a16z

The New Rule for Picking AI Winners | The a16z Show

33 minMay 29, 2026

Key Themes

AI adoptionvalue capturetoken economicsventure capitalstartup outcomespublic marketsmodel competitionAI-native companies

Summary

AI has moved from hype to scale, but the winners are still being sorted out

This episode argues that AI is already producing unusually fast revenue growth, yet adoption across the broader economy is still early. The speakers focus on how value capture may concentrate around model providers, token paths, and AI-native companies, while also examining venture strategy, bubble risk, and what a wave of large AI companies could mean for public markets. The overall message is that AI outcomes are becoming bigger, faster, and harder to predict, so picking winners now depends on understanding market structure, cost curves, and where the next layer of software value will sit.

AI adoption is still early despite rapid revenue growth

The episode repeatedly contrasts the extraordinary pace of frontier AI revenue with the fact that broader economic diffusion remains small. That means the current phase may still be early relative to the eventual market size, and adoption patterns in enterprises and consumer workflows are still forming.

Value capture may matter more than raw AI value creation

A central theme is that many companies can benefit from AI, but only some will sit in the most advantageous position along the stack. The discussion emphasizes token paths, model market structure, and AI-native product design as the places where durable advantage may concentrate.

AI-native products are not just faster software, they can change workflows entirely

The speakers distinguish between tools that simply automate existing processes and products that are built around new operating models. That distinction matters because AI-native companies may reorganize how work gets done, not just make old workflows cheaper or quicker.

High uncertainty can coexist with huge outcomes

The episode stresses that AI winners may be much larger than past software winners, while the path to identifying them is also becoming harder. Startup churn, rapidly shifting rankings, and widening valuation dispersion all point to a landscape where the upside is large but selection is difficult.

Venture firms need to adapt to companies scaling earlier

Because AI startups can become complex sooner, investors are expected to provide broader support beyond capital alone. The chapter suggests that go-to-market help, pricing, international expansion, and other platform support may become more important as companies cross major milestones faster.

The episode argues AI is not in a classic bubble yet

Rather than focusing only on high valuations, the speakers frame current AI conditions as constrained by compute, memory, data centers, power, and supply chains. In that view, the market can stay elevated without looking like a demand-fueled bubble, at least for now.

Public markets may eventually need more AI growth exposure

The discussion notes that there are relatively few public companies with true hypergrowth, which makes large AI IPOs potentially valuable for broad market participation. If that happens, AI could become more visible in index construction and public portfolios.

Select any chapter text to Deep Dive with AI

01AI Scale, Value Capture, and Token Economics

The discussion examines how AI has rapidly moved into a commercial scale phase while still being early in real-economy diffusion. It contrasts frontier model growth with broader enterprise adoption, explains the difference between skeuomorphic and AI-native products, and argues that the hardest question is which companies will capture the value created by AI. The chapter ends by emphasizing token economics, open source pressure, and the tension between declining compute costs and rising frontier demand.

Frontier AI companies are growing revenue at a pace that rivals or exceeds major hyperscalers.

Real-world AI diffusion remains low, and enterprise adoption is still concentrated in a few use cases.

AI-native companies are described as structurally different from traditional software businesses.

The conversation highlights growing uncertainty around where value will accrue across the AI stack.

Token path and model-market structure are presented as the central strategic issues.

Open source and local models are increasingly important as cost pressure builds.

02Loss Ratios, Bubble Risk, Public Markets, and Venture’s Future in AI

This chapter shifts to venture strategy in an AI era where companies scale faster, stay private longer, and create bigger outcomes sooner. It argues that high loss ratios can be normal in venture if the firm is consistently backing the best founder in a major category, and it rejects the idea that AI is currently in a bubble because supply constraints remain dominant. The discussion closes with a broader view of public markets and the future of venture, suggesting that large AI IPOs, token economics, and the model ecosystem will shape how value is distributed across the industry.

Venture investing in AI still accepts meaningful loss ratios as part of real risk-taking.

Backing the eventual category leader matters more than minimizing misses.

AI companies are reaching scale and complexity faster, requiring more support from investors and platform firms.

The speakers argue current AI conditions are supply-constrained rather than bubble-like.

Large AI IPOs could broaden public-market exposure to growth.

Future venture outcomes will depend on model labs, token pricing, and the ecosystem built on AI infrastructure.