Big Tech just committed $650B+ in 2026 capex.
Here's how that money flows through every layer of the AI stack:
Compute: GPUs and Accelerators
$NVDA - Blackwell shipping now, Rubin next
$AMD - MI300/MI400 giving hyperscalers a second source
Custom Silicon and Chip Design
$AVGO - Custom AI chips and networking ASICs
$MRVL - Interconnect and custom silicon for AI racks
$ARM - CPU architecture powering hyperscaler fleets
$CDNS - EDA tools behind hyperscaler chip programs
$SNPS - Design software enabling next gen silicon
Memory and Storage
$MU - HBM and advanced DRAM feeding AI accelerators
$SNDK - NAND flash across hyperscaler storage
Networking and Connectivity
$ANET - Switching infrastructure inside AI clusters
$ALAB - Rack scale connectivity silicon
$CRDO - High speed SerDes for AI interconnect
$COHR - Photonics and lasers for optical networking
$GLW - Fiber and optical materials connecting it all
Servers and Infrastructure
$DELL - AI server systems for enterprise and hyperscale
$SMCI - Liquid cooled GPU racks, first to market on new chips
$HPE - Enterprise AI servers and private cloud infrastructure
Foundry, Equipment, and Packaging
$TSM - Manufactures the chips everyone else designs
$ASML - EUV lithography, no substitute
$AMAT - Wafer fab equipment for leading edge nodes
$LRCX - Etch and deposition for chip scaling
$KLAC - Process control and yield management
$AMKR - Advanced packaging and test
$INTC - Domestic foundry bet, execution still unproven
GPU Cloud and AI Infrastructure
$CRWV - On demand GPU cloud for AI workloads
$NBIS - Full stack AI infrastructure and developer tools
$IREN - Vertically integrated power and compute
$CIFR - Long-term hyperscaler data center leases
Power and Energy
$VRT - Cooling and power distribution inside data centers
$CEG - Nuclear baseload for large clusters
$VST - Dispatchable power for grid demand
$BE - On site fuel cells, behind the meter
$NEE - Renewable capacity for grid expansion
$FSLR - Utility scale solar
$EOSE - Long duration storage for 24/7 uptime
$GEV - Turbines, transformers, grid hardware
$OKLO - Next gen nuclear, early stage but Altman-backed
$SMR - Small modular reactors, longest dated play here
40+ stocks. One thesis. $650 billion has to go somewhere.
Which layer are you most positioned in?
Here's how that money flows through every layer of the AI stack:
Compute: GPUs and Accelerators
$NVDA - Blackwell shipping now, Rubin next
$AMD - MI300/MI400 giving hyperscalers a second source
Custom Silicon and Chip Design
$AVGO - Custom AI chips and networking ASICs
$MRVL - Interconnect and custom silicon for AI racks
$ARM - CPU architecture powering hyperscaler fleets
$CDNS - EDA tools behind hyperscaler chip programs
$SNPS - Design software enabling next gen silicon
Memory and Storage
$MU - HBM and advanced DRAM feeding AI accelerators
$SNDK - NAND flash across hyperscaler storage
Networking and Connectivity
$ANET - Switching infrastructure inside AI clusters
$ALAB - Rack scale connectivity silicon
$CRDO - High speed SerDes for AI interconnect
$COHR - Photonics and lasers for optical networking
$GLW - Fiber and optical materials connecting it all
Servers and Infrastructure
$DELL - AI server systems for enterprise and hyperscale
$SMCI - Liquid cooled GPU racks, first to market on new chips
$HPE - Enterprise AI servers and private cloud infrastructure
Foundry, Equipment, and Packaging
$TSM - Manufactures the chips everyone else designs
$ASML - EUV lithography, no substitute
$AMAT - Wafer fab equipment for leading edge nodes
$LRCX - Etch and deposition for chip scaling
$KLAC - Process control and yield management
$AMKR - Advanced packaging and test
$INTC - Domestic foundry bet, execution still unproven
GPU Cloud and AI Infrastructure
$CRWV - On demand GPU cloud for AI workloads
$NBIS - Full stack AI infrastructure and developer tools
$IREN - Vertically integrated power and compute
$CIFR - Long-term hyperscaler data center leases
Power and Energy
$VRT - Cooling and power distribution inside data centers
$CEG - Nuclear baseload for large clusters
$VST - Dispatchable power for grid demand
$BE - On site fuel cells, behind the meter
$NEE - Renewable capacity for grid expansion
$FSLR - Utility scale solar
$EOSE - Long duration storage for 24/7 uptime
$GEV - Turbines, transformers, grid hardware
$OKLO - Next gen nuclear, early stage but Altman-backed
$SMR - Small modular reactors, longest dated play here
40+ stocks. One thesis. $650 billion has to go somewhere.
Which layer are you most positioned in?
• • •
Missing some Tweet in this thread? You can try to
force a refresh
