VC Contagion: Is Venture Capital Killing Itself?

Summary

The earliest of Kyle’s podcast appearances in the corpus (May 2023), on Eric Newcomer’s show, and the one that most reads like a founding document — nearly every argument he develops over the following three years is already here in first draft. Built around the essay VC Contagion, which Kyle had shared with Eric but had not yet published (see the dating note below).

The mood, and the opportunity inside it. Eric opens from a Slow Ventures conference full of young VCs anxious about their future. Kyle’s own writing, he says, “started as philosophical questions and has mounted into existential dread over the last 18 months.” But his read is not despair — via a Don Valentine line he quotes in The Unbundling of Venture Capital: when there is disruption there is confusion, and when there is confusion there is the most significant opportunity. “For people who actually want to do things differently and create new products for founders, I think there’s actually a lot of opportunity.” What he doesn’t think survives is “a couple of white dudes in a room making all the decisions based on gut and influence.”

Two trends people conflate. The first is AUM expansion, from The Blackstone of Innovation: picture any income-inequality chart and apply it to venture. Capital agglomerators build their business model — stated or not — around aggregating as much capital as possible, which becomes a fee business. His tell, and it recurs across the corpus: when big PE shops go public, the market discounts carry almost entirely and prices them on fees. Kyle’s constraint is the arithmetic: “your fund size is your strategy.” You cannot earn 5–10x on multiple billions; the power law doesn’t operate at that scale. Which is fine, because those LPs don’t need it — deploying hundreds of millions into one fund, 2–3x is an acceptable outcome. The second trend is momentum investing — firms rushing into AI — against genuinely thesis-driven shops. He’s careful that these are different phenomena that get blamed on each other.

The sustainability argument, via Bill Gurley. Gurley’s 2016 line: raising $30M before an IPO used to be remarkable; now it’s a Series A. That capital produces “voraciously hungry unicorns” — and venture’s real distortion is that it lets a company stay unsustainable for a very long time, then demands an overnight shift. His example is Uber: Dara saying the company needs to step back and check whether the unit economics work before going big, at which point Kyle asks the obvious question — “what is the definition of big if not Uber?” The underlying gospel is Paul Graham’s since 2013: if more money won’t help you, don’t raise it. Eric reframes it as does blitzscaling work outside a zero-interest-rate environment, and can you come back from it if you did it?

On Uber specifically, an unusually careful verdict. “If the game is Hungry Hungry Hippos — open your mouth as much as you can and take as much as you can — Uber’s crushing it.” And in a commodity market, which ride-hailing became once the black-car-versus-your-buddy’s-car distinction collapsed (“remember the pink mustaches”), spending like a drunken sailor to be the only name people look for is a defensible strategy. But roughly $30B of losses since IPO, before counting everything spent before it: “was that the best capital allocation output equation that we’ve ever seen? I feel like the answer is probably no.” Early investors and early employees did great. The question is the whole-life return on capital, not the early slice — which is also his read on the Midas-List canon of Coinbase, Uber, WeWork and Snowflake: “some of the big wins that define the venture returns don’t even sustain.”

A generational taxonomy of companies. The Facebook/Google/Microsoft/Apple near-monopolies are “a class unto itself” — so rare that including them distorts any comparison. Then the first-generation SaaS companies — Adobe, Workday, Salesforce — which he calls, warmly, “your grandpa’s SaaS companies… just really good salt-of-the-earth companies,” and then asks whether anyone builds them anymore. His counterexample is Twilio: a long-time darling whose profitability was always structurally capped by carrier fees and low gross margins, always defended with at scale it’ll be better, which never proved true. Even Snowflake and Datadog, which he calls exceptional, prompt the question of whether they needed that much capital to get there. And the closing note: he loves Alibaba’s hundred-year vision, and “nobody in Silicon Valley has a hundred-year vision. Nobody talks that way. I really wish we did that more.”

Product-led venture, and the everything-store problem. The question every firm should answer: what is the job to be done that a founder hires your money for? He presses it on a16z — name a company they’d say isn’t a fit — and the answer is essentially none, because they’ll raise a fund for it. Large firms have solved this with fiefdoms inside the empire (crypto, American Dynamism), but Kyle’s worry is that when you try to be everything to everyone it’s hard to be very good at anything specific. He does concede that Marc Andreessen’s borrowed credibility is a real product: if the logo closes customers even when the partner never calls you back, that is a value proposition. He singles out Index Ventures as one of the most focused top firms — a growth fund that isn’t $10B, whole categories skipped — and quotes partners telling founders: “if you want somebody who’s going to write you a massive check and never call you, awesome, there are firms who will do that. That’s not us.”

How Contrary actually works. Identify the sharpest people as early as undergrad, then be useful across their whole career — scouting, first jobs, comp negotiation, placement — so that by the time they found something, “sometimes we know people for five, six years before they ever start a company.” Contrary Research was born from a recurring question inside that ~500-person community: I have offers from three private companies, what’s the investor’s perspective on these businesses? Kyle joined to build the Series A+ practice, because the signal works in the other direction too — five of the first fifty employees at Ramp came from Contrary, six to eight at Retool, a handful at Anduril — so where sharp people go is itself an investment signal. Funds around $100M, leading pre-seed and seed, participating later.

The best practical advice in the episode is about accepting a lower valuation: the higher price is “a weight around your neck.” Run it forward — that raise buys you eighteen months to two years, so what if six things go wrong? Because in startups it’s usually not six things, it’s twelve. Can you still hit milestones that support a markup? If not, take the dilution now. And he argues the partner matters more than the number, because “at every firm I could point to a handful of people and say: if you are doing this thing in this space at this stage, that is who you want in your corner.”

