All podcasts / Invest Like the Best / Summary

David George - Building a16z Growth, Investing Across the AI Stack, and Why Markets Misprice Growth (EP.450)

2025-12-02 - 64 min - source - Read full transcript

Patrick O'Shaughnessy (host)David George

Key insights

Markets systematically underprice sustained high growth because analysts cannot naturally model growth persistence.

George says above roughly 30% growth, the market still doesn't fully value the growth rate, because it is unnatural to build a five-to-ten-year model where 80% growth decays slowly (80, 75, 65) rather than quickly (80, 65, 50, 40, terminal). He cites Apple's 2009 consensus estimates for 2013 revenue being off by roughly 3x as evidence that even the most-covered companies get mispriced this way, and says the valuation gap between slow and fast growth-rate decay can be a 3x difference in outcome.

growth-investing-philosophy

a16z's growth fund entered the last year at 21x revenue for a portfolio growing 112% dollar-weighted, and George argues that is less risky than a 12%-growth private-equity buyout at 15x EBITDA.

He frames this as a deliberate contrarian stance: growth 'de-risks' an investment because it compounds optionality, whereas paying a supposedly conservative multiple for a slow grower has its own concentrated risk. The public software/consumer/fintech universe currently has fewer than five companies growing above 30%, which he uses to argue private markets now hold structurally better growth assets than most of the shrunken public small-cap universe.

growth-investing-philosophy

The AI foundation-model layer is likely to play out like the cloud infrastructure market (multiple durable, profitable winners) rather than a winner-take-all outcome, because the addressable market is enormous.

George frames the choice as 'aircraft manufacturing' (high margin, high capital intensity, few players) versus 'airlines' (brutally competitive, prone to bankruptcy) and concludes cloud - and likely the model layer - resembles aircraft manufacturing at cloud's scale: AWS, Azure, and Google Cloud are all excellent independent businesses despite intense competition, because the market is so vast it fragments into multiple large profit pools rather than collapsing to one winner.

ai-market-structure

Winner-take-most dynamics apply far more broadly across technology markets than most growth investors admit, including many enterprise categories, not just consumer network-effect businesses.

George says peers often claim 'even the number two player will be viable,' but in his experience that's usually wrong: there is no real number two to Salesforce, Workday, or ServiceNow. The one carve-out he identifies is the single dominant consumer chat interface category, where being a clear number two is the outcome he considers genuinely bad, unlike enterprise software where a strong number two can still be a good business.

ai-market-structure

The best AI application companies combine ease of customer acquisition, durable/growing (not decaying) customer engagement, and gross margins that are given temporary slack because usage growth outweighs current cost structure.

George names three concrete evaluation criteria for the current wave: organic/viral acquisition (Cursor), retention and increasing engagement over time rather than novelty decay (Harvey's usage step-change after reasoning-model improvements), and gross margin, where the firm now tolerates lower margins than it demanded of SaaS/cloud companies on the assumption inference cost will keep falling as it has even through the reasoning-model era.

ai-market-structure

George's preferred founder archetype, the 'technical terminator,' starts as a deeply technical builder and only later develops commercial instincts - and the archetype is not always obvious at the start.

His prototype is Databricks' Ali Ghodsi, who was one of seven authors of the original open-source project and was not even CEO at first; George also cites Roblox's Dave Baszucki, who reads as quiet on the surface but is ruthlessly competitive about market-cap creation, and a bridge founder Shiv, a practicing cardiologist who converted an office into a place to sleep so he could keep working. His explicit counter-example is Uber's Travis Kalanick: a non-technical, operationally ruthless founder suited to a market (fighting mayors and entrenched competitors) that rewarded raw competitive intensity over product depth.

founder-evaluation

'Poll' businesses, where the market pulls the product without being sold, are the rarest and most valuable finds; every great company has unique product or unique distribution, and the best have both because unique product creates unique distribution on its own.

George keeps a sticky note reading 'is the market demanding more of your product?' and calls it the most important evaluative question, especially in consumer. He illustrates with GitHub, which for years barely talked to customers - closing a $400k Walmart deal without a single sales call - because the product was so good it sold itself, and with ChatGPT's billion-user growth being organic and brand-driven despite having no network effect. He contrasts this with 'push' businesses like ad-driven consumer platforms, where incumbents such as Google and Facebook accumulate increasing pricing power over advertisers over time, making it progressively harder for competitors relying on paid acquisition.

ai-market-structure

The biggest mistake of the 2021 venture vintage was misreading where the industry stood in the product cycle, not misreading valuations per se.

