AI Investing Framework: Identifying Outlier Growth Startups

Our job as investors is to try to predict the future. We pontificate over how the AI landscape will evolve and “where value will accrue” based on the current state of affairs. The reality is, in such a dynamic environment where “value capture” in AI is constantly in flux and the winning hand rotates every three months, it’s safe to assume there are fewer “knowns” than “unknowns” about the future.

This is evidenced by the fact that there’s a highly compelling bear case for every bull case. As late as Anthropic’s Series E ($61B valuation), skeptics in the market worried that the API layer would quickly commoditize, that gross margins would remain negative given compute and pricing pressures, and that labs would live in a perpetual state of high burn given exorbitant training costs that had to be constantly re-run. Skeptics worried OpenEvidence would be capped at a Doximity-sized outcome. They held unwavering conviction that inference engines would commoditize. They argued ClickHouse would never expand beyond real-time analytics and thus be confined to a smaller outcome. In aggregate, these well-intentioned skeptics will leave billions of dollars worth of returns on the table.

The hard truth

We don’t know nearly as much about the future as we’d hope. In each of the above cases, the rational argument at the time fell in opposition to the reality that unfolded (e.g. the API layer didn’t commoditize). The rational argument might land one at a coin-toss outcome at best, or an overanalyzed wrong decision at worst.

Terminal market structures in emerging markets are, by definition, unknowable. So what do we do in lieu of logic while a market is still in the making? The three options are: 1) Wait until markets settle and remove market structure or TAM as a risk, yet miss out on generational returns that accompany such risk. 2) Attempt to apply old frameworks to new markets, in turn falling into the aforementioned trap of paralysis analysis. 3) Abstract recent successes to new frameworks that help simplify decisions down the line. I’ll make the case for option 3.

The framework of Outlier Components

The following framework applies to mid- and late-stage growth companies. Series As and early Series Bs are excluded. The framework is as follows: If a company has at least one Outlier Component, engage seriously. If a company has two or more, lean in. Outlier components are defined as:

Outlier growth: They are in the top 0.1% growth of their cohort.
Outlier customer access: There is a captive or hard-to-penetrate set of relationships.
Outlier team: Not just a “great” or “persuasive” team. Not even a “technical prodigy” - there are plenty of those in Silicon Valley. The filter is: have they accomplished something that puts them in the top 0.1% of the industry?
Examples: Anthropic founders created GPT 3. RJ Scaringe built Rivian. Arkady Volozh built Yandex. Bret Taylor is Bret Taylor.

How does this compare to the status quo?

The outlier framework can at times be in opposition to traditional growth technology investing, in which one filters for tailwinds, poignant value props, and top-quartile metrics. In today’s markets, a company could score well on each of those dimensions yet lack an Outlier Component, and thus never break out.

Examples of outliers are as follows - note, these are not all examples of clearly large outcomes, as many of them have yet to exit - they are simply examples of the inputs.

\denotes Meritech portfolio company.*

Outlier growth at the time of a growth stage round:

Caveat 1: Outlier growth should be accompanied by a large vision. The point is to not overthink the likelihood of a company’s ability to achieve said vision.

Caveat 2: Outlier growth with a leaky bucket (i.e., <100% NDR) is excluded from this bucket.

Outlier customer access at the time of a growth stage round:

Anthropic (Amazon GTM partnership)
Abridge (unique Epic relationship)
Roblox* (marketplace liquidity)
Kalshi* (marketplace liquidity)
Anduril (gov’t relations)
Palantir (gov’t relations)
True Anomaly* (gov’t relations)
Castelion (gov’t relations)
Mind Robotics* (Rivian, VW relationships)
Lumilens* (large hyperscaler)
Sierra (Bret Taylor relationships)
ClickHouse (OSS base)
Vercel (OSS base)

Outlier team:

Anthropic (GPT-3 creators)
Glean (Rubrik co-founder)
Mind Robotics* (Rivian CEO/Founder)
OpenEvidence* (Kensho co-founder)
Nebius (Yandex founder)
Sierra (Salesforce Co-CEO)
SSI (OpenAI co-founder)

Note: One component missing from the above is “outlier technical moat”, largely because that is increasingly hard to come by these days. Short of SpaceX, Waymo, and Tesla, outlier technical moats are hard to find in early growth investing. Even Anthropic’s moat has more to do with scale, capital, and early-mover advantage than it does technical IP.

Guidance: Consider the company if it has at least one outlier; lean in strongly if the company has 2+ of the above. Examples of 2+ include:

Anthropic (outlier growth, customer access, team)
ClickHouse (outlier growth, customer access)
Kalshi (outlier growth, customer access)
OpenEvidence (outlier growth, customer access, team)
Sierra (outlier growth, customer access, team)

What about moats?

This is the whole point - moats are difficult to discern when market structures are in flux. Do Cursor or Cognition have moats, or are they wrappers? It depends on the enterprise context they can capture and build around as they expand their product suite. It depends on how much value accrues to the model vs. the harness. The terminal market structure isn’t yet known. In the meantime, these companies scale first and build moats after. Scale itself is often one of the most compelling moats given the second-order effects around capital availability, cost structures, and cross-sell benefits.

Why doesn’t this apply to early-growth companies?

Is this knowable for early growth companies? When companies are <$20M ARR, Outlier Components are nascent at best and often non-existent, particularly for top-down enterprise IT businesses. Predicting the future still matters for these investments.

What are the counterfactuals?

There are plenty of counterfactuals of non-outlier outcomes despite holding at least one Outlier Component. It wouldn’t be very charitable to discuss companies that didn’t work, but suffice it to say that for every pre-revenue growth round with an A+ team that works, there’s a graveyard of companies behind it that were overfunded and never found PMF.

There are similarly many outlier outcomes from companies for which outlier components weren’t visible during growth rounds. Cerebras is a great example that many growth investors overlooked.

The reality is, there is no one-size-fits-all framework for any stage of investing. 2/20 fee structures wouldn’t exist if it were that simple. Yet this is my best attempt at making sense of the current landscape.