The information economy — the most original stretch of the conversation. Eric’s frame is that VC runs on legal insider information, and that these are loosely held secrets: if a founder tells fifty investors, the fact is effectively public to an elite class and withheld from everyone else. Kyle splits it. Some privacy is legitimate and worth defending — a customer list with names and prices, in a company still figuring out its product, shouldn’t be handed to competitors. What he objects to is the gossip economy, where status accrues to knowing who did what, and which “reinforces dramatically this feeling of FOMO.” His illustration is a Redpoint partner’s clip about asking a founder post-close who else gave them term sheets — “oh, nobody else cared about this thing” — where the rumor doesn’t just inform the decision, it makes you question it afterward. Against that he sets the public markets, where you wear your pride and your shame in the open: a 13F is your thesis, and Cathie Wood selling most of an Nvidia position has to be worn publicly. “That’s a different vibe than in venture, and I think that is unhealthy.”

On whether elite firms have a real data edge, he’s deflationary. There is constant information leakage — a company with a hundred customers has a hundred people who know things, which is the entire premise of Tegus — and he’ll sometimes relay unflattering customer feedback back to a founder. The magic-eight-ball data mostly doesn’t exist: GitHub stars can be bought the same way Twitter followers can, so using them to decide is “probably making a bad decision.” What is genuinely unequal is purchased data — credit-card panels mapped onto private companies, at six-figure contracts only the largest AUM firms carry — and the fact that anyone with a Pitchbook login can see returns for funds whose LPs must disclose. Asked whether more disclosure would be healthier, he says yes — “being a publicly traded company is a good way to force you to be healthier” — while granting the obvious limit: an early-stage company changes weekly, and “I don’t know if I’m gonna make payroll next week, I’m not worrying about those forms.”

On AI, the line that names the whole problem: “AI fervor and the dynamics of building AI companies is a match made in heaven for the capital agglomeration industrial complex.” If you’re trying to raise an enormous amount and deploy it into things that consume enormous amounts, AI is a very good place to do it. He expects a repeat of the 2021 hangover in a lot of these companies — while noting training costs are falling step-function and Sam Altman has already said the next generation isn’t about the largest model.

He is emphatic about not being read as a skeptic — his mother tells him he caveats too much — “I am an optimism maximalist through and through.” The worry is specific: concentration of power. It’s why he finds OpenAI’s enthusiasm for regulation unnerving (“they’re going to be in the room where it happens, helping write the regulation that protects them”), and why he describes OpenAI as, in some ways, a ridiculously successful subsidiary of Microsoft — an AWS-like business that happens to live inside a corporate parent, which is a different thing from a startup success story. On the product itself: ChatGPT is “good at a lot of things, great at basically nothing” — which he immediately concedes is a high bar, since “every human being sucks at most things.” And on the name: “OpenAI is the most open — they really need to take it out of the name.”

He also pre-announces The Openness of AI here, and gives its central finding early: every time OpenAI explains closing up, safety comes first and competition second, almost without fail.

What he’s actually looking for in AI is not a spike but a functional funnel: open-source engagement and traction at the top (real, but on its own a trap — he heard 2021 valuations quoted as a multiple of Discord users, which he calls “deeply unhealthy”), then genuine contribution, then an enterprise-grade product someone will pay real ACVs for. Hugging Face he rates a phenomenal business that has nailed the first two rungs, while noting the third isn’t yet clear — and that GitHub took a long time and ultimately needed Microsoft’s commercialization DNA to make its long term make sense. His worked example of the full funnel is Nomic: GPT4All ran open-source traction comparable to LangChain, but the business underneath is Atlas, the data-visualization tool they used to build it — including the Stable Diffusion portrait hanging in Contrary’s New York space, with its Kermit the Frog corner. (He tells the same story five months later on The Centralization of Power in AI.)

Transcript

The Newcomer Podcast, ~64 minutes, hosted by Eric Newcomer. ASR errors and names cleaned, text paragraphed; nothing reordered or summarized. Corrections applied: Coatue, USV, NEA, Andreessen Horowitz, Katherine Boyle, Amjad Masad, Nomic, GPT4All, LangChain, Tegus, Cathie Wood, Twilio, Dara Khosrowshahi, SPVs, ACVs, 13F, mea culpa. Sponsor reads are marked rather than transcribed. Two names the audio left uncertain are marked [?].

Intro

Eric Newcomer: Hey, it’s Eric Newcomer — welcome to the Newcomer podcast. This week I have venture capitalist Kyle Harrison from Contrary, and we talked about the doom and gloom in venture capital. He’s written a piece called VC Contagion: Is Venture Capital Killing Itself — you can imagine how enticing I found that topic.

I’d just gone to the Slow Ventures conference, where there was a lot of fear among venture capitalists — young VCs earlier in their careers — about their future, amid doom and gloom everywhere outside of AI. And I think everyone is very aware that there’s a lot of hype in AI that feels perhaps unsustainable, so there’s a lot of soul-searching about that.

Kyle’s a former investor at Index and Coatue, and we dug into information sharing and private information in the private markets. Contrary publishes reports on private companies, so in some ways they’re a kindred spirit in terms of exposing information about startups to the world — at the same time as, being a venture capital firm, they give private advice to their startup founders. So we talked about how he balances that, and also got philosophical about the differences between the public and private markets. And then finally we talked about artificial intelligence in earnest, and interrogated how to find opportunity in a moment of so much hype and so much large spending from hyperscaler companies like Microsoft and Google.

[Sponsor read — Vanta.]

How screwed is VC?

Eric Newcomer: I was just at a Slow Ventures conference in New York with a bunch of up-and-comers — the concept was the partners doing the work, the more junior people hustling — and it definitely felt like the nature of VC itself was in question. Which, if you’re a young professional, you never want to hear. You’ve been writing on this topic. How screwed is VC right now?

Kyle Harrison: I’ve been writing now for a little over a year. It started as philosophical questions and has mounted into existential dread over the last 18 months — which was very timely. I was just kind of thinking about stuff, and then it’s become very prominent, with lots of people thinking about it.

My very first piece was this idea I called the unbundling of venture capital — one of the first things I wrote that took off. In it there’s this really good quote from Don Valentine, where he’s talking about companies, but he says: when we invest in companies we look for specific points of disruption, because when there is disruption in a business model or an industry there’s confusion — and when there is confusion, there is the most significant opportunity.