George argues 2021 was actually a late-product-cycle period (mobile, cloud, SaaS, and e-commerce build-outs were maturing) that got head-faked by COVID into looking like a fresh cycle; the real tell of a good vintage is being early in a new technology wave, which he believes 2022-early2025 AI investing represents, more than any specific price discipline.

growth-investing-philosophy

a16z's growth fund replaced a centralized investment committee with individual trigger-pullers because committee dynamics create incentives to politick rather than fully explore investment risk.

Modeled on the firm's venture process, decisions are made by a single accountable person (his first was made over breakfast with Marc Andreessen and Scott Kupor before George had even formally joined). The stated goal is intellectual honesty and full exploration of both the risks and rewards of a deal, with explicit 'disagree and commit' norms, rather than the vote-building and pitch-refining that a formal IC process incentivizes.

firm-building-and-culture

About 70% of growth-fund dollars go into companies where a16z already has early-stage 'game film' - deep prior knowledge of the founder and company - which the firm treats as its primary source of edge.

George says roughly half of growth investments by count, and about 70% of dollars deployed, are follow-ons into existing a16z venture-stage portfolio companies. He argues growth investors mostly compete on the same forecasting and unit-economics analysis (widely available skill), so real edge comes from product, market, and people insight accumulated through years of proximity to founders before the growth check is written.

growth-investing-philosophy

a16z's growth fund deliberately avoids reserving capital for large future follow-ons, treating every big check as a brand-new investment decision to prevent 'we already committed' complacency.

Small reserves are kept only for minor, non-lead follow-on participation. For sizable investments, George says pre-reserving capital would create lazy decision-making ('oh, we reserved for it, let's do it'); instead every large follow-on, even into a top holding like Databricks, SpaceX, OpenAI, or Figma, is fully re-underwritten, which is why those companies show up as separate investments across multiple growth funds.

exit-and-portfolio-strategy

Exit and hold decisions at the growth stage are semi-qualitative, anchored on whether the founder is still running the company and whether it remains the clear market leader, layered against a hard-to-execute read on price versus intrinsic value.

George contrasts this with the venture-stage 'Fred Wilson' heuristic (hold a third, sell a third, hold a third forever), which he considers sensible only because venture entries are so early. a16z will not build a buyout fund, which George frames as a cultural mismatch: the firm's identity is backing challengers to beat incumbents, not acquiring and preserving incumbents.

exit-and-portfolio-strategy

Media referenced

Glengarry Glen Ross - movie - David George cites the Alec Baldwin 'first prize Cadillac, second prize steak knives, third prize you're fired' sales-contest scene as his working metaphor for how technology markets tend to be brutally winner-take-most.

Companies

Andreessen Horowitz (a16z) - David George leads the firm's growth investing practice; the episode is largely about how that practice is built and run.
Databricks - Flagship growth investment; CEO Ali Ghodsi cited as George's prototypical 'technical terminator' founder.
Waymo - a16z's first growth-fund investment in 2020, initially made against George's own financial analysis at the urging of Marc Andreessen and Ben Horowitz; became the episode's central case study in recognizing when a speculative technology 'starts to work.'
Figma - Two-year courtship of founder Dylan Field; internal team debate over whether the design-tool market was big enough, resolved by trusting the venture team's read on a structural market shift and George's own overnight reconsideration.
OpenAI / ChatGPT - Used repeatedly as the reference case for AI monetization, the reactive-vs-proactive product debate, and for reaching a billion users while under-monetizing most of them.
Cursor - Cited as the clearest current example of unique product creating unique distribution via viral, largely unpaid growth, plus fast enterprise pull-through ('PMF' relabeled 'product fucking market fit' internally).
Anduril - Example of a pure 'pull' business outside consumer/network-effects, driven by AI capability, operator know-how from Palantir/SpaceX alumni, and geopolitical demand.
Harvey - Legal AI company (founder Shiv described as a practicing cardiologist turned founder); cited as evidence that reasoning-model breakthroughs produced a step change in real customer engagement, not just growth-rate noise.
GitHub - Told as an anecdote: for years the company barely spoke to customers because the product sold itself (e.g., a $400k Walmart deal closed with no sales call), illustrating unique product driving unique distribution.
Roblox - Early relationship with founder Dave Baszucki, cited as an example of a quiet-seeming but ruthlessly competitive 'technical terminator' founder, and of a poll business with two network effects.
Salesforce - Used as the incumbent case study for how AI could displace sticky, database-driven SaaS via a reimagined proactive UI, new unstructured data sources, and a new business model.
Uber - Travis Kalanick cited as the counter-example to the 'technical terminator' archetype: a non-technical, ruthlessly competitive operator suited to a fight-the-incumbents, regulatory-heavy market.
CrowdStrike - George Kurtz cited as another example of a technical founder who became an excellent business operator.

