I've Invested in Nearly 200 Companies — AI Is Rewriting VC's Fundamentals

Over the past two decades, as an angel investor and VC, I've invested in nearly 200 technology companies. From mobile internet to blockchain, from VR to AI, I've seen countless waves rise and fall. But no technological shift has shaken the foundations of the VC industry as fundamentally as AI is doing right now.

The Three Pillars of Old VC Are Being Dismantled by AI, One by One

Traditional VC investment logic rests on three assumptions:

First, team size equals execution power. When evaluating a deal, you look at team size and pedigree. A 30-person team is "more reliable" than a 10-person one. Founders from Big Tech are "safer bets" than grassroots entrepreneurs.

Second, burning cash for growth is the only path. SaaS companies need to burn to $100M ARR before breaking even. E-commerce platforms need to burn to monopoly before turning profitable. VC money exists to be "burned."

Third, due diligence is backward-looking verification. Financial statements, customer interviews, competitive analysis — all fundamentally "looking in the rearview mirror."

AI is dismantling all three.

Team Size No Longer Matters

Midjourney: 40 people, nearly $200M annual revenue — $4.8M per capita. WhatsApp: 55 employees when acquired for $19B. Instagram: fewer than 15 people when acquired for $1B.

These are not anomalies. My own experience confirms this: at 26, with only a few dozen people, I defeated state-owned enterprises with thousands of employees. Back then, the advantage was diligence + tools. Today, AI is that "super-tool" — a 10-person team that knows how to use AI can outproduce a 100-person team that doesn't.

The iron law of VC is being overturned: don't invest in "big teams" — invest in "small teams that know how to use AI."

The End of the Burn Model

AI-era startups have a fundamentally different cost structure. Traditional SaaS companies' biggest cost is people — engineers, sales, customer support. AI companies' biggest cost is compute — and the marginal cost of compute is dropping by over 70% per year.

What does this mean? An AI-native startup can be profitable at $1M ARR — no need to burn to $100M. When your primary cost is GPUs rather than headcount, your break-even point is fundamentally restructured.

For VCs, this means: the "invest, burn, wait for exit" model is obsolete. You need to invest in companies with positive unit economics from Day 1.

Due Diligence: From Archaeology to Prophecy

Traditional due diligence looks at historical data — last year's revenue, last quarter's growth rate. But AI-era companies change so fast that last year's data may be meaningless.

When I diligence an AI company today, I focus on three questions:

1. How deep is the AI moat? Is the model just a wrapper? Has the data flywheel started spinning?
2. How high is the token efficiency? Not "how much AI are you using" but "how much revenue does each dollar of compute generate?"
3. How fast does the team learn? In an era where AI evolves week by week, what matters more than "what you know" is "how fast you learn."

A Pattern I've Only Recently Understood

After nearly 200 investments, I've noticed a pattern: the most successful founders are rarely the ones who "know the industry best" — they're the ones who "learn to use new tools fastest."

At 26, I knew nothing about retail, but I knew computers. At 28, building SMS applications, I knew nothing about telecom, but I knew data. Every time I crossed into a new domain, I relied not on industry experience but on "jumping in first and using tools to accelerate learning."

In the AI era, this pattern is amplified. The founders who excite me most today are not those who spent ten years doing AI at a Big Tech company — they're the ones who "used AI to go from zero to number one in a completely unfamiliar domain."

Token-First: The Next VC Paradigm

Internally, I've coined a concept called "Token-First" — in the future, a company's core competitiveness will not be measured by headcount, patents, or customer count, but by how efficiently you convert compute into revenue.

Token-First companies share three characteristics:

• Revenue correlates positively with token consumption, but token efficiency continuously improves
• The human-to-revenue ratio steadily declines (headcount per million in revenue halves each year)
• The moat is not data — it's "the speed at which you use data to train better models"

As an investor, the first question I now ask founders is not "What's your TAM?" but "What's your token efficiency?"

---

Twenty years of investing have taught me one thing: don't bet against the trend. AI is not just another wave — it's the operating system underneath all waves. And the VC industry is standing at the boot screen of this new OS.