I actually feel that way about venture now. I do feel like there is opportunity for people who want to be creative. For people who are like, no, I’d really like to keep it the couple of white dudes in a room making all the decisions based on gut and influence — well, I don’t think that’s going to stick around. But for people who actually want to do things differently and try to create new products for founders, I think there’s a lot of opportunity there. There is, in the interim of disruption, a lot of that existential dread.

Is the pie growing?

Eric Newcomer: A classic thing in VC is to say there’s too much money, people don’t want to pay the high prices — you’d rather be the only investor with all the time in the world. When I started covering VC it was Andreessen Horowitz coming out of the gate paying too much, then SoftBank and Tiger. Some of those were wrong prices, but overall VC weathered it and kept going, so you can grow fatigued of the is there too much VC money question. But now, with a real winter in everything but AI, it feels like maybe there is. Do you think the pie is going to grow or shrink?

Kyle Harrison: There are two big trends that are correlated but not exactly the same thing, and sometimes people conflate what’s happening because of one with the other.

On one side — and I wrote a piece called The Blackstone of Innovation unpacking this — you have AUM expansion. Picture any chart you’ve seen about income inequality in the US, that sort of the richest get richer. There’s a function of income inequality, so to speak, in venture, where I call them capital agglomerators: firms who have built their business model, whether they say it explicitly or not, around aggregating as much capital as possible. And it largely becomes a fee business.

Somebody was talking about this recently — when you look at some of these big PE shops that have gone public and how they trade, people basically discount carry entirely. There’s some aspect of it, but public markets caring about a PE firm as a publicly traded company — they don’t super care about carry. It is essentially based on the value of their fees.

Eric Newcomer: At some point LPs wise up — if you’re a terrible product they stop paying those fees. That’s another trend in family offices, people sick of fees on fees on fees for a tiny slice. But for people with products — Andreessen Horowitz probably has something like $500 million a year in management fees. NEA raising a $6 billion fund where the message seems to be slow, steady, reliable — consistent more than astronomical returns, unlike a USV. Lightspeed has raised a huge fund, General Catalyst is enormous now. And you talk to smaller VCs and it’s is that even VC anymore? If you’re investing in Stripe at $50 billion like Thrive did — is that venture capital?

Kyle Harrison: There’s a saying that your fund size is your strategy. Anybody managing ten-plus billion dollars — you can’t earn 5–10x returns on multi-billions of capital. That just doesn’t happen. It’s taking this old framework where you have a small venture fund and if you hit it big there’s the power law, where you can have 10x, 1500x returns. That doesn’t work when you’re allocating billions and billions.

Because even the hits — I remember the sine waves of the Information criticizing Andreessen because their returns were so-so, and then Coinbase went public and they returned their entire AUM on the Coinbase position alone, and now that’s come down. But they did sell.

It’s just different. It’s not public equities per se, it’s not venture classic — it’s something in the middle. The term growth equity has been bandied around and misused, so I don’t know the perfect way to articulate what the strategies are. I tie it more to what are the return focuses. When you look at these massive capital agglomerators, they’re totally okay with 2–3x returns, because people trying to deploy hundreds of millions of dollars into one fund at one time are fine with 2–3x. They don’t need 10x on that size of capital.

Eric Newcomer: So to distill it — there are these huge funds, that’s one vector. The other is the momentum-investor category: firms rushing into AI, right or wrong. Spark clearly wants to be known for paying high prices for AI, believes in it — that’s a return of momentum investing. And on the other hand you have the truly contrarian, thesis-driven Lux posture. How do you see that overlaying on the capital piece?

Voraciously hungry unicorns

Kyle Harrison: Bill Gurley had this line I quote a lot — I quoted it in the piece I shared with you. It’s from around 2016, but he talks about how in the ’90s, if a company had raised $30 million before they went public, that was crazy. Now $30 million is a Series A in some cases.

All this capital going into all these companies — and this has only gotten worse since 2016 — makes these just voraciously hungry unicorns that have built their models around unsustainability. In that same piece I talk about this sine wave of how long you can maintain unsustainability, and venture has allowed that to be prolonged. You can stay unsustainable for a really long time.

And then you get to the point where — Dara at Uber, with whatever, $30 billion of revenue, saying a year or so ago: hey, this next phase is going to be different, we need to take a step back and figure out if our unit economics really work before we go big.

And it’s like — right. What is the definition of big if not Uber?

Venture has allowed these companies to remain in unsustainable territory for a really long time, and then all of a sudden overnight they have to make this shift. Especially when they’re big, they don’t shift like that.

So the outcome for venture is going to depend on: we have a generation of really cash-hungry businesses. Can they turn off the cash hunger and still continue to grow and still be big outcomes? And if they do turn off the cash hunger and go on a diet and still become big outcomes — that’s the real existential question. Oh crap, you’re telling me we didn’t need to funnel hundreds of millions of dollars into a company to get a multi-billion-dollar outcome? They could have done that anyway.

Which is basically the gospel Paul Graham has been preaching since 2013: there are going to be situations where more money is not going to help you, or you could grow without it — in which case don’t raise money, or definitely don’t raise a ton of money. And everybody’s just been like, that’s just what you do.

Eric Newcomer: I’d almost put it as: does blitzscaling work in a non-zero-interest-rate environment? And if you were blitzscaling, can you come back from it? That’s the question with the Ubers and Airbnbs — VC made businesses possible that maybe were impossible, but are they sustainable now that they’ve been made possible? Bill Gurley loves the SNL sketch where a bank gives you two dollars for a dollar — people will take that trade all day. Drivers are rational. If you pay them more than is sustainable, for a while it’s a great business and the top line looks amazing.

On the other hand, Lyft is dying — crashing, clearly trying to sell, tiny market cap, founders gone. That could give Uber an opportunity to be a nice monopoly business. I personally think Uber has gotten to the point where it’s working on some level. Do you disagree?

Kyle Harrison: I find myself all the time talking about Hungry Hungry Hippos. If that’s the game — just open your mouth as much as you can and take as much as you can — Uber’s crushing it at that.

