Advice for investing in the AI era

Check whether AI startup revenue comes from outside the venture ecosystem before investing.

“The major risk is whether the ecosystem is an ouroboros: is all the spend just within startups?”

A key risk in AI investing is that spending may be circular—startups buying from other startups rather than generating revenue from the broader economy. The author acknowledges this concern but counters that venture-backed firms now represent about a sixth of the US economy, making the circle meaningfully larger than in the dot-com era. The shorter gap between infrastructure spend and application revenue suggests the payback cycle is healthier this time around.

Don't pass just because the idea looks bad.

“the best startup ideas seem at first like bad ideas.”

Nearly all of an early-stage portfolio's return comes from a few huge winners, and those often look unpromising at first. Investors can afford far more risk than they usually take.

Pass or Fund note: Ask "what if this works?" before "what could go wrong?".

If everyone builds on the same models, ask what makes this app different.

“applications lack strong product differentiation because they use similar models”

Looking at where generative AI value accrues, a16z found application companies growing fast but struggling with retention, differentiation and gross margins, while most of the money went to infrastructure.

Pass or Fund note: Look for an edge the model can't copy: proprietary data, a workflow customers live in, or distribution.