Everyone is still arguing about ChatGPT.

That's already yesterday's conversation.

The real money is now moving underneath the models—into AI infrastructure Singapore is betting on, alongside global giants deploying over $700 billion into AI infrastructure over the next few years.

While retail investors debate which AI app wins, Meta, Microsoft, Amazon, and Google are collectively committing actual board-approved capital expenditure into the layer beneath the models.

Not ideas.

Not prototypes.

Actual capex.

And almost nobody understands what that means.

The biggest wealth transfer of the next decade probably won't happen in the chatbot layer.

It'll happen in the AI infrastructure Singapore and other strategic hubs are building right now—the infrastructure layer underneath it.

The new economy is being built right now:

• power grids
• semiconductors
• memory
• cooling systems
• networking
• data centres
• energy logistics
• sovereign compute

The picks-and-shovels beneath the picks-and-shovels.

That's where the leverage is moving.

And quietly, Singapore is positioning itself directly in the middle of it.

Why AI Infrastructure Matters More Than Models

AI infrastructure is the mandatory foundation every model needs to survive. While consumer apps compete for attention, infrastructure owners capture compounding economics regardless of which chatbot wins. This isn't software—it's energy, semiconductors, cooling, and capital deployed at industrial scale. Singapore understands this distinction better than most. Every AI breakthrough increases demand for compute, electricity, cooling, and memory—making infrastructure the less crowded but higher-leverage investment layer that compounds regardless of surface-level app competition.

I Just Killed Two Months Of Work

For the past two months, Hermes Claw was my main AI agent.

It ran:

• my mornings
• my inbox
• my calendar
• my workflows
• the operational heartbeat of my company

And honestly?

Parts of it weren't scaling the way I needed them to.

I could've kept patching it.

I could've kept defending it because I already invested months into building around it.

That's exactly the trap I refused to fall into.

The sunk cost fallacy destroys builders.

Most people keep funding systems that are clearly losing because their ego is attached to the time already spent.

But the market doesn't care what you already invested.

The only thing that matters is:

from this point forward, what compounds best?

So this week, I rebuilt the stack.

Simpler.

Cleaner.

More reliable.

This is the part most people still haven't processed:

One person with the right AI infrastructure can now operate at the output level of an entire small company.

That changes the economy permanently.

The Market Is Looking At The Wrong Layer

Most people still think AI is:

• apps
• agents
• prompts
• wrappers
• content tools

That's the visible layer.

But visible layers rarely capture the deepest economics.

The internet didn't create the biggest fortunes in blogs.

It created them in:

• AWS
• cloud infrastructure
• payment rails
• semiconductors
• logistics
• distribution

AI is following the exact same pattern.

Every AI breakthrough increases demand for:

• compute
• electricity
• cooling
• memory
• physical infrastructure

Which means every new model launch creates second-order winners underneath it.

This is why NVIDIA became the symbol of the cycle.

But even NVIDIA may only be the first derivative.

The deeper money may sit beneath even that.

The New Oil Is Electricity

AI is not software anymore.

AI is energy.

Every serious frontier model requires absurd amounts of power.

And this changes the global economy completely.

Countries with:

• stable grids
• energy security
• advanced infrastructure
• political stability
• semiconductor access

suddenly become disproportionately valuable.

This is where Singapore becomes fascinating.

Singapore has almost no natural resources.

Which means Singapore survives by becoming strategically indispensable.

And historically, Singapore has been elite at seeing economic transitions early:

• shipping
• trade finance
• global banking
• oil logistics
• semiconductors
• wealth management

Now it's moving toward AI infrastructure.

Quietly.

Systematically.

Very Singapore.

Singapore's AI Bet Isn't Consumer Apps

Singapore understands something most countries still don't:

Consumer AI is noisy.

Infrastructure AI compounds.

The government is already heavily investing into:

• sovereign AI capability
• data centre expansion
• semiconductor relevance
• AI regulation frameworks
• cross-border digital infrastructure
• power resilience

Because Singapore's goal is never to become the loudest player.

It's to become the player nobody can bypass.

That's a completely different strategy.