And I think it’s also an argument for a commodity market. For a long time, 2013–2014, investors in Uber and Lyft used to argue it was not a commodity, that they offered very different experiences — the black car versus your buddy with a car. Remember the pink mustaches. They tried to be super different products, and they have entirely commoditized. So if you’re going to chase a high-volume, low-value commodity, there is an argument for: spend like a drunken sailor, sprint as fast as you can, get everywhere as quickly as possible so you’re the only thing people look for in that market. They did that really well.

Did they need to? They’ve lost $30 billion or whatever since they went public — not even including all the money they spent before. That’s a huge amount of capital for the value they have. Does the long-term blended return-on-capital equation make sense? I’m not sure.

The early investors made a crap ton of money — for them, perfect. For the LPs in those early funds, perfect. For the early executives who made a ton of money, perfect. It’s the greater fool game: who gets left holding the bag? Is Uber going to stick around forever as a long-term sustainable company? It hasn’t yet proven it can repeatedly produce profitable results. I think it could get there, and if Lyft goes away that makes it easier. So I’m definitely not out shorting Uber. But there’s a broader question of was that the best capital-allocation output equation we’ve ever seen — and I feel like the answer is probably no.

Eric Newcomer: And some of the core Midas-List-driving investments — Coinbase, Uber, WeWork, and Snowflake is obviously a great investment but falls dramatically — so much of the narrative in venture gets written and then these companies don’t sustain.

Kyle Harrison: There’s a generational shift in the kinds of companies that got built. And part of the difficulty is that a lot of the best companies in tech built over the last 20 years were very rare, borderline monopolies — the Facebooks and Googles and Microsofts and Apples. Those are difficult for me to put in the same bucket, because they’re so rare that the likelihood of building one is astronomical even for venture. That’s a class unto itself.

Then you have the Adobes and Workdays and Salesforces — first-gen SaaS products. I view those as solid, sustainable, your grandpa’s SaaS companies. Just really good salt-of-the-earth companies. And I don’t know — do they build companies like that anymore?

Because you look at something like Twilio — for a long time a darling, but it did not have a sustainable profile. Its ability to turn a real profit is based on this carrier network that’s constantly charging huge fees, so it’s always had low gross margins, and it has always said ah yes, but at scale it will be so much better. Never proved true. That doesn’t mean Twilio is a bad business — but for how it was capitalized relative to the output it can eventually produce, I don’t think that was a great outcome.

I think businesses like Snowflake and Datadog are exceptional businesses. But there are also arguments about could they have been that without a huge amount of capital? Did they need to be as big? Could they have grown more sustainably and, in the long run, been better businesses? I don’t know.

Say what you will about Alibaba — they’ve got their own issues — but I always loved Alibaba’s hundred-year vision. And nobody in Silicon Valley has a hundred-year vision. Nobody talks that way. I really wish we did that more.

Product-led venture, and the everything store

Eric Newcomer: You’re a VC expressing a lot of skepticism. How do you mix that — I get to be skeptical, but I don’t have to turn around and invest. How do you be contrary and then decide where to invest?

Kyle Harrison: Another concept I talk about all the time is product-led venture firms. Everybody rolls their eyes at that — any time a venture fund talks about differentiation, everyone’s like yeah whatever, money is money, you have nice marketing.

But when I talk about product-led VCs, it’s not lip service, and it’s also not an AI bot — I saw a fund in Europe doing something like that, creating a bot that ingests all the information. I don’t think it’s that, and I don’t think it’s hand-wavy stuff. I think it’s clearly articulating a value proposition for what is the job to be done that a founder might have, and why do they hire your money.

One of the biggest problems in the capital-agglomerator world is this race for everyone to become the everything store of venture. Name me a company that Andreessen would say is not a fit for us. Maybe they pass on metrics, or on the founder — but this is a company that does this at this stage, and we don’t really do that? Absolutely not. They’ll raise a fund for that.

Eric Newcomer: It’ll be totally separate, and it’s American Dynamism — if it’s good for America, it’ll be an investment.

Kyle Harrison: And these might be good for the country. But it’s a very broad mandate. I know Katherine and I really like those guys, and I think that is a strategy. But when you try to be everything for everyone, it becomes really difficult to be really good at a specific thing.

The way they’ve tried to solve it — I call them fiefdoms. These large firms have created fiefdoms within their empire: a crypto fiefdom, the American Dynamism team. Sequoia has done the same thing. But a lot of those folks struggle to have a really clear, articulate product, and so what they become is brand value. If I can attach the Sequoia name to my company — Marc Andreessen calls it borrowed credibility — that’s totally legitimate. There’s a world in which just having borrowed credibility is enough: if that VC never called you back, but it helped you close customers because people know the name, that’s a product. That’s fine for that to be your product.

But then you have those firms with big pools of capital and maybe some brand value — and then a long tail. And I think the folks in the biggest trouble are the ones with no discernible value proposition, who have no way to articulate here is why you hire us, and no way to say hey, we’re not the best fit for that.

One of the things I respected about Index is that of the most successful venture firms, Index was one of the most focused. They have a multi-billion-dollar growth fund, but it’s not a $10 billion growth fund. There were a lot of deals they just didn’t chase, a lot of spaces they didn’t get into — they didn’t do that much in crypto. Several of the partners would always say: if you want somebody who’s going to write you a massive check and never call you — awesome, there are firms who will do that. That’s not us. Whereas other firms are like, I’ll take that. If you want me to be super involved, I’ll take that too. Having a very focused mandate becomes more valuable.

I like that world better, when VCs are held accountable for whether they can actually articulate a value proposition. That’s what we try to do at Contrary.

How Contrary actually works

Eric Newcomer: I think of you as doing great writing, and you have these reports on private startups that everybody gets — even startups you don’t invest in. What’s the distinction between that and content marketing? And what do founders get that’s differentiated once they’re in your portfolio?

Kyle Harrison: We think of our content as a SKU within our broader product offering.

The core of Contrary is that we’re very people-centric investors — which is another thing everybody says, but we put our effort and our money where our mouth is. The thesis years ago was: if we can identify the sharpest people in the world as early as possible — sometimes years before they start companies — and build an offering that’s valuable to them throughout their career, we have an unfair advantage to work with them as they build the next generation of great companies.

