Brian Manning

AI Investor · Operator · Practitioner
// Operating Thesis

The AI buildout is a five-layer cake — and I invest the whole cake. Jensen Huang frames AI infrastructure as five layers: energy at the bottom, then chips, the AI-factory infrastructure, models, and the applications where the value lands. The frosting gets the headlines; the constraints live in the base — power, memory, interconnect, packaging, and the minerals under it all. Thirty years of building digital businesses taught me the pattern: value pools at the choke points.

The Five-Layer Cake

After Jensen's five layers · book % · Jul 2026

Layer · top of cake firstWeight — tap a layer for the full map%
APPLICATIONS
AI Assistants & Chatbots
33
GOOGL9.4MSFT1.9META1.4
Agentic AI Platforms multi-step reasoning, tool use
10
MSFT1.9PLTR1.2NOW0.4CRM0.4
Enterprise AI SaaS Copilot-style integrations
9
MSFT1.9SNOW0.9NOW0.4CRM0.4SHOPMDBADBE
AI-Native Vertical Apps coding, video, drug discovery
0
TEMSHOP
EDGE & PHYSICAL AI
Autonomous Vehicles
33
GOOGL9.4TSLA3.1UBERMBLYAUR
Humanoid Robotics
8
TSLA3.1OUSTVPGTER
Drones & UAVs
0
AVAVKTOSRCAT
Edge Inference Chips
44
AAPL9.9NVDA5.9AMD0.4QCOM0.4ARM
AR / VR Devices
54
AAPL9.9GOOGL9.4META1.4
Quantum: Gate / Trapped-Ion
2
IBM0.4IONQ0.4RGTIQBTS
Quantum: Superconducting Qubits
26
GOOGL9.4IBM0.4RGTIQBTS
Neuromorphic emerging
1
IBM0.4INTCBRCHF
Optical / Photonic Computing emerging
0
SIVEFPOET
Space / Earth-Observation AI
4
SPCX1.4ASTSPLRKLB
SECURITY — CROSS-CUTTING
AI Model Security alignment, jailbreak defense
1
CRWD0.5PANW
Cybersecurity for AI Infrastructure
1
CRWD0.5PANWZSFTNT
Post-Quantum Cryptography
1
IBM0.4
Optical Network Encryption
0
NOK0.1CIEN
Hardware Root of Trust
0
INTCRMBS
AI MODELS
Foundation Models LLMs, multimodal
69
GOOGL9.4AMZN6.1MSFT1.9META1.4
Fine-tuned / Specialized Models
39
GOOGL9.4PLTR1.2INOD
Inference Serving Infrastructure
29
NVDA5.9MSFT1.9
Model Orchestration & Agentic Frameworks
10
MSFT1.9NOW0.4CRM0.4
SOFTWARE INFRASTRUCTURE
ML Frameworks training
47
GOOGL9.4MSFT1.9META1.4
GPU Programming Layer low-level kernels
23
NVDA5.9AMD0.4
Distributed Training Systems
63
GOOGL9.4NVDA5.9MSFT1.9
Container Orchestration Kubernetes
64
GOOGL9.4AMZN6.1MSFT1.9
Inference Optimization Stack
23
NVDA5.9AMD0.4
CLOUD INFRASTRUCTURE
Hyperscaler Clouds
81
GOOGL9.4AMZN6.1MSFT1.9ORCL
NeoClouds GPU-specialized
1
NBIS0.2APLDIRENCIFRCORZWULFCRWV
Edge / Inference Clouds
0
NETFSLY
Data Center Colocation
0
EQIX0.1DLR
POWER DELIVERY & DC BUILD
Grid: Transformers
8
GEV1.3ETN0.4HUBBVMIPOWLHTHIY
Grid: Substations & HV Distribution
5
PWR0.6ETN0.4HUBBPOWL
Data Center Power: UPS Systems
7
VRT1.1ETN0.4SBGSY
Data Center Power: 800V DC Bus
5
VRT1.1NVTSVICRPOWIMPWR
Server / Rack Assembly ODM / EMS
2
SMCI0.4DELLJBL
Data Center Power: Rack-Level Delivery
7
VRT1.1SMCI0.4VICRDELLMPWR
DC Build: Mechanical / Electrical Contractors
3
PWR0.6FIXIESCSTRL
THERMAL MANAGEMENT
Air Cooling legacy
5
VRT1.1NVT
Direct-to-Chip Liquid Cooling
7
VRT1.1SMCI0.4DELLNVT
Immersion / Two-Phase Cooling
5
VRT1.1
Heat Exchangers & CDUs
5
VRT1.1MODNVT
Thermal Interface Materials (TIM)
0
HON
COMPUTE HARDWARE
Training GPUs
20
NVDA5.9AMD0.4CBRS
Inference GPUs
20
NVDA5.9AMD0.4
Custom AI ASICs custom silicon
36
GOOGL9.4AVGO1.8MRVL
EDA / Chip-Design Software
1
SNPS0.3CDNS
Dedicated Inference Accelerators
3
AMD0.4QCOM0.4
Server CPUs orchestration, agentic workloads
20
NVDA5.9AMD0.4ARMINTC
Networking ASICs switch silicon
25
NVDA5.9AVGO1.8MRVL
MEMORY
HBM AI server standard
5
MU1.6SKHYYSSNLF
HBF high-bandwidth flash
