Sep 4, 2025 · 55m · a16z
Is Non-Consensus Investing Overrated?
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The a16z Podcast, hosts Martin Casado and Erik Torenberg join guest Leo Polovets to debate whether non-consensus early-stage investing is overrated, exploring how market efficiency, follow-on capital dependency, fund mechanics, and sector hype cycles impact startup valuations and long-term venture returns.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 18.1% of the talking time here. How this is scored →
The host as informed peer 4.5 Guest teaching 4.6 Guest disagreement 3.5 The host pushing back 2.9
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 33:08 Martin rejects Scale AI non-consensus label
Martin directly interrupts and challenges Erik when Scale AI is presented as a non-consensus seed deal, pointing out that top-tier investors made it intensely competitive.
Hardest push from the host ▶ 43:06 Erik reframes VC identity vs market realityErik directly challenges the guests' reliance on consensus terminology, asserting that VC ego and identity drive the desire to appear non-consensus rather than economic logic.
Biggest teaching moment ▶ 33:08 Martin schools Erik on Scale AI's investor profileMartin corrects Erik's characterization of Scale AI by citing the involvement of elite investors like Dan Levine to prove the round was highly consensus.
The host holds their own ▶ 33:39 Erik demonstrates expertise on outcome expansionErik counters Martin's challenge by demonstrating how expanding outcome scale ($100B+ companies) mathematically justifies paying historically high seed valuations.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| The a16z Podcast Title Sequence | 3 | 2 | 1 | 1 | Erik prompts Martin to recap his viral tweet regarding non-consensus investing. Martin clarifies his view that ignoring market consensus is dangerous due to startups' dependence on follow-on capital. | |
| Early-Stage Non-Consensus and Return Multiples | 4 | 5 | 4 | 3 | Martin pushes back against Keith Rabois' list of non-consensus winners, arguing that hard fundraising rounds should not be conflated with non-consensus market views. The group discusses Anduril's valuation history as an example. | |
| Valuation Trajectories and Market Efficiency Analysis | 5 | 4 | 3 | 3 | Erik brings up Peter Thiel's thesis on up-rounds, prompting Martin and Leo to explore valuation trajectories. Martin proposes testing whether prior hot rounds predict future hot rounds as evidence of market efficiency. | |
| Productive Asset Value vs. Market Hype Cycles | 4 | 5 | 4 | 3 | Leo contends that non-consensus companies build better financial discipline, whereas consensus deals often skip diligence. Martin agrees, noting that most startups die from indigestion rather than starvation. | |
| Market Efficiency Trends and Fund Vintage Cycles | 4 | 5 | 3 | 3 | Erik questions if venture markets are becoming more efficient overall. Leo and Martin debate whether inflated consensus valuations indicate market inefficiency or efficient pricing of high-returning assets. | |
| Martin's Founder Journey and Deep Tech Milestones | 3 | 5 | 2 | 1 | Martin shares Nicira's funding journey through the 2008 crash to illustrate shifting consensus. Leo explains how seed investments transition from non-consensus to consensus as technical milestones are met. | |
| Sector Dynamics: AI, Deep Tech, and Unit Economics | 4 | 5 | 6 | 2 | Martin sharply criticizes VCs investing in hyped deep tech like humanoids without proven unit economics. He mockingly recounts partner meetings that ignore physical laws in favor of infinite TAM calculations. | |
| Fund Size Mechanics and Expanding Return Scale | 7 | 8 | 7 | 5 | When Erik offers Scale AI as an example of a non-consensus seed investment, Martin aggressively cuts in to reject the framing. Erik defends his point by explaining how expanding outcome scale justifies higher entry valuations. | |
| Decacorns, Sizing, and Power-Law Portfolio Models | 5 | 5 | 3 | 3 | The panel discusses power-law distribution models and whether decacorn outcomes are frequent enough to dictate fund sizes. Martin names several a16z mega-winners to show that massive scale outcomes are becoming more common. | |
| VC Identity vs. Market Reality and Product Disruption | 6 | 4 | 4 | 5 | Erik challenges the guests' terminology, arguing that VCs' obsession with non-consensus is tied to identity rather than reality. Martin expresses strong belief in creative destruction and growth investing over public market predictability. | |
| Ecosystem Incentives and the Barbell Fund Landscape | 5 | 3 | 2 | 3 | Erik outlines how individual VC interests diverge from ecosystem incentives, advocating for a barbell fund structure. Martin agrees that intense competition ultimately drives technological progress. |
Statements from this episode (37)
Casado: Non-consensus investing is dangerous because solitary views are often wrong
“It's dangerous to do non-consensus investing. Like, that's a dangerous idea. If you're alone in your view, you may just be missing something.”
