Investor Insights: Finding the right recipe for Nvidia's AI Five-Layer Cake

Chips with everything? When Jensen Huang talks about the AI industry as a “five-layer cake”, he’s describing how the modern AI ecosystem is stacking up — from the electricity powering data centres at the bottom, all the way to AI applications used by consumers and businesses at the top.

The key point for investors: AI is no longer just about semiconductor winners. The opportunity set is broadening across the entire stack. The cake is not fully baked yet but the appetite from investors is insatiable.

Here’s a simple breakdown of the five layers, what they mean, and which companies and ETFs are best aligned with each part of the AI value chain.

1. Energy: The Foundation of AI

AI models require enormous computing power — and therefore enormous electricity demand. Training large language models and running inference workloads is becoming one of the biggest power consumption stories in modern markets.

This is why utilities, nuclear, grid infrastructure and power equipment companies have suddenly become AI plays.

What matters here

• Electricity generation

• Grid upgrades

• Cooling systems

• Nuclear and gas baseload power

• Energy infrastructure for hyperscale data centres

Stocks exposed to this layer

Constellation Energy — nuclear generation leveraged to AI power demand

Vistra Corp — power demand and merchant electricity pricing

Schneider Electric — data-centre electrical systems

Siemens Energy — grid and transmission infrastructure

GE Vernova — gas turbines and grid infrastructure

ETF idea

Global X Uranium UCITS ETF — indirect play on nuclear-powered AI electricity demand

Alternative broad exposure: iShares Global Clean Energy UCITS ETF

2. Infrastructure: Data Centres & Networking

This is the “picks and shovels” layer of AI. Before AI can function, companies need massive data centres, fibre networks, cooling systems and high-speed connectivity.

This has become one of the fastest-growing capex booms in decades.

What matters here

• Hyperscale data centres

• Networking equipment

• Cloud infrastructure

• Optical connectivity

• Cooling systems

Stocks exposed to this layer

Equinix — premium hyperscale data-centre exposure

Digital Realty

Arista Networks — AI networking backbone

Vertiv Holdings

Broadcom — networking plus custom AI chips

ETF idea

UCITS alternative: L&G ROBO Global Robotics and Automation UCITS ETF

3. Chips: The AI Compute Arms Race

This is the layer most investors associate with AI today. GPUs and accelerators are the engines powering model training and inference.

NVIDIA dominates this layer, but competition is increasing rapidly.

What matters here

• GPUs

• AI accelerators

• Memory chips

• Semiconductor manufacturing

• AI servers

Stocks exposed to this layer

NVIDIA — dominant AI GPU supplier

Advanced Micro Devices — emerging AI accelerator challenger

Taiwan Semiconductor Manufacturing Company — manufactures advanced AI chips

SK Hynix — leader in high-bandwidth memory

ASML — critical lithography supplier

ETF idea

VanEck Semiconductor UCITS ETF

iShares Semiconductor ETF

4. Models: The AI Brains

This layer refers to the companies building the actual AI models — the foundational software systems trained on enormous datasets.

These firms control the intelligence layer and increasingly compete on reasoning, multimodal capabilities and agentic AI.

What matters here

• Frontier AI models

• AI platforms

• Cloud AI ecosystems

• Enterprise AI integration

• Model distribution

Stocks exposed to this layer

Microsoft — OpenAI partnership and enterprise AI integration

Alphabet — Gemini ecosystem and AI search

Amazon — AWS AI infrastructure and models

Meta Platforms — open-source Llama ecosystem

Baidu — China AI model exposure

ETF idea

WisdomTree Artificial Intelligence UCITS ETF

Global X Artificial Intelligence ETF

5. Applications: Where AI Meets the Real Economy

This is likely where the biggest long-term economic value ultimately accrues.

Applications are the businesses using AI to improve productivity, automate work, personalise services or create entirely new products. This is also potentially the broadest and most underappreciated layer.

What matters here

• AI software copilots

• Automation

• AI-enabled healthcare

• Cybersecurity

• Customer service and enterprise tools

Stocks exposed to this layer

Salesforce — enterprise AI agents

Adobe — generative AI for creators

ServiceNow — workflow automation

Palantir Technologies — AI operating systems for enterprises and governments

CrowdStrike — AI-driven cybersecurity

ETF idea

iShares Automation & Robotics UCITS ETF

Global X Robotics & Artificial Intelligence ETF

The Bigger Investor Takeaway

Huang’s “five-layer cake” framework matters because it suggests the AI trade is broadening beyond a handful of mega-cap semiconductor names. The first phase of the AI boom was dominated by chips. The next phase could increasingly shift toward:

electricity providers,

infrastructure builders,

enterprise software companies,

and eventually the businesses that successfully apply AI to boost productivity and profits.

In other words, the AI economy is evolving from a narrow hardware story into a full industrial ecosystem. For investors, the challenge now is not whether AI matters — but deciding which layer of the cake offers the best risk-reward from here.