9
MU1.6SNDK1.2
DRAM system memory
5
MU1.6SKHYYSSNLF
NAND Flash storage
4
SNDK1.2STXWDCKIOXY
LPDDR edge / mobile inference
5
MU1.6SSNLFSKHYY
INTERCONNECT
Scale-up: Co-Packaged Optics (CPO)
11
AVGO1.8COHR0.9LITE0.7MRVLFN
Scale-up: NVLink-style Fabrics
19
NVDA5.9ALAB
Scale-up: On-package Optical Engines
11
AVGO1.8COHR0.9LITE0.7FN
Scale-out: Linear Pluggable Optics (LPO)
4
COHR0.9CRDO0.2MRVLAAOI
Scale-out: High-speed Ethernet / InfiniBand
28
NVDA5.9AVGO1.8ANET1
Scale-out: Pluggable Optical Transport
5
COHR0.9LITE0.7CIEN
Scale-across: Coherent Optical Transport
3
COHR0.9NOK0.1CIEN
Scale-across: Long-haul DWDM
0
NOK0.1CIEN
Scale-across: Submarine Systems
0
NOK0.1CIEN
Components: External Light Sources CW lasers
5
COHR0.9LITE0.7LASRIPGP
Components: Optical Connectors / Cables
1
CRDO0.2APHAAOIGLW
Components: Optical Engines / FAU
5
COHR0.9LITE0.7SIVEFFN
Components: Fiber & Test
0
VIAVGLW
ADVANCED PACKAGING
Wafer-Level Packaging CoWoS, SoIC
6
TSM1.9AMKRASX
HBM Stacking & Integration
11
TSM1.9MU1.6SKHYY
FC-BGA Substrates server chips
0
AMKR
Glass Core Substrates emerging
0
INTCGLW
Thermal Interface Materials (TIM)
0
HON
SEMICONDUCTOR FOUNDRY
Leading-Edge Logic 3nm, 2nm
6
TSM1.9INTCSSNLF
Specialty / Mature Nodes analog, mixed-signal
0
TSEMTXNGFSUMC
Silicon Photonics Foundry
6
TSM1.9TSEMINTCGFS
Compound Semi Foundry GaN, SiC
0
WOLFTSEMNVTS
OSAT assembly & test
0
AMKRASX
SEMICONDUCTOR EQUIPMENT
Lithography / EUV + photomasks
6
ASML1.9PLAB
DUV mature nodes
6
ASML1.9CAJ
NIL nanoimprint, photonics
0
CAJ
Deposition CVD, PVD, ALD
1
LRCX0.3AMATASMIY
Etch
1
LRCX0.3AMAT
Metrology & Inspection
2
KLAC0.6AMATONTO
Burn-in & Reliability Testing
0
AEHRTERATEYY
Compound Semi Growth MOCVD InP/GaN
0
VECO
SEMICONDUCTOR MATERIALS
Silicon Wafers 300mm
0
SHECYSOIGY
SOI Wafers for SiPh
0
SOIGY
InP Substrates lasers
0
AXTIIQEPY
GaAs Substrates RF, optoelectronics
0
AXTIIQEPYSIVEF
SiC Wafers power
3
COHR0.9WOLF
Photoresist litho chemicals
0
SHECY
Specialty Gases neon, krypton, xenon
0
LINAPD
CRITICAL MINERALS — BEDROCK
Silicon base material
0
GSM
Copper interconnect, power
3
FCX1MLISCCO
Gallium GaN, GaAs
0
AXTI
Germanium fiber, compound semi
0
UUUU0.1
Hafnium / Specialty Metals high-k, beryllium
1
CRS0.4ATIMTRN
Indium InP, ITO
0
FPLSF
Cobalt energy storage
0
GLNCY
Tungsten / Antimony capacitors, strategic metals
0
UAMYALM
Lithium energy storage
0
ALBLACSQM
Rare Earths Nd, Dy magnets / motors
3
MP0.7UUUU0.1USARCRMLLYSDY
POWER SEMICONDUCTORS
Power Semis: GaN HF switching
0
NVTSPOWIIFNNY
Power Semis: SiC high-voltage
0
WOLFNVTSAOSLONSTMIFNNY
Power Semis: Power Management ICs
0
POWIAOSLTXNMPWRADI
Power Semis: MLCC & Passives
0
APHVSH
POWER GENERATION
Power Generation: Nuclear baseload
35
CEG1.5VST1.2TLN0.3NEE0.1NRGCCJ
Power Generation: Natural Gas Turbines
24
GEV1.3EQT0.4BE0.4CGEHINIOWFRDSMNEY
Power Generation: SMR emerging
5
BWXT0.1LEU0.1UUUU0.1OKLO0.1NNESMR
Solar / Storage behind-the-meter
36
TSLA3.1NXT0.1FSLRENPHSEDGTEFLNCEOSEGNRC
TICKER%= held, weight in the active mode TICKER= watchlist, or no figure in this mode

Developments

LAYER NOTES · STUB · INGEST LATER
L2
HBM is still the choke

Memory bandwidth, not FLOPs, is the near-term bind on inference. Hand note — not a news feed.

L1
Power before silicon

Interconnect and packaging matter; the constraint underneath is still electrons. Hand note — not a news feed.

Core Positions

ON THE MAP · TAP A NAME TO OPEN ITS LAYER

Contact

Investing, advisory, an operating role, or an argument about the AI trade — all welcome.

0 / 500