Polovets: Startups dependent on capital markets must eventually reach consensus to survive
“Eventually, you have to get to consensus. If you're dependent on capital markets, it's very hard to keep the company alive if nobody wants to fund it.”
Casado: Ignoring VC consensus is dangerous due to follow-on capital needs
“Being blinkered to how VCs view companies is, is actually quite dangerous because you're so dependent on follow on capital.”
Casado: Early venture markets are far more efficient than people realize
“My underlying belief is early markets are actually pretty darn efficient, a lot more efficient than people realize.”
Polovets: Many of my top returns came from non-consensus investments
“I would say like for me, and maybe we invest like a tick earlier more like towards pre-seed and seed, but for me, a lot of my best investments have been more on the non-consensus side.”
Polovets: Return multiples fall significantly after startups establish proof points
“Once they get good, the valuation skyrocket so fast that like you could still get good multiples, but they're just much lower than it, you know, early stages.”
Casado: Venture markets are efficient and good companies command high prices
“Markets are actually quite efficient. If the market's efficient and it's a good company, the price is going to be high.”
Casado: VCs should seek great companies rather than cheap valuations
“You shouldn't be looking for good deals with respect to other investors. You should be looking for good companies and price shouldn't sway you from that.”
Polovets: Founders with prior unicorn exits are never non-consensus
“I'm not sure an ex-unicorn founder would ever be, not consensus really.”
Casado: Previous round heat best predicts future high up-rounds
“So I'll bet the best prediction of a The best correlate of a high up round outside of the business is the fact that the previous round was hot.”
Polovets: Most successful startups originate from non-hot batches
“Most of that, like most of the hot companies end up coming from the not hot batch, right?”
Polovets: E-commerce valuations swung wildly year-to-year despite stable business fundamentals
“E-commerce didn't, I think the fundamentals didn't change that much year to year, but like the valuations and the like appetite for investing and maybe starting companies changed a lot year to year.”
Casado: Many top a16z deals had no competing investors
“Even my own portfolio, many of the top deals I've done, nobody else was in the deal, you know, et cetera.”
Casado: Most startups fail from indigestion rather than starvation
“Most companies fail from indigestion, not starvation, which is they just raise too much money too easily. They don't listen to the actual market, which is, you know, the customer base. And as a result, they just have a bunch of bad practices and end up running…”
Casado: 2021 billion-dollar startup cohort will be a massive capital wipeout
“I actually think in 20, 21, If you look at, if you just did a study of that cohort, the companies that had these, you know, these billion dollar bees, if you remember that time, it was totally crazy. I'll bet that's probably one of the biggest wipeouts of capi…”
Polovets: VC expansion makes market efficient for non-consensus startups
“For non-consensus companies, it's getting more efficient because the more investors there are, the more likely you are to find at least one or two that like what you're doing.”
Casado: Traditional infrastructure startups currently struggle to raise capital
“In my area of traditional infra, or of infra, a lot of the traditional companies that, you know, two years ago would be great, like, can't even raise right now, just because they're not in the sweet spot.”
Casado: OpenAI, Anthropic, and Cursor show tremendous real growth
“The reality is, is open AI has grown tremendously, and Anthropic has grown tremendously, and Cursor has grown tremendously, and so, like, there is some underlying market signals To fuel the chaos.”
Casado: Nicira sale returned a fund with historic enterprise revenue multiples
“When we actually sold the company, I mean, it, you know, it returned a fund. You know, it was one of the highest acquisitions on multiples of revenue at the time in enterprise software.”
Polovets: Six to eight of his top ten investments struggled raising seed
“I think at least on my side, for a lot of the precedes and seeds I've done, I went back, I think over my top like 10 investments, maybe six or seven or eight took them months to raise a seed round.”
Polovets: Deep tech startups rarely have functional products by Series A
“At seed, it's very rare to see like, oh, there's an asset that's going to be working here at the series A. Cause usually the asset's still going to be like being developed at the series A or maybe series B.”
Polovets: Humba Ventures paused defense investing due to inflated valuations
“We invested in defense a lot three, four years ago, and then we kept looking, we're basically paused for a year and a half or two, because after the Ukraine and Israel thing, you know, prices just went up like two, three, four times, but the company fundamenta…”
Casado: ElevenLabs and Midjourney have excellent unit economics
“For AI, for better or for worse, like, you have great unit economics. I mean, you know, everybody knows kind of like when we always talk about the opening eyes and the anthropics, but do you talk like