By Esther Wong

Ever since the Industrial Revolution introduced machines to replace physical labor, the scarcest resource in business was human intelligence. Good ideas were exceptional, good managers were rare, great workers were expensive, and decision-making simply did not scale. 

That is now changing. Drastically, and fast. 

We are entering a world where intelligence is becoming abundant and cheap. Whether it is intelligence in the form of coding, analysis, design, writing, customer support, or even strategic planning, the cost is plummeting. As that happens, the entire logic of investing must change. 

The old world was built on cheap tools and expensive minds; the new world is built on cheap tools and cheap machine minds. That sounds exciting, but it is also profoundly deflationary. If everyone can build, then more people will build. If everyone can think faster, competition accelerates, compressing margin to no end.

The Great Compression and the Death of Human Inefficiency

In the last era, many great businesses were essentially organized intelligence. Consulting firms sold expert opinions (always questionable to me but hey…people do pay for it), software companies sold workflows encoded by engineers, services companies sold human coordination. White-collar labor was the moat.

Most, if not all, of our current businesses only exist because humans are slow, inconsistent, and prone to error. They make money from essentially a life-size excel file -  handoffs, queues, approvals, formatting, summarizing, routing, and reconciliation. Once intelligence becomes abundant, much of that friction will get designed out entirely.

We are moving from legacy processes to purely AI-native work, where workflows are redesigned around outcomes rather than around what humans can reasonably understand or pass to one another.

Take insurance industry. A legacy software stack helps human claims handlers process files more efficiently. But an AI-native company like Lemonade does not use AI to help an adjuster work faster; the AI is the workflow, ingesting data, reasoning, and executing the payout in seconds. Or consider coding. Claude helps engineers write code faster, but it also act as autonomous software engineers, shifting the paradigm from "AI that answers questions" to "AI that builds systems".

The Application Trap

This fundamental shift is why at 3C AGI Partners, we do not touch application-layer startups that simply use AI to make existing human procedures "faster or easier."

Why invest in a software wrapper designed to lubricate a human workflow that is fundamentally obsolete? The old software pitch was: "We help the analyst do in two hours what used to take eight." The AI-native pitch is: "Why is there still an analyst in the loop at all?"

Why “Good Companies” May Get Cheaper

This is the uncomfortable part for traditional investors: many companies may become operationally better because of AI, yet financially less valuable. 

If AI makes a company 50% more efficient, that is good. But if AI also enables 100 new competitors to enter the same market with the exact same capabilities, that is bad. We are already seeing early tell-tale signs of this in China, where everyday people are rapidly embracing AI, and we believe this will soon be a worldwide phenomenon. 

In economics, when an input becomes abundant, the value migrates away from the user of the input and toward whoever controls the scarce bottleneck. 

The New Scarcity Map

If intelligence is free, the moat shifts to things AI cannot easily commoditize. That usually means one or more of the following: 

• Proprietary data and distribution

• Brand, trust, and regulatory permission

• Physical infrastructure and manufacturing capability

• Energy, compute, and advanced chips

• Real-world integration and hard scientific know-how 

In other words, the winners in the AI era are less likely to be companies that merely use intelligence, and more likely to be companies that control the scarce assets around intelligence.

Why Our Investment Thesis Makes (even more) Sense

This brings me to our core philosophy at 3C AGI Partners: we focus on the technology that builds a better AI factory. Our thesis does not break in a world of zero-cost intelligence, rather, it becomes more relevant than ever. 

Take compute and chips: If AI demand explodes, the bottleneck is not another chatbot. The bottleneck is the hardware, systems architecture, memory bandwidth, and networking needed to run intelligence at scale. That is why infrastructure names matter, and why backing custom AI inference accelerators like Cerebras and Groq fits the exact logic of the moment.

Take energy and industrial systems: If intelligence becomes close to free, the world will use vastly more of it—but intelligence still runs on electrons. The constraint moves downstream into power generation and energy density. This is why a picks-and-shovels approach to frontier energy, like backing commercially viable fusion companies such as Kyoto Fusioneering and Type One Energy, makes sense when tied to real system bottlenecks. 

Take unconventional infrastructure: If the future requires massive compute and resilient data processing, platforms that expand where computing can happen become strategically vital. Space-linked data infrastructure, like Starcloud building data centers in orbit, is not science fiction cosplay—it is the ultimate expression of the physical scarcity thesis. 

Conclusion: How and Where to Invest?

Not everyone cares for deep-tech or infrastructure investing, but the rational investor must move up one level of abstraction and ask,

• Does this company benefit when AI becomes universal?

• Does this company benefit if open-source model reaches parity with proprietary ones? 

• Does this company sell “smartness” or “results and trust”?

• Does this company own the NEXT bottleneck?

• Does this company get stronger as competition increases?

That is where the money will be made. In a world flooded with synthetic intelligence, the premium will go to whoever can bring results, and to us, that hinges on whoever owns what thinking needs.