Erik Torenberg on joining a16z: the firm's internal debate on AI defensibility and why they should have been more aggressive

Summary

Erik Torenberg is a few weeks into his role at Andreessen Horowitz — a nearly 600-person firm that has raised over $40 billion to date — and the conversation quickly moves from onboarding logistics to the firm's internal debates on AI strategy.

The defensibility question

The sharpest internal debate at a16z right now is whether AI companies are actually defensible. The core tension is between two competing moat theories: context (the more a user interacts with a product, the more locked-in they become, similar to how ChatGPT learns preferences over time) versus brand (users stick with familiar tools regardless of context accumulation). A third possibility, which Torenberg acknowledges, is that many AI application companies may simply not be as defensible as prior platform-era winners.

What makes the debate uncomfortable is that the revenue and user growth numbers across the category are, in his words, "astronomical" — which makes it hard to stay cautious. Torenberg is direct that a16z has historically been too conservative: the firm wrongly assumed OpenAI's dominance would foreclose other major winners at the model and application layers. The market has expanded instead, with winners emerging across the stack rather than consolidating around a single player.

"We should have been more aggressive"

The self-critique is pointed. Torenberg says the firm made a mistake by treating category leadership as a signal that further investment was unnecessary. The actual pattern has been multiple winners at each layer, and a16z believes there is still room for more. That view is now shaping how aggressively the firm wants to deploy going forward.

Multi-stage advantage

Torenberg flags one structural benefit of being at a multi-stage fund: missing a Series A is not terminal. He cites an unnamed example where a16z wasn't the earliest backer but became the second-largest by following on aggressively at later rounds. For founders, LP day is also positioned as a genuine business development opportunity — a16z's LP base includes large international partners, and portfolio companies with global ambitions have used the event to find direct investors or strategic partners.

Narrative management

On a separate thread, Torenberg describes an ongoing internal debate about when to respond publicly to criticism. His example: Eric Newcomer implied on his podcast that a16z's crypto fund made most of its money on speculative positions it sold off. Torenberg flatly denies this, saying the fund holds most positions and pointing to Coinbase — where Chris Dixon invested from seed through late-stage rounds — as evidence of a long-term strategy. His instinct coming in is to correct the record more often; the firm has historically leaned toward ignoring critics.