This started with finding people as early as undergrad and grad school who wanted to be founders someday, or had other ways of demonstrating high slope in their career and real ambition. We’d work with them in school, they’d be scouts with us and help us find companies — which was also a way to introduce them to the world of startups. Then they graduated, and we built a talent function: we helped them find their first jobs, with comp negotiation, thinking through all of it. Then they’d say maybe I want to join an early-stage startup rather than a big company, and we’d help them think about how different private companies are positioned. And then — sometimes we know people for five or six years before they ever start a company — once they start one, great, we want to be your first check, we can introduce co-founders, help with early customers.

Two things came out of that. Contrary Research was born from a recurring question inside what is now a roughly 500-person community. These are sharp people who don’t necessarily need help getting jobs — they want a trusted person in their corner. So they’d say: I’ve got offers from these three private companies, one Series B, one Series A, one Stripe — how do I think about them? I know the people, I like the product, but what’s the investor’s perspective? We’d do it ad hoc, and eventually said: let’s productize this.

The second piece is why I joined in the first place. At Coatue and Index I’d been doing Series A, B and beyond, and when these people start a company I’m not always the best help — I hadn’t done pre-seed and seed in a dedicated way, and that was Contrary’s bread and butter. But what we started to notice is that when you pay attention to a lot of really sharp people, where they go is actually pretty strong signal.

Five of the first fifty employees at Ramp came from Contrary. We saw that and thought, there’s something special here — we should invest in this company. So we invested in Ramp’s Series B. Not a huge check, but an opportunity to put our money behind people we think are really sharp. And even from the Series B, that’s worked out really well. Similarly, six, seven or eight people at Retool; a handful at Anduril; a bunch at Stytch [?]. So we get to not only place great people there and stay close, but also invest in those companies. I joined to build out that Series A-plus practice — before, we were doing it in SPVs.

Eric Newcomer: What’s the fund size and round focus?

Kyle Harrison: Historically around $100 million or so — that’s grown every fund. We’ll lead pre-seed and seed, and participate at Series A and beyond. We’re rarely leading later-stage rounds; it’s more that folks bring us into the syndicate, either because we’ve already helped them hire before we ever invest, or because they know our network and want access to that talent community.

The weight around your neck

Eric Newcomer: How are you playing the downturn in terms of pace?

Kyle Harrison: One benefit is that Contrary historically — even before I got here — never had a lot that got caught up in runaway valuations. So there hasn’t been much scar tissue to deal with, no hangover to come back from.

Our biggest focus is still fairly first principles. We haven’t gotten super heavy into the AI hype or companies raising the same ways. We’ve made investments — we’re pretty excited about some things happening in open-source AI, we have portfolio companies around tooling. But we haven’t gotten crazy where it’s at any cost we need to do these things.

It’s not rocket science, but it is arguably contrarian: we want to find exceptional people, help them build businesses that people actually care about, that solve real problems, in a way that isn’t solving for short-term benefit. You have a generation of investors — myself included — who started investing within this bull market, so there are a lot of folks whose entire education was built around pretty irresponsible ways of building companies.

Part of it for me is that I’d been a founder before I got into investing, so I know how scary it is to run something up in a way that makes you feel like I have no idea what to do with this now. So I’ve always had a hesitation around ripping off all the safety measures. It’s weirdly contrarian to just say we want to build companies that can be around for a long time, rather than companies that can light a bunch of money on fire and — who knows, maybe they’ll be here in a year, maybe they won’t.

Eric Newcomer: But you’re still participating in the venture model — if you invest in a seed company, you’re assessing whether it’ll raise a Series A.

Kyle Harrison: When I criticize venture, a lot of what I criticize is the attitude of excess — hype-driven or FOMO-driven decision-making. I don’t have a fundamental issue with you raise money to do these things and hit these milestones, then raise more because you’ve proven traction. That’s beautiful, I love that. If anything we’ve moved away from that, where companies were able to raise without hitting the milestones.

There are instances where we’ve had conversations with our companies who get offers at a certain valuation from one firm and lower offers from others. And we help them think about it: here’s the weight around your neck at that valuation. Can you raise it? Yes. Is it nice to not take that extra dilution? For sure. But it’s important to be honest with yourself: what does that give you? If you really ramp up and grow, that gives you a year and a half to two years before you need to raise again.

And in two years — what if these six things go wrong? Because in startups it’s usually not six things, it’s twelve. Can you still get to a point where you’re comfortable with the milestones you hit? And at those milestones, can you raise at a valuation that’s a nice markup from where you were? If you can’t, it makes more sense to accept a little more dilution now than to put a weight around your neck.

That’s also true of the partner. We’re a pretty collaborative firm, and because we’re not leading at Series A and beyond, people often treat us as a trusted voice — I’m having conversations with these two or three firms, how do we think about that? We spend a lot of time advising not just on valuation but on the partner. This partner might give you a way higher valuation than that one — but hey, we know so-and-so, we’ve worked with them before, we know your business, and we know what you’re going to deal with over the next 12–18 months, whether that’s go-to-market or commercialization. That person is going to be an incredible partner to you.

Which I think is the last remaining solid part of venture: at almost every firm there are really exceptional individual partners. Maybe the firms have strategies that are good fits for different things, but at every firm I could point to a handful of people and say — if you are doing this thing, in this space, at this stage, that is who you want in your corner.

The information economy in private markets

Eric Newcomer: You’re like me in that you publish reports on how private companies are doing — and at the same time you have private information about which partners are trustworthy, who’s worth taking a lower valuation from. Part of what I’m doing takes advantage of the fact that in Silicon Valley it isn’t illegal to trade on inside information — if anything it’s the job of VCs to have tons of it. So there are these public secrets: if you’re a top SaaS investor you probably know what deals your competitors did over the last couple of months even if they haven’t been published. In some ways I’m saying — VC, you’ve grown up, you can’t hold these secrets among a hundred people and expect everyone else competing not to have that information. How do you think about what you expose and what you hold?

Kyle Harrison: There are a few things happening that get conflated.

On the one hand there’s the trust between a founder and an investor. A founder gets to decide how much building in public they want to do. It’s not on me to decide how much information should be out there — if the founder doesn’t want their valuation known, that’s their decision. Or how many term sheets they had.

