📺 Today’s recommended deep-dive video: https://www.youtube.com/watch?v=AiM9mZCmVPY

The AI Super-Cycle: Why This Isn’t a Bubble and How Exit Sizes Are Exploding

While many fear an AI bubble, the underlying numbers tell a story of unprecedented revenue scaling and massive supply-side constraints. We are witnessing a fundamental shift where the threshold for a “top 1% exit” has tripled in value in just twenty-four months.

Core Question: How is the rapid scaling of frontier AI models redefining the upper bounds of enterprise value and the structure of the venture capital industry?

Highlights

• Anthropic and OpenAI are adding more monthly revenue than Google, Meta, or Microsoft.
• Top 1% venture exit thresholds have surged from $10 billion to over $32 billion in just two years.
• AI diffusion into the broader economy is currently estimated at less than 5%, suggesting massive latent growth.
• The market is currently supply-constrained—compute, power, and data centers—rather than demand-constrained.

⏱️ Reading time: approx. 7 minutes · Saves you about 26 minutes vs. watching.

Want to take notes while watching? Click the image below and let AI Notebook capture the key points for you 👇

The Revenue Explosion and the Diffusion Gap

Beyond the “Nebulous Promise”

The current landscape of artificial intelligence has moved beyond the “nebulous promise” of early cloud computing, shifting toward a phase of tangible, aggressive revenue generation that is outpacing even the most established hyperscalers in terms of incremental growth. This isn’t just a trend in software; it is a fundamental reconfiguration of how value is being captured at the frontier of technology.

OpenAI and Anthropic are currently adding more monthly revenue to their top lines than tech giants like Google, Microsoft, or Meta.

This acceleration is occurring while actual diffusion into the real economy remains remarkably low, likely sitting at less than five percent, signifying that we are only at the very beginning of the massive utilization phase for white-collar automation and coding. In tech-forward sectors like software development, adoption is high, but the vast majority of Fortune 500 functions haven’t even begun to fully integrate these capabilities into their daily operations.

If you pair the current revenue scale with the fact that we are at less than five percent diffusion, the potential outcomes for the next five years are nearly impossible to overstate. The S&P 500 generates roughly two trillion dollars in profit annually; if AI can capture even ten percent of that through efficiency gains, the revenue run rate for model companies could easily reach two hundred billion dollars by the end of this year alone.

💡 Digging Deeper

Q: Is this growth driven by consumer or enterprise spending?
A: Both, but the shift in priors suggests the enterprise side is much larger than initially expected, as companies move from the documentation phase to full automation.

Q: Why is diffusion still so low?
A: Most mature companies are slow adopters; they are currently in the “context capture” phase, converting internal knowledge into markdown files to prepare for AI integration.

The New Math of Venture Exits

Re-Evaluating the 1% Threshold

Venture capital thresholds are being completely rewritten by the sheer magnitude of AI-driven outcomes, with top-tier exit values doubling or tripling at a pace that has shocked even veteran investors. The power law is becoming more extreme, concentrating value in a few “generational” companies that solve massive problems faster than their predecessors ever could.

We are looking at a future where individual AI companies could soon command valuations north of one hundred billion dollars before they even reach the public markets.

Between 2020 and early 2024, a top one percent exit was benchmarked at approximately ten billion dollars; however, by late 2024, that number has rocketed to over thirty-two billion dollars based on recent closed deals and market momentum. This ten-fold increase over a two-year period suggests that the winners of this cycle will be an order of magnitude larger than the winners of the SaaS or mobile eras.

This shift is forcing venture firms to scale their own platforms to meet the needs of founders who encounter “big company problems”—like international expansion and complex supplier negotiations—much earlier in their lifecycle. When a company like Cursor or Wiz reaches massive revenue in a fraction of the traditional time, the venture partner must act more like a global consultant than a simple board observer.

