The Rule for Picking AI Winners | The a16z Show
When evaluating AI investments, focus on being "in the token path" - companies that directly process or monetize AI tokens. With Anthropic and OpenAI already adding more monthly revenue than Meta, Google, or Microsoft combined, yet less than 5% diffusion into the real economy, the biggest opportunit
33mKey Takeaway
When evaluating AI investments, focus on being "in the token path" - companies that directly process or monetize AI tokens. With Anthropic and OpenAI already adding more monthly revenue than Meta, Google, or Microsoft combined, yet less than 5% diffusion into the real economy, the biggest opportunities lie in identifying which companies will capture value as AI scales across every enterprise function over the next 12-24 months.
Episode Overview
This episode explores the explosive growth of AI companies, with discussion of how top venture capital exits have increased from $10 billion to $32 billion in just 24 months. The conversation examines the supply constraints in AI infrastructure, the shifting landscape of value capture between model companies and applications, and why we're not currently in an AI bubble despite massive valuations.
Key Insights
The AI Revenue Explosion is Real and Accelerating
Anthropic and OpenAI are adding more revenue per month than established tech giants like Meta, Google, or Microsoft. Combined, these two companies are projected to reach $200 billion in annual revenue run rate by the end of the year, yet AI has achieved less than 5% diffusion into the real economy. This suggests we're in the early stages of an unprecedented technology adoption curve.
Top 1% Exit Valuations Have 10x'd in 24 Months
Between 2020 and 2024, the threshold for a top 1% venture exit started at $10 billion. As of February 2025, it reached $20 billion. Updated data from yesterday shows it's now $32 billion for closed exits only - representing a 10x increase over just 24 months. This reflects the massive value creation happening in AI-native companies.
Supply Constraints, Not Excess Demand, Define This Cycle
Unlike traditional bubbles characterized by excess supply destroying economics, the AI market is massively supply-constrained across compute, memory, data centers, and power. Data center capacity at scale isn't available until late 2028 or early 2029. This supply constraint makes a near-term bubble unlikely and suggests sustained growth ahead.
Being 'In the Token Path' is the Critical Investment Criteria
The number one thing to look for in AI companies is whether they're "in the token path" - directly processing AI tokens and capturing that value. Cost pressure is already hitting buyers of AI technology, and companies not in the token flow will struggle to justify their existence as AI costs consume existing software budgets.
Native AI Companies Run Fundamentally Differently
The new generation of AI-native founders run their companies with extreme efficiency and aggression. Researchers are whispering into voice interfaces rather than typing, running swarms of agents, and encountering big company problems (complex business deals, supplier relationships, international expansion) far earlier in their lifecycle than previous technology generations.
Notable Quotes
"Anthropic and OpenAI are adding more revenue per month than Meta, Google, or Microsoft. And I wouldn't be surprised if the combination of those two companies is doing 200 billion of revenue run rate by the end of this year."
"Between 2020 and 2024, top 1% exit started at $10 billion. We updated those numbers in February this year, $20 billion. We just updated them yesterday. It's now at $32 billion. So, we've 10x over the space of kind of 24 months."
"I can't think of a time in my career where I have changed my mind about things at a faster clip, which is good, but is also humbling, right? Two big areas are scale and value capture."
"The world kind of changed in November as it relates to our business and I think sort of productivity in the workforce. Basically, Anthropic and OpenAI are adding more revenue per month than Meta, Google, or Microsoft. They are already at that scale of revenue getting added and actual diffusion of this technology into the real economy is tiny. It's like less than 5%."
"I feel pretty confident saying that we're not in a bubble right now. I'm less confident, you know, that we won't be in a bubble 3 years from now. But all I can speak to is where we are right now. We're massively supply constrained."
Action Items
-
1
Evaluate AI Investments Through the 'Token Path' Lens
When assessing AI companies or opportunities, ask whether the business is directly in the token processing path. Companies that monetize AI usage directly will be better positioned as cost pressures force enterprises to consolidate spending. Avoid companies that sit adjacent to AI workflows but don't capture the core value flow.
-
2
Focus on Documentation and Context Capture
The most cutting-edge companies are in the 'documentation phase' - turning everything into markdown files and maximizing context capture. Start converting your organization's processes, knowledge, and workflows into AI-readable formats to prepare for proactive AI engagement rather than reactive tools.
-
3
Back Leading Entrepreneurs in High-Conviction Spaces
In markets with multiple talented entrepreneurs building where there are clear tailwinds, pick the best founders and market leaders at the early stage. If the space works out and you have the leader, excellent. If it doesn't work out but you had the leader, that's acceptable risk. The bad outcome is when the space works but you picked the wrong company.
-
4
Plan for Faster Company Maturation Timelines
AI companies are hitting big company problems (complex deals, international expansion, large sales forces) far earlier than previous generations. Whether you're building or investing, plan for 3-5 year growth trajectories to be compressed into 12-24 months and allocate resources accordingly.