Eric Newcomer: But if a founder goes to fifty people — there’s a level of non-public information that, if you’ve shared it so widely that it’s chattered about, you haven’t done the diligence to hold it. You’re basically saying only an elite class gets to know it, and the aspiring class of investors or potential employees doesn’t. Isn’t there a level of care you need to maintain if you expect to maintain secrets?

Kyle Harrison: Founders are in a tricky situation there. Going to fifty people is so many people — but especially in this environment, it’s hard to know who will do what. There have been instances where I was surprised: man, I thought so-and-so was closed for business, but it looks like they did this thing, and they got really aggressive about it — whereas every other thing I sent them they weren’t even willing to look at. VCs are manic depressives in some regard — they don’t know what they want, so how are you supposed to know? You have to cast a broad net.

And in terms of you’ve shared all this with those people — yeah, but that’s because I need those people to do a thing. If I have a customer list with all my customers’ names and how much they’re paying me, I don’t want my competitor to have that, because it’s early stage, we’re still figuring out our product, and that person can go to them and say we’ll give you a $5,000 discount, or we can do this, when they can’t — and we didn’t even know you were working with them. So there is a spectrum where they should still have some privacy while they’re trying to build something very difficult.

What you’re really talking about is more the gossip economy in venture, where you get rewarded for knowing who did what, where, why and when. I struggle with that — not because gossip is unique to venture, there’s gossip everywhere, lawyers gossip about who’s defending whom. The difference is that it reinforces the FOMO.

I have another piece I’m working on — sort of a play on Patrick O’Shaughnessy’s podcast, but Invest Like the Rest — this idea that we are just a bunch of lemmings of capital chasing everybody off the cliff. One of the things I don’t love is that the rumor mill dramatically reinforces that feeling of FOMO.

There’s a partner at Redpoint — Rashaun [?], who does a lot of their content — who had a clip about this. After they closed an investment: so who else did you have term sheets from? And the founder says, oh, it’s just you guys. And he’s like, no, but you were talking to other people — who were you talking to? Well, we sent them the data room but they never responded. And he’s like, right, right… oh man.

The fear is not only that it helps you make the decision — it makes you question the decision after the fact, when you find out nobody else cared about this thing. That’s the stuff I think is pretty unhealthy.

Whereas in public markets you kind of wear your pride and your shame right out in the open, because you’ve got a 13F. You have to say: my portfolio is my thesis, here you go. You look at Cathie Wood — AI is one of five pillars, and she sold most of her Nvidia position four or five months ago. And it’s like, I don’t know, man, that’s the thing you should have been betting on if that’s your vibe. But she has to wear that shame publicly. That’s a different vibe than in venture, and I think that is unhealthy.

Eric Newcomer: Having spent time at Coatue, you’ve seen how public market investors work — a specific paradigm of disclosure, companies have to share things, everybody has the same information, it’s illegal not to.

Kyle Harrison: There’s a lot the private markets could learn from that. I wrote another piece, The Professionalization of Startups, which is about venture but also true of startups: this is not the same as the ’80s and ’90s, everybody-building-in-a-garage era. Building startups has become professionalized — the playbooks are known, the principles are known, the tools are there and everybody has them. So as you lift that, both venture and startup building could learn something from public market disclosure.

Eric Newcomer: What’s your view on how much of a data edge the top private market investors actually have? A hedge fund buys satellite imagery to count cars in a Walmart parking lot ahead of the quarter. How much intelligence do good firms have relative to what’s on the open internet?

Kyle Harrison: You’d be surprised how much information is out there — because to your point, founders go tell fifty investors, and if they have a hundred customers, those customers also know quite a bit. Anytime you interact with anybody there is information leakage. Tegus literally exists based on the ability to go talk to a ton of customers and ask questions. And not all of them are flattering — I’ve had plenty of calls where people say yeah, we’re probably going to churn off this anytime soon, and that’s a relevant insight. There have been times I’ll relay that back to a founder: just so you know, that person is not happy with this — you should know that.

But the magic eight ball data doesn’t really exist. A lot of open-source investors — this is true in AI and in data infrastructure — think about GitHub traction and GitHub stars. You can buy GitHub stars the same way you can buy Twitter followers. You can use it as a relevant data point that things are up and to the right, but if you’re making a decision on that, you’re probably making a bad decision.

For really specific businesses, though — I’ve seen firms buy credit card data and map it to private company information, so you can see certain transactions for consumer companies. That’s not accessible to everyone; in many cases those are six-figure contracts with data providers. Not everybody has that, but the largest AUM firms definitely do.

Eric Newcomer: My intuitions go back and forth, because the internet is such a powerful force — if information is valuable to some people there’s a lot of pressure to put it online, so beating the internet is challenging. On the other hand there are these private conversations. When I publish slides outlining an investor’s returns, I wonder how widely known that already is.

Kyle Harrison: For most people it’s a pretty deliberate exercise. Anybody who has taken money from specific types of institutions — anybody with a Pitchbook login can see their returns, because those institutions have to disclose. So people make really deliberate choices to protect certain types of information. There’s always this back and forth between people trying hard to protect information and people who are really good at finding it. That economy will always exist.

The bigger question is: would it be healthier if more of this was out in the open? If more private funds had to release performance data, more companies had to release financials — and I feel like the answer is yes. I’m in the bucket of thinking that being a publicly traded company is a good way to force you to be healthier. It does make it hard to innovate, because you’re so focused on quarterly things, but you can choose not to care about that to some extent and do okay — and some of it keeps you in check.

But there’s a give and take, because building an early-stage company is really freaking hard and changes on a dime every week. To then turn around and say yeah, but did you file your 10-Q? — dude, I don’t know if I’m going to make payroll next week. I’m not worrying about those forms.

AI, and the capital agglomeration industrial complex

Eric Newcomer: Is AI a transformational shift, or do VCs have a great desire to deploy capital and momentum-invest, and this is the best excuse they have to get back to their old ways?

Kyle Harrison: Multiple things can be true at one time.

AI fervor and the dynamics of building AI companies is a match made in heaven for the capital agglomeration industrial complex.