Techniques and frameworks

Technical terminator - George's preferred founder archetype: someone who starts as a deeply technical builder (often not the original CEO) and later develops strong commercial and management instincts; contrasted with the 'ruthlessly competitive operator' archetype like Travis Kalanick.
Push vs. poll businesses - George's core evaluative lens: is the market demanding more of the product organically ('poll'), or does the company have to go sell it ('push')? Poll businesses tend to compound; push businesses tend to get harder to sell into as they scale.
Single trigger-puller decision process - a16z's growth fund replaced the traditional growth-equity investment committee with individual decision-makers, mirroring the firm's venture-stage process, to force intellectual honesty and avoid consensus-softened bets.
Game film - George's term for the deep, multi-year knowledge a16z accumulates on founders and companies through its early-stage relationships, used to de-risk growth-stage decisions; roughly 70% of growth-fund dollars go into companies where the firm has prior investment history.
Zero reserving for large follow-ons - a16z's growth fund treats every large follow-on investment as a fresh decision rather than pre-reserving capital, to avoid lazy 'we already committed' thinking; small reserves are kept only for minor non-lead follow-ons.

Summary

David George, who runs Andreessen Horowitz's growth investing practice, walks Patrick O'Shaughnessy through both his macro view of the AI platform shift and the granular mechanics of how a16z's growth fund actually operates. On the macro side, George is careful to separate what he's confident about from what he isn't: he believes the consumer AI opportunity is enormous and still barely monetized (ChatGPT users spend roughly 30 minutes a day in the product versus 50 on Instagram and 70 on TikTok, and only a small fraction of its billion users are monetized at all), but he doubts the chat interface is the durable form factor, expecting AI products to shift from reactive to proactive with long-term memory. On enterprise AI business models, he is openly skeptical of the popular "software is $400 billion but white-collar labor is trillions" slides, noting that only a couple of categories - task-completion pricing in customer support, consumption pricing in coding - have actually found a defensible new business model so far.

A large section of the conversation is a case study of Waymo, which a16z first backed in 2020 against George's own financial modeling, at the insistence of Marc Andreessen and Ben Horowitz, who argued the category itself ("the mother of all markets") outweighed near-term unit economics. George recounts the 2019 test ride where the car handled unprotected lefts but couldn't park, and contrasts that slow, narrow-domain problem (stay in lane, avoid collisions, obey speed limits) with the much higher degrees of freedom required for general-purpose home robotics, which is why he expects robotics to take meaningfully longer than the LLM crowd currently assumes. He uses Waymo's eventual inflection - just 400 cars achieving the effective coverage of 10,000 due to route optimization and full utilization - as his template for recognizing when a speculative technology "starts to work": not projected economics, but observable customer pull.

On founder evaluation, George's central archetype is the "technical terminator" - a deeply technical founder (Databricks' Ali Ghodsi, Roblox's Dave Baszucki, Figma's Dylan Field) who later develops strong commercial instincts, contrasted with a pure operator archetype like Uber's Travis Kalanick, suited to markets that are won by raw competitive intensity rather than product depth. His investment philosophy is summarized as "pay fair prices for great companies," with the differentiating skill being recognition of underpriced greatness through product, market, and people insight rather than financial modeling, which he considers a commodity skill in growth investing. He is explicit that markets structurally mis-model sustained high growth (citing Apple's 2013 consensus estimates missing by roughly 3x) and that this mispricing is the core edge behind a16z's willingness to pay 21x revenue for a 112%-growing portfolio.

The back half of the episode covers how a16z's growth fund is actually run: a deliberately small ten-person team that relies on "game film" from the firm's much larger early-stage practice (about 70% of growth dollars go into pre-existing portfolio companies), a single-trigger-puller decision process built to avoid committee politics, zero capital reserving for large follow-ons to force fresh underwriting every time, and a promotion system that holds even junior investors accountable for contributing to collective investment judgment from day one. George also describes the two-year courtship of Figma's Dylan Field, where the venture team's structural read on the design-to-engineering market ultimately overrode the growth team's narrower designer-headcount-based sizing model, and frames this as the central tension growth investors must resolve: quantitative market-sizing discipline versus qualitative conviction earned through years of proximity to the founder and market.

Notable Quotes

"The best business in the world don't have customers, they have hostages." - David George (quoting a16z partner Alex Rampell)

"Is the market demanding more of your product? It's the most special thing when it happens." - David George

"90% of the technological surplus is going to go to the end users. Just start with that as the assumption, whether it's consumer, whether it's enterprise." - David George

"You need those product and market insights or you're just going to live in a spreadsheet and die in a spreadsheet." - David George

"We are the Yankees, and we're going to act like it." - David George, describing a16z growth fund's internal culture principle