While the US builds frontier models and China builds scale, Singapore is positioning itself as:

• trusted
• stable
• neutral
• infrastructure-dense
• capital-efficient

In geopolitical uncertainty, those become premium traits.

The Constraint Nobody Wants To Talk About

Everyone talks about AI models.

Almost nobody talks about bottlenecks.

But bottlenecks are where fortunes get made.

A few months ago, people laughed at the idea of RAM shortages.

Then the AI buildout accelerated.

Suddenly memory constraints became real.

Micron exploded.

Same pattern everywhere:

• GPU shortages
• transformer shortages
• energy constraints
• cooling shortages
• data centre scarcity

The AI economy is becoming a bottleneck economy.

And the winners of bottleneck economies are usually not the flashy consumer brands.

They're the infrastructure owners.

Why I'm Personally Watching Cooling, Power, and Data Infrastructure

The obvious trade is models.

The less crowded trade is everything models require to survive.

Every new AI system creates:

• more heat
• more electricity demand
• more inference load
• more networking traffic
• more memory dependency

That means the infrastructure layer compounds regardless of which AI company wins.

The model layer is competitive.

The infrastructure layer is mandatory.

That distinction matters.

Massively.

The Bigger Shift Nobody Sees Yet

AI is not just a technology cycle.

It's an economic reordering.

The countries that control:

• compute
• energy
• semiconductors
• infrastructure
• capital flows

will shape the next 20 years.

That's why this isn't a normal tech boom.

It's closer to the early internet era mixed with the industrial revolution.

And Singapore understands this better than most.

Small countries survive by adapting faster than large ones.

Singapore has always been elite at that.

The next decade may belong to countries that become indispensable nodes in the AI economy — not necessarily the countries building the loudest apps.

That's the real signal underneath the noise.

Everyone is watching the chatbot war.

I'm watching the infrastructure underneath civilization being rebuilt in real time.

And honestly?

I think we're still early.

Frequently Asked Questions

What is AI infrastructure and why does it matter?

AI infrastructure includes data centres, semiconductors, cooling systems, power grids, and networking hardware that enable AI models to function. Unlike consumer AI apps that compete in crowded markets, infrastructure captures compounding economics because every new AI breakthrough increases demand for compute, energy, and memory. Infrastructure owners win regardless of which chatbot dominates, making it the less obvious but higher-leverage investment layer.

Why is Singapore investing heavily in AI infrastructure?

Singapore's strategy has always been becoming indispensable rather than loudest. With no natural resources, the nation survives by positioning itself as a critical node in global systems—shipping, finance, semiconductors, and now AI infrastructure. The government is investing in sovereign AI capability, data centre expansion, and power resilience to become a trusted, stable, neutral hub that major AI deployments cannot bypass in an increasingly fractured geopolitical landscape.

What are the biggest AI infrastructure bottlenecks right now?

GPU shortages, memory constraints, cooling capacity, energy availability, and data centre scarcity are the critical bottlenecks. As AI model training and inference scales, demand for electricity, advanced cooling, high-bandwidth memory, and physical compute space outpaces supply. Companies like Micron saw explosive growth when memory became the constraint. These bottlenecks create investment opportunities in infrastructure far more predictable than betting on which consumer AI app wins market share.

Is AI infrastructure investment only for governments and hyperscalers?

No. While Meta, Microsoft, Amazon, and Google deploy the bulk of the $700B, SMEs and individuals benefit indirectly. Businesses investing in on-premise compute, edge AI hardware, or partnerships with data centre providers capture infrastructure leverage. Even solo operators using the right AI stack can match the output of small teams—shifting the economics of labour. Infrastructure thinking applies at every scale, not just at the hyperscaler level.

How does AI infrastructure differ from previous tech infrastructure booms?

AI infrastructure combines internet-era compute scalability with industrial-revolution energy and hardware demands. Unlike cloud infrastructure (mostly software and networking), AI requires massive physical resources: electricity grids, advanced semiconductors, liquid cooling, and memory at unprecedented scale. This makes it capital-intensive, geographically constrained, and politically strategic. Countries controlling compute, energy, and semiconductors will shape the next two decades, similar to how oil defined the 20th century.