Eric Newcomer: There’s our headline.

Kyle Harrison: If you’re trying to raise a crap-ton of money and deploy it into things that eat a crap-ton of money, AI is a pretty good place to do it.

Even though — somebody pointed out recently that there’s a step-function decline in the cost of training models, which is true and a really important trend to understand. Building AI is not always going to be as expensive as it is today. Sam Altman has talked about how the next generation isn’t going to be about the largest model; there are other ways to drive higher-quality models than adding more parameters.

But in this moment you have companies that eat a ridiculous amount of capital, and people who have a ridiculous amount of capital and are too afraid of most things but very excited to deploy into this one. I don’t know that that’s a good thing. I think a lot of these AI-hyped investments are going to be another 2021 — the same hangover, again.

Eric Newcomer: For listeners — why are we still talking about Uber at the front of this conversation? In some ways the AI thing is the blitzscaling. There it was subsidizing the network effect; here we’re subsidizing synapse firings in large language models. Same question of whether that large investment pays off, or whether it’d be wiser to work with a smaller amount.

Kyle Harrison: One of the things that’s gotten away in the narrative is how much of this is marketing. At Cerebral Valley, Amjad Masad — CEO of Replit, one of our companies, and I’m a huge fan — made the point that Microsoft would talk about the first contact with AGI, then change it to sparks of AGI. For them it’s marketing. They’re trying to get everybody talking about this thing.

And then there’s a whole other world of people who have made their entire careers Doom-storming around AI, and they love it. So they feed off each other. It’s this nightmare cloud. The companies that want to be hyped say it’s going to be the new coming of God, and the professional critics who want to testify before Congress say it’s not going to be a benevolent god.

What VCs get caught up in — looking at OpenAI as a startup success story. OpenAI is a great company, has done a lot of great things, and anybody questioning that isn’t paying attention. But thinking of OpenAI’s success as just another startup success story, just like Stripe and Airbnb — no. This is a company that built what it built off ten or eleven billion dollars from Microsoft. That’s a different thing, with different dependencies. Not to discredit what they’ve done, but in some ways they are a ridiculously successful subsidiary of Microsoft — in the same way AWS as a standalone business is exceptional, it just lives inside Amazon. This is not a thing that could be easily accomplished outside a massive, entrenched, incentivized corporate hierarchy.

And that’s true of most foundation models: building them is ridiculously expensive and ridiculously general purpose. ChatGPT is good at a lot of things and great at basically nothing — which is fine, because every human being sucks at most things. It’s really good.

Eric Newcomer: I literally copy-edit my posts with it. It tells me to put things in the active voice. It’s amazing.

Kyle Harrison: It is. And that’s one of the reasons people are so afraid to draw a line. My mom reads my writing and tells me I caveat too much — just take a stand. But I have all those caveats because I want to be calling things out and framing things, and then people come back at me and say you don’t get it, AI is game-changing. And I’m like — listen, I absolutely get it. I fundamentally believe the world as we know it is going to be transformed by this technology. I am an optimism maximalist through and through. I’m so excited about the opportunity and the implications.

But there are pockets within this world, and I think massive concentration of power and influence with a technology like AI is really dangerous — and that is where we’re heading. It’s really scary to see OpenAI talking so much about regulation, because they’re going to be in the room where it happens, helping write the regulation that protects them.

And I hesitate to criticize Sam Altman, because as far as I know he has no economic interest in OpenAI per se — which, if that’s true, is remarkable.

Eric Newcomer: OpenAI is the most open — they really need to take it out of the name. Beyond not being open source, they’re not a very open company.

Kyle Harrison: So for Contrary Research, probably next month, we’re going to put out a deep dive that digs into the openness of AI. And in one section we talk about this — several folks at OpenAI have been saying the quiet part out loud. Every time OpenAI talks about the shift they took — from we’re going to share our research with the world to just kidding, this is not a good idea — the first thing they’ll talk about is safety, and the second thing they’ll talk about is competition. Almost every single time, without fail. It’s basically the mea culpa. And it’s true that it’s really difficult to build GPT-4. But just say the quiet part out loud: you guys are not this benevolent, open-for-the-benefit-of-the-world thing.

What he’s actually looking for

Kyle Harrison: Rather than looking for one thing that spikes, you look for a functional funnel — even if it’s very early.

At the top you have some kind of user or open-source engagement — people want to use this thing, they’re excited about it. GitHub stars, community size, anything that looks like top-of-funnel engagement. That’s valuable, and many people have made investments purely off that. I heard more than once in 2021 people talk about valuations in terms of a multiple of Discord users — and that is a deeply unhealthy thing to do. People are turning over in their graves.

Then you translate that into actual contribution — are there people meaningfully using and taking advantage of the tool they’ve built?

And then: does that translate into an actual enterprise-grade product? Is there something more than facilitating? There are a lot of really great companies today that are still largely facilitating. Hugging Face is a really great company — from a business model perspective, most of what they’ve done is engagement and enablement. But turning that into a foundational enterprise-grade product they can sell with multi-figure ACVs — that doesn’t quite exist yet, and there’s not a really clear path to it.

Eric Newcomer: People say GitHub figured it out, so they’ll figure it out.

Kyle Harrison: For what it’s worth, I think Hugging Face is a phenomenal business. My fear is that there’s a difference — because even GitHub took a long time to become what it is, and a lot of that came from Microsoft having a really established DNA for commercialization and distribution. I don’t want to say GitHub didn’t do anything — they built a really important place for people to be — but the long term was fuzzy for a really long time, and Microsoft kind of made the long term make sense.

For a lot of these companies it’s just a question of: I don’t know. The long term could be phenomenal, maybe they crush it. But right now it’s not clear. My preference is for companies that can at least articulate what that long-term potential is — how do you take the engagement and the contribution and translate it into something commercial?

One of our portfolio companies is Nomic, still very early. They built a model called GPT4All — effectively an open-source GPT model that uses a fraction of the compute and can run locally, so for healthcare and government use cases they’re way more likely to use that than something like OpenAI. It took off; it now has about as much open-source traction as something like LangChain.

