216B revenue = $162B in gross profit. That's more gross profit than most companies have total revenue. The CUDA ecosystem lock-in is the moat.
L4 Infrastructure Cloud, Data Centers, Power, Cooling

The landlords of compute. Cloud providers rent GPU time. Data center REITs sell physical rack space. Power utilities sell the electricity. Cooling companies keep it all from melting.

Cloud Providers

CompanyProductRevenueAI SegmentOp MarginGrowth
AWS
AMZN
Cloud, Trainium chips$128.7B$10B+ custom chips35%+20-24%
Azure
MSFT
Cloud, OpenAI API$75B+$13B AI run rate~43%+34-40%
Google Cloud
GOOGL
Cloud, TPUs, Vertex AI$70B+ run rateN/A23-30%+28-48%

Physical Infrastructure & Power

CompanyProductRevenueAI ExposureOp MarginGrowth
Vertiv
VRT
DC cooling, power mgmt$10.2BNearly all AI-driven22%+28%
Equinix
EQIX
Colocation (280+ DCs)$9.2B60% of top deals20%+5-11%
Digital Realty
DLR
DC wholesale space$6.1B20% of bookings26%+10%
Constellation
CEG
Nuclear power$25.5B20-yr PPAs (Meta)12%+8%
Vistra
VST
Nuclear/gas power$17.7BPPAs (AWS, Meta)11%+3% rev
The unsexy winners

Vertiv has orders up 252% and a $15B backlog. Every GPU deployed needs cooling. Liquid cooling market growing at 31.5% CAGR through 2033. Power is the bottleneck below the bottleneck. Constellation and Vistra have signed 20-year PPAs with hyperscalers. Nuclear plants can't be built fast enough.

L5 Model Labs Frontier AI

They train the models that create all the token demand. Their product is intelligence itself, sold per token. Every company in L7 depends on them. And every one of them is burning cash at industrial scale.

CompanyProductARRValuationMarginGrowth
OpenAI
Private
GPT models, ChatGPT~$25B$852BNegative (-$9B/yr)3x/yr
Anthropic
Private
Claude, Claude Code~$19B (Mar 26)$380BNegative~10x/yr
DeepMind
GOOGL
Gemini modelsLow single-digit BPart of AlphabetN/AN/A
Meta AI
META
Llama (open-source)$2-3B directPart of MetaN/A22% (Meta)
Mistral
Private
Mistral Large, Mixtral~$400M$13.7BNot disclosed~20x/yr
Can't invest yet

All private or embedded in larger companies. Anthropic's IPO (possibly Oct 2026) would be the first pure-play frontier AI stock. Until then, you're investing through cloud providers (L4) or chip makers (L3). OpenAI burned $9B in 2025, projecting $17B in 2026. The bet is that models get cheaper to run faster than the competition catches up.

L6 Inference Platforms Token Distribution

They buy GPU time wholesale and sell tokens retail. The spread between cost and price is the business. This is the fastest-growing layer by percentage.

CompanyProductRevenueValuationMarginGrowth
CoreWeave
CRWV
GPU cloud for AI$5.1B~$35B mkt cap61% EBITDA+170%
Fireworks AI
Private
Inference API$315M ARR$4B~50% gross+416%
Together AI
Private
Open-source inference~$300M-$1B$3.3-7.5BN/A3x+
Groq
Private*
LPU inference chips~$500M proj$6.9BN/AN/A
The play

CoreWeave (CRWV) is the only public pure-play. 170% revenue growth, 61% EBITDA margins. But more than half of revenue comes from one customer (likely Microsoft). Customer concentration is the risk.

L7 Applications Token Consumers

The addicts. Every company at this layer is 100% dependent on the layers below. Cursor can't function without inference tokens. If Anthropic raises prices or Together AI goes down, their product breaks. This is the Token Daddy thesis playing out.

CompanyProductARRValuationMarginGrowth
Cursor
Private
AI code editor~$2B$29-50BN/A10x+
Databricks
Private
Data lakehouse + AI$5.4B$134B~Breakeven FCF+55-65%
Perplexity
Private
AI search~$200M$21.2BN/A+230%
Jasper AI
Private
AI content marketing~$88M~$1.8BN/A-27% from peak
Winner-take-most

Cursor at $2B ARR is 10x Perplexity's $200M. Jasper's revenue declined from its peak, proving that commodity AI features get eaten by platforms when ChatGPT shows up. All private. Watch for Databricks IPO.

The Margin Waterfall

Gross/operating margins by layer. The pattern: margins are highest where you're closest to the silicon and where you hold a monopoly position.

L3 NVIDIA/AVGO
75-77%
L6 CoreWeave
61%
L2 TSMC
60%
L1 SK Hynix
49%
L4 AWS/Azure
35-43%
L3 Arista
43%
L4 Vertiv
22%
L4 Power
10-12%
L5 Model Labs
Negative

The Key Insight

Money flows downstream. Margins flow upstream.

TOKEN FLOW (downstream) MARGIN FLOW (upstream) L7 Apps pay for tokens ───▶ L1 Raw materials L6 Inference marks up ───▶ L2 Fabrication L5 Models charge per token ───▶ L3 Chip design ★ HIGHEST MARGINS L4 Cloud rents GPUs ───▶ L4 Infrastructure L3 Chips sell to cloud ───▶ L5 Model labs (NEGATIVE) L2 Fabs print chips ───▶ L6 Inference L1 Materials feed fabs ───▶ L7 Applications
Applications (L7) are the most visible. Chips (L3) and memory (L1) capture the most value. The addiction creates the demand, but the dealer keeps the margin.

Where to Put Money

Public, investable companies ranked by conviction.

Tier 1 - Strongest conviction

NVIDIA

NVDA
91% AI revenue, 75% gross margin, +65% growth. The Token Daddy of chip design. CUDA ecosystem lock-in is the moat.
Tier 1 - Strongest conviction

TSMC

TSM
Irreplaceable monopoly. 65% foundry share, 51% operating margin. Every AI chip flows through their fabs. AI CAGR 54-56% through 2029.
Tier 1 - Strongest conviction

Broadcom

AVGO
Stealth winner. 77% gross margins, 70% custom ASIC share. Designs Google TPUs. AI revenue doubling quarterly.
Tier 2 - Strong conviction

SK Hynix

000660
HBM memory monopoly. 49% margins, 60% share. Every GPU needs it. Memory demand scales linearly with chip shipments.
Tier 2 - Strong conviction

Vertiv

VRT
Every GPU needs cooling. Orders +252%, $15B backlog. The unsexy winner. Liquid cooling CAGR 31.5% through 2033.
Tier 2 - Strong conviction

Arista Networks

ANET
68% AI revenue, 43% operating margin. GPU clusters need networking. 800G/1.6T Ethernet is the next wave.
Tier 3 - Speculative

CoreWeave

CRWV
Only public pure-play GPU cloud. 170% growth. But single-customer concentration risk (50%+ from one buyer).
Tier 3 - Speculative

Micron

MU
HBM ramp to $8B run rate, 47% DC margins. #3 behind SK Hynix/Samsung. Solid but not dominant.
Tier 3 - Speculative

Constellation Energy

CEG
Nuclear PPAs with hyperscalers. 20-year contracts. Power is the bottleneck. Down 30% from peak, potential entry.

IPO Watchlist (Can't invest yet)

CompanyARRValuationExpected IPO
Anthropic~$19B+$380BPossibly Oct 2026
Databricks$5.4B$134BPreparing
Cursor~$2B$29-50BTBD