The Great Inversion: Why AI Is Making Big Tech Asset-Heavy (And Everyone Else Asset-Light)

Last week, I had attended the Singapore Fintech Festival 2025. This is something that struck me about AI adoption in businesses.

For years, business strategy had a simple mantra: Be asset-light.

The most valuable companies in the world—Google, Microsoft, Facebook—were celebrated for scaling globally without owning the physical assets. Meanwhile, traditional institutions like banks were seen as slow, capital-intensive, "heavy" businesses.

The rise of AI has quietly flipped this playbook.

Today, the companies building AI infrastructure are becoming some of the most capital-intensive businesses in history. At the same time, forward-thinking institutions like DBS Bank are becoming more efficient by strategically not owning those assets.

This is the “Great Inversion” of the AI era.

Big Tech Have Become Industrial Giants

The public still thinks of Microsoft, Google, Amazon, and Meta as software companies. In reality, they now look more like 21st-century utilities.

Training frontier models and delivering AI services on the cloud requires physical assets of unprecedented scale.

The core "capital asset" of the AI age is the GPU. An NVIDIA H100 GPU, costing over $30,000 each, is the new shovel in a digital gold rush. Hyperscalers are purchasing them by the hundreds of thousands. A single world-class training cluster costs billions of dollars. 

 

These chips require specialized facilities, the data center. You can't just plug 100,000 H100s into a normal building.  This level of compute requires special demands in high-density power and advanced liquid cooling systems.  

 

Recent earnings calls show AI CapEx (Capital Expenditure) rising to tens of billions of dollars per quarter. This isn't optional innovation spending; it's a required, non-negotiable cost to remain competitive. The companies we once called “asset-light” now depend on heavy, irreversible capital commitments.

The New Asset-Light Winners: AI Adopters

While AI builders are forced into an asset-heavy approach, the most efficient adopters are taking the opposite path. They are choosing not to own the infrastructure, rather adopting cloud-based solutions. 

 

A company like DBS would rather consume cloud services from partners like AWS and Azure, use managed AI APIs for general tasks and convert fixed CapEX into variable costs, scaling capacity with flexibility.  

 

Owning physical infrastructure for AI will not differentiate a company like DBS. What does differentiate DBS is owning customer relationships, decades of transaction histories, trusted risk models, and financial insights. 

 

By avoiding the heavy infrastructure spending, an AI adopter like DBS can spend on what actually builds an advantage. This includes fine-tuning models on their private, domain-specific data, building a world-class product experience with AI integration, and investing in cybersecurity and regulatory compliance. 

Two Diverging Moats In The AI Economy 

The AI economy is crystallizing into two distinct strategic positions. This split mirrors past industrialization: a small number of firms build the utilities; the rest build the applications.

The "asset-heavy" hyperscalers have a moat in their scale of capital investment, coming to tens of billions of dollars a year. Their risk also lies in the expensive CapEx, with billions in assets that depreciate rapidly. 

The "asset-light" AI adopters hold moats in their respective sectors, with proprietary data, domain expertise, and customer trust being important factors. Their AI spending comes as variable costs, on pay-as-you-go scalable OpEx instead. 

Conclusion

This "great inversion" is my key to navigating the hype around AI. While we all talk about the "AI Bubble" and how it may burst in the quite-near future, I believe that the huge risk is being taken on by the asset-heavy hyperscalers. They are locked in a brutal, capital-intensive war which may or may not pay off. 

Other established, dividend-paying businesses in banking, healthcare, logistics and many other sectors are not participating in this arms race. They are its prime beneficiaries, using AI to bring real, measurable business improvements in automation and efficiency. 

The AI revolution is real, and my way to invest in it is by backing the adopters who are turning the technology into tangible, boring, and beautiful profits.