But where that translates down the funnel is that the company originally started building a data visualization tool for fine-tuning models. We have this really sick portrait up in the Contrary New York space — it’s Stable Diffusion visualized in dots. You can use it to understand how different parameters play with each other, and then train the model based on we think these sections aren’t relevant, or dig in and fine-tune. It’s awesome. I keep telling them they should sell it, because it’s so cool. They have things like Kermit the Frog corner, where you can see all the images in Stable Diffusion that get leveraged to create Kermit the Frog output.

So they’ve done a really good job of taking open-source traction, translating it into contribution — exceptional people engaging with their tool — and then that translates down into: when you want to train a unique model in a way that’s scalable but doesn’t require massive compute, Atlas, their data visualization tool, is how they built GPT4All. By being a customer of Atlas you can engage with models, even open-source models, in a much more meaningful way.

I love that funnel. There are few businesses that have it, but that is what I’ve been looking for in AI — as opposed to boy, is this hot, who knows where it goes, but it’s hot. That stuff makes me so anxious, especially having paid attention in 2021. A lot of that did not end well. And in a small subset of venture, this feels almost more hype, more fervor, more money — obviously complicated by the hyperscalers.

Eric Newcomer: Thank you so much for coming on. When I launched Newcomer I articulated a vision of contrarian optimism — I think tech journalism had gotten so negative — so I certainly feel a kindred spirit in trying to be optimistic while being somewhat contrarian.

Kyle Harrison: This is super fun. I feel like we see things eye to eye for sure. Also, I’ve been told multiple times that we look very similar — I think it’s the beard and the glasses.

Eric Newcomer: The headphones aren’t helping. They’re not a normal accessory.

Kyle Harrison: I think they’re the same exact headphones.

Connections

The show

Eric Newcomer · Newcomer — his stated editorial posture is contrarian optimism, which is why the conversation lands where it does; both of them are publishing information about private companies while sitting inside the ecosystem.

Essays this episode is built on — one of them unpublished at the time

VC Contagion — the piece the episode is named for. Kyle refers to it as “the piece I shared with you”, and it published three days after the episode aired. See the provenance note.
The Unbundling of Venture Capital — his first piece that travelled, and the source of the Don Valentine disruption → confusion → opportunity line.
The Blackstone of Innovation — capital agglomerators, AUM expansion, and public markets pricing PE firms on fees rather than carry.
The Professionalization of Startups — invoked on why private markets could learn from public-market disclosure.
The Openness of AI — pre-announced here (“probably next month”), published six weeks later, with its central finding given away on air.

The capital argument

“Your fund size is your strategy” — you cannot earn 5–10x on multiple billions, and the LPs deploying at that scale are content with 2–3x.
Power Law — and why it stops operating above a certain fund size.
a16z / Coinbase — returning an entire fund’s AUM on one position, and what that says about how these outcomes get narrated.
NEA · Lightspeed · General Catalyst · Thrive Capital · USV — the scale contrast Eric lays out.
Bill Gurley — $30M pre-IPO was once remarkable, now it’s a Series A; voraciously hungry unicorns; the sine wave of prolonged unsustainability.
Paul Graham — the since-2013 gospel that if more money won’t help, don’t raise it.
Blitzscaling — Eric’s reframing: does it work outside ZIRP, and can you come back from it?

Companies as evidence

Uber · Lyft — Hungry Hungry Hippos, the pink-mustache commoditization, ~$30B of post-IPO losses, and “was that the best capital allocation output equation we’ve ever seen? Probably no.” Dara Khosrowshahi’s unit-economics line prompting “what is the definition of big if not Uber?”
Coinbase · WeWork · Snowflake — Midas-List winners that don’t sustain.
Adobe · Workday · Salesforce — “your grandpa’s SaaS companies… salt-of-the-earth.”
Twilio — the darling whose margins were structurally capped, defended by at scale it’ll be better, which never arrived.
Alibaba — the hundred-year vision, and “nobody in Silicon Valley has a hundred-year vision.”

Firms and differentiation

Product-led venture — “what is the job to be done that a founder might have, and why do they hire your money?”
a16z as the everything store, fiefdoms inside the empire, and American Dynamism / Katherine Boyle.
Marc Andreessen — borrowed credibility conceded as a genuine product.
Index Ventures — named as one of the most focused top firms: “if you want somebody who’ll write you a massive check and never call you — that’s not us.”
Contrary — the ~500-person community, the talent function, Contrary Research born from what’s the investor’s perspective on these three offers?, and the signal that five of the first fifty at Ramp came from the network (plus Retool, Anduril).
The valuation advice: the higher price is “a weight around your neck”; run it forward and ask what happens when six things go wrong, “because in startups it’s usually not six things, it’s twelve.”

The information economy

The gossip economy and how it manufactures FOMO — the Redpoint clip about asking a founder post-close who else bid.
Cathie Wood / Nvidia — the 13F as the public-market contrast: “she has to wear that shame publicly.”
Tegus — information leakage as the premise of an entire business.
GitHub stars can be bought like Twitter followers — engagement metrics as a decision input are a trap.
Purchased credit-card panels mapped to private companies, at six-figure contracts only the largest firms carry; Pitchbook exposing returns wherever LPs must disclose.

“A match made in heaven for the capital agglomeration industrial complex.”
OpenAI as “a ridiculously successful subsidiary of Microsoft”, on the AWS analogy; ChatGPT “good at a lot of things, great at basically nothing.”
Sam Altman on the next generation not being about the largest model; and on regulation, “they’re going to be in the room where it happens.”
Amjad Masad at Cerebral Valley on first contact with AGI → sparks of AGI as marketing, and the professional doomers feeding off it — “this nightmare cloud.”
The funnel: open-source engagement → contribution → enterprise-grade product. Discord-user multiples in 2021 as the cautionary case.
Hugging Face — phenomenal at the first two rungs; GitHub took years and needed Microsoft’s commercialization DNA for its long term to make sense.
Nomic · GPT4All · Atlas — the full-funnel example, including the Stable Diffusion portrait and Kermit the Frog corner. He tells the same story on The Centralization of Power in AI five months later.