The AI Stack Explained: 7 Layers Driving Growth

I think we can all agree that the world is changing fast and that AI will revolutionize entire industries (if not all of them) and our way of living. How the world will look like in 2035 will be hard to imagine if you ask me. Some big questions I sometimes ask myself are:

• What does work look like? Will that even be a thing anymore?

• Will the global balance of power shift because of this?

• How does a world of abundance look like? Can we get rid of inequality?

• Will AI control all major decisions on governmental levels?

• How will daily life look like? A lot more fluid vs the now rigid?

I can go on and on. You might also wonder, how to benefit from it besides making sure to stay on top of things and take your business (or work) to a new level? In order to give you a possible better understanding of where growth will happen, we today are going to talk about investing. So I guess this is the right moment to state it clearly: this is NOT financial or investment advise.

The goal of this article is to give you a better understanding of how to look at the AI industry from a wide perspective, so you have the whole playing field in sight when placing your bets. Therefore, the companies I will name in this article are not my “picks” or any sort of advise whatsoever. Additionally, I am mostly focused on “western” markets, instead of China and others. Not all companies on the below overview might be public or investable. Always do your own research. Now let’s dive in.

The AI industry is not one thing

We have talked about it many times before. AI is not a single tool, a single company or just one solution that everyone can buy. It’s general technology that can be implemented anywhere and everywhere. What does this mean for investing? Let’s start by zooming out.

When most people talk about “investing in AI,” they immediately think of two names: Nvidia and OpenAI (which is still private and not available on public stock markets). That’s like looking at the car boom of the 1900s and only buying Ford stock. Sure, Ford was important, but the real boom touched steelmakers, tire companies, oil refineries, and even the builders of highways. Entire ecosystems rose with it.

AI is the same story. It’s not one company, it’s an economy. You can think of it as a stack with several layers and orbits:

• At the core sit the foundation models and algorithms (the OpenAIs and Anthropics of the world).
• Surrounding that core is the infrastructure: the chips that train models, the hyperscalers that host them, the energy that powers them, and the security that protects them.
• Beyond infrastructure lies the applications layer: the tools, robotics, and financial rails where AI becomes tangible in everyday life and business.

Once you see it this way, the field of opportunity gets much wider. Betting only on Nvidia or OpenAI is like buying Apple in 2007 but ignoring Qualcomm (chips), Foxconn (assembly), or the App Store ecosystem (applications). You’d be missing the bigger story.

So let’s widen the lens. To make this practical, I’ve broken the AI economy into seven buckets. Think of these as the categories where growth, disruption, and opportunity are most likely to show up.

Chips – The Foundation Layer

Every AI model is powered by chips. The faster and cheaper they get, the more ambitious AI becomes. Without this layer, the entire AI revolution stalls. The battle for chips has become a global power race. The dominance of Nvidia has scared many companies and even global superpowers. The United States has restricted Nvidia from selling their best chips to China for example.

Interestingly enough, some people claim this kind of backfired as the AI breakthroughs coming from China not only happened because of these restrictions, but it also has woken up China’s manufacturing power. For example, Huawei is rapidly ramping up its chip capabilities.

Key players:
• Nvidia – Dominates the GPU market with ~80% share in AI training chips. Its CUDA software ecosystem makes it hard to switch away.

• AMD – Closing the gap with its MI300 series, offering competitive performance at lower costs. A real second supplier to watch.

• Intel – Still struggling, but betting big on Gaudi accelerators and foundry services. If it executes, it could re-enter the game.

• TSMC – The world’s most advanced chip manufacturer. Every Nvidia, AMD, and Apple AI chip relies on TSMC’s fabs. TSMC, located in Taiwan, is a global treasury and therefore in the middle of the global power battle as China still wants (and believes) Taiwan to be a part of China. If this happens, it could have major repercussions for the Western technology world.

• ASML – Our national pride! The Dutch firm that makes EUV lithography machines. Without ASML, no one can manufacture the leading-edge chips that AI depends on. They’re the bottleneck behind the bottleneck and currently one of the most important companies in the world.