💡 Digging Deeper

Q: How does this change the “loss ratio” for VC firms?
A: A low loss ratio actually suggests a firm isn’t taking enough risk; in this high-velocity market, firms must pick the best entrepreneurs and accept that many will fail while the winners return the entire fund many times over.

Q: Are these high valuations justified?
A: While 80% of companies may be overvalued, a small subset of leaders is likely massively undervalued because they will capture the majority of the market’s ultimate profit.

Bubble or Bottleneck?

The Supply Side Defense

Unlike the dot-com era, the current market is characterized by a massive shortage of critical infrastructure rather than an oversupply of useless services. We are in a supply-constrained environment where the physical reality of hardware prevents the kind of irrational exuberance that typically precedes a crash.

The one thing that could shift us into an oversupply bubble would be a radical algorithmic breakthrough that makes models a thousand times more efficient overnight.

We are currently seeing a situation where data center capacity, compute, memory, and even basic power are so supply-constrained that most major operators cannot scale fast enough to meet the voracious demand of the frontier models. You effectively cannot get data center capacity at scale in the United States until late 2028 or early 2029, which creates a significant barrier to entry and a natural limit on speculative growth.

This physical bottleneck acts as a natural stabilizer for the industry, ensuring that capital is being deployed into actual hardware and intelligence production rather than just speculative marketing. Even if we spend five trillion dollars on capital expenditures, the potential for a one or two trillion dollar revenue return makes the investment equation look surprisingly rational for long-term investors.

💡 Digging Deeper

Q: Why is there so much resistance to building data centers?
A: Local communities often cite water usage and power consumption, though the economic benefits—like high-speed internet for schools and tax revenue—usually outweigh the footprint.

Q: Will small, local models replace the frontier?
A: While optimization is coming, the current appetite for “absolute frontier intelligence” is so high that it exceeds the cost-savings of smaller models for most high-value tasks.

Key Takeaways

The AI landscape is undergoing a “Skeuomorphic to Native” transition. We are moving away from simply using AI to do existing jobs faster and toward building entirely new types of organizations that run on “swarms of agents” rather than traditional headcount. This shift will likely lead to proactive, rather than reactive, software engagements in both consumer and enterprise sectors.

The magnitude of the current leaders is already larger than the entire Russell 2000 combined, yet the pace of value creation is still accelerating. For investors and founders alike, the primary challenge is no longer just finding a market, but navigating the “shifting sands” of a technology that updates its capabilities every few months. The winners of this decade will be those who can stay in the “token path” and leverage frontier intelligence before it becomes a commodity.

Q&A

Q1: Is the AI market currently in a bubble?
A1: No, because it is currently supply-constrained. There is a genuine shortage of compute and data centers, whereas bubbles are usually characterized by excess supply destroying economics.

Q2: How has the definition of a “successful exit” changed?
A2: The threshold for a top 1% exit has moved from $10 billion to over $32 billion in just 24 months, with $100 billion outcomes becoming a realistic target.

Q3: What is the biggest unknown in the AI value chain?
A3: The market structure of the model labs. If there are only two leaders, token prices remain high; if there are five or more, prices drop, which benefits the application layer but pressures the labs.

Q4: How does AI diffusion today compare to previous tech waves?
A4: It is still very early. While coding and tech-forward firms are high users, diffusion into the broader Fortune 500 economy is estimated at less than 5%.

Q5: Why are AI companies staying private longer?
A5: They are reaching massive scale so quickly that they encounter complex “big company” problems while their teams are still small, requiring venture firms to provide more platform support.

Q6: What is the “token path” and why does it matter?
A6: Being in the token path means your product is central to where the AI processing happens. This is crucial because buyers are currently shifting budgets from old software to cover their growing AI costs.

Q7: Will open source models eventually win?
A7: It depends on “distillation.” If it remains cheap to distill a giant frontier model into a smaller one, open source will thrive; if not, the frontier labs will maintain a massive defensive moat.