• Groq – A serious newcomer with a unique tensor streaming architecture. Known for blistering inference speeds, it shows that challengers can innovate outside the Nvidia mold. (Last minute update: they just announced that they have secured $750 million in funding, lifting its valuation to $6.9 billion as investors seek alternatives to Nvidia.)

• Tesla (xAI) – Developing its own AI5 chip, expected in 2026, designed for RoboTaxis, Optimus humanoid robots, and Tesla data centers. Musk claims it will be 40x more powerful than Tesla’s current chip. Unlike Nvidia’s products, it will likely remain Tesla-exclusive, but it’s a clear move to reduce dependence on Nvidia and vertically integrate AI hardware like Apple did with its M-series chips.

Takeaway: Chips are the picks and shovels of the AI boom. Whoever makes them cheaper and more efficient will stay upstream of every AI application. But don’t underestimate the geopolitics, this layer is where national power struggles over AI will play out.

Hyperscalers – The Landlords of Compute

Chips are the foundation, but chips alone don’t run AI. You need to connect thousands of them into giant data centers. That’s what the hyperscalers do: they stack GPUs into clusters the size of football fields, cooled by rivers of electricity.

Every time you type a prompt into ChatGPT, Midjourney, or Perplexity, that request is powered by compute from one of these giants. In practice, every consumer and business using AI is paying rent to the hyperscalers.

Key players:
• Microsoft (Azure) – Partnered with OpenAI early and integrated GPT models across Office, Teams, and GitHub. Azure is now positioned as the most AI-native cloud platform.

• Google (Cloud / DeepMind) – A research powerhouse (transformers, AlphaFold) with its own TPU chips. Rolling AI into Search, Ads, YouTube, and Workspace while serving enterprises with custom AI infra.

• Amazon (AWS) – The largest cloud provider by revenue. Pushing custom AI silicon (Trainium, Inferentia) and Bedrock, its model platform. Less consumer-facing, but critical for enterprise adoption.

• Oracle – Long the outsider, now a major player. In September 2025, Oracle signed a $300 billion deal with OpenAI to provide compute over five years, one of the biggest cloud contracts in history . This cements Oracle as a hyperscaler to watch, especially with its Stargate project (building 4.5 GW of AI data centers with OpenAI, SoftBank, and the U.S. government).

• Chinese hyperscalers (Alibaba, Tencent, Baidu) – Racing to build vast clusters for domestic AI needs. With U.S. export restrictions, they’re forced to develop local supply chains and architectures.

Takeaway: Compute comes from chips, and hyperscalers are the ones who stack and connect those chips into usable AI power. With the OpenAI–Oracle mega-deal, the market is no longer just “the big three” Oracle has forced its way into the conversation.

Applications – The Gold Rush Towns

If chips are the picks and hyperscalers are the landlords, applications are the towns springing up along the frontier. This is the layer where consumers and businesses actually see AI. Productivity tools, creative platforms, healthcare breakthroughs, gaming, legal automation, all sit here.

This is definitely the most challenging category as it’s hard to predict who will last with these speeds of change. For example, software. What will happen to software in general? In my article titled “The Future of Work Looks a Lot Like a Whatsapp Group” I argue that software might be changed by “just for you” apps and applications that just last a few days or weeks, just like we now watch content.

Some key players and interesting startups (public + private mix):
• Adobe – Embedding generative AI directly into Photoshop, Premiere, and Illustrator. • Canva – Democratizing design with AI tools for teams and brands. • Runway – GenAI-powered video editing and content creation. • Synthesia – AI avatars and video generation for training and enterprise comms. • Higgsfield – Mobile-first generative video aimed at creators. • ElevenLabs – Industry-leading AI voice generation and dubbing. • Perplexity – AI-powered search engine challenging Google’s dominance. • Midjourney – High-quality image generation with a strong creator community. • Character.AI – AI companions and character chats with mass consumer adoption. • Vibecoders (Cursor, Replit, Lovable) – AI-native coding environments and copilots.

This is where AI turns into user experience. The chips and clouds stay invisible; the applications are what people pay for directly.

Takeaway: Applications are the most exciting (and the riskiest) part of the AI stack. A few of today’s scrappy startups will become the next $100B giants, but most will vanish. For investors, the safer plays are the incumbents embedding AI into their platforms; the big upside bets are in startups with defensible moats.

I do believe a bunch of them will remain, who the winners will be is just hard to predict. I have mentioned many times before that I think the real battle for (consumer) AI will happen on the application layer as I believe the intelligence itself will become a commodity. Meaning, who will create the stickiest, easiest to use, most interesting features where users will want to pay for and stay with?

Robotics – AI in the Physical World

Robotics is where AI leaves the screen and enters the real world. Smart models plugged into capable hardware mean machines that can walk, lift, deliver, and assemble. Robots don’t get tired, don’t need breaks, and only get better with more data. As AI advances, the line between “software agent” and “physical worker” starts to blur.

Key players:
• Tesla (Optimus) – Building humanoid robots designed to work in factories and eventually in households.

• Boston Dynamics – Famous for its “dog” and humanoid robots; now testing logistics and industrial use cases.

• Figure AI – A leading humanoid robotics startup backed by major investors and corporate partners who just announced a massive One Billion Dollar investment round.

• Agility Robotics – Focused on warehouse and logistics robots like Digit.

• UBTech / Unitree – Chinese robotics companies scaling humanoids and quadrupeds rapidly.

• Nvidia (Cosmos platform) – Powering robot training in simulated 3D worlds, accelerating development.

Takeaway: Robotics turns AI into labor. If chips are the picks and hyperscalers the landlords, robots are the workforce. The cost of robot labor is falling fast, and as adoption scales, this could reshape entire industries from logistics to manufacturing. If you want to read more, I wrote a deep dive article about it a few months back: The Robot Next Door. How Close Are We To Tomorrow? (please note that I published this article on March 6th 2025, meaning some information could already be outdated.)

Energy – The Oxygen Supply

AI runs on electricity. Estimates suggest training GPT-4 consumed tens of gigawatt-hours of electricity (thousands of U.S. homes’ annual use) though exact figures aren’t public. By contrast, GPT-3’s training has been estimated at ~1,287 MWh (about 120 U.S. homes for a year). Scale that to thousands of models, plus the inference costs of billions of daily prompts, and you see why energy is the oxygen of the AI revolution. Without cheap, abundant, reliable energy, the whole system chokes.

This also makes energy a hidden bottleneck and a massive opportunity. Whoever solves AI’s hunger for power (through renewables, nuclear, or grid innovation) becomes essential to the entire stack.

Key players:

• Nuclear startups (Oklo, TerraPower, Helion, Commonwealth Fusion, Valar Atomics) – Betting on fission and fusion to deliver dense, clean power for data centers.

• Renewables (NextEra, Ørsted, Enel) – Scaling wind and solar capacity, often tied to hyperscaler long-term contracts.

• Utilities & grid tech (Duke Energy, Siemens, Schneider Electric) – Modernizing transmission and building smarter grids to handle surging demand.

• Tesla Energy – Deploying batteries and solar to complement its AI compute ambitions.

Takeaway: Chips and hyperscalers get the headlines, but energy is the true constraint. The AI revolution won’t be limited by imagination or capital, it will be limited by how much clean power we can generate and move.

Awesome interview on the possibilities of nuclear fusion power:

Security – The Seatbelt

As AI systems grow more powerful and more embedded in critical infrastructure, security becomes non-negotiable. The stakes are high: protecting valuable model weights, preventing AI-driven cyberattacks, and ensuring sensitive data isn’t leaked or manipulated. In short, AI needs a seatbelt before the ride gets too fast. (please note, this is the area I am least knowledgeable about so this is all just information, again no recommendations whatsoever).q

Key players:

• Palo Alto Networks – Expanding its platform to detect and defend against AI-powered threats.

• CrowdStrike – Endpoint protection leader now training AI models to spot anomalies at scale.

• Cloudflare – Sitting at the edge of the internet, crucial for filtering malicious AI-driven traffic.

• Darktrace – Using AI to fight AI with adaptive cyber defenses.

• Startups (HiddenLayer, Robust Intelligence, Lakera) – Focused on securing AI models themselves, from prompt injection to data poisoning.

• National initiatives – Governments are moving to classify model security as a matter of national defense.

Takeaway: If chips are the picks and hyperscalers the landlords, security is the seatbelt. As AI systems power banks, grids, healthcare, and defense, the cost of failure skyrockets. Expect this sector to become one of the fastest-growing and most strategically important in the AI economy.

Financial / Crypto Rails – The Money Layer

AI agents won’t just think, they’ll transact. Paying for data, services, or other agents requires rails that are fast, programmable, and always online. Traditional finance wasn’t built for autonomous 24/7 commerce. Or do you think AI agents will be able to open up traditional bank accounts? 😅 That’s why crypto, and more especially Ethereum, may become the financial nervous system of the AI age.

Ethereum – Positioning itself as the coordination layer for AI.

The Ethereum Foundation just launched the dAI Team, working on:
• The AI economy on Ethereum: enabling agents and robots to pay and coordinate without middlemen.
• A decentralized AI stack: keeping AI open, censorship-resistant, and verifiable. Standards like ERC-8004 (for proving agent identity) are early steps toward making Ethereum the trusted settlement layer for human + AI economies.

Bitcoin – The reserve asset of the machine economy.

While Ethereum builds the rails for programmable payments, Bitcoin stands as the neutral, non-sovereign store of value, digital hard money for agents and humans alike.

Google + Coinbase (AP2 + x402)

In September 2025, Google introduced the Agentic Payments Protocol (AP2), and Coinbase added x402, a stablecoin extension . Together, they allow AI agents not just to talk to each other, but to pay each other directly.

That means:
• A research agent can pay an archival agent per document crawled.
• A bug-finder agent can charge cents per issue flagged.
• A support agent can instantly compensate a translation agent.

In a demo with Lowe’s Innovation Lab, agents handled shopping end-to-end (research, inventory, payment, and checkout) and settled instantly in stablecoins. This is agent-to-agent commerce in action.

Takeaway: Ethereum provides the programmable money layer, Bitcoin provides the neutral reserve, and new protocols like Google’s AP2 with Coinbase’s x402 show how agents will actually transact.

Pulling It All Together

When people think of “investing in AI,” they usually picture Nvidia’s GPUs or OpenAI’s models. But as we’ve seen, the AI revolution is an ecosystem play. Chips provide the raw horsepower. Hyperscalers connect them into usable compute. Applications turn that compute into tools for businesses and consumers. Robotics brings AI into the physical world. Energy keeps the whole system running. Security makes sure it doesn’t crash. And financial rails give AI agents a way to transact.

Each bucket is its own growth story. Together, they form the AI economy. That’s the real insight: don’t look for the one big winner. Look for the categories that AI can’t function without. That’s where durable value will be created.

The AI revolution will not be won by one company or one country. It’s a stack of interdependent players. If you want to benefit from it, don’t chase the headlines—learn to see the whole playing field. That’s how you place smarter bets, whether in the market or in your business.

The future is exciting and I hope we will ride that wave together, full of joy and prosperity!

Much love,

Funs ❤️ 

Disclaimer: Nothing in this article is financial advice. The goal is simply to give you a wider perspective on how the AI economy is structured, so you can understand the playing field before deciding where to place your own bets. Always do your own research.

PS... If you’re enjoying my articles, will you take 6 seconds and refer this to a friend? It goes a long way in helping me grow the newsletter (and help more people understand our current technology shift). Much appreciated!

PS 2... and if you are really loving it and want to buy me some coffee to support. Feel free! 😉

^{Who am I and why you should be here:}

^{Over the years, I’ve navigated industries like advertising, music, sports, and gaming, always chasing what’s next and figuring out how to make it work for brands, businesses, and myself. From strategizing for global companies to experimenting with the latest tech, I’ve been on a constant journey of learning and sharing.}

^{This newsletter is where I’ll bring all of that together—my raw thoughts, ideas, and emotions about AI, blockchain, gaming, Gen Z & Alpha, and life in general. No perfection, just me being as real as it gets.}

^{Every week (or whenever inspiration hits), I’ll share what’s on my mind: whether it’s deep dives into tech, rants about the state of the world, or random experiments that I got myself into. The goal? To keep it valuable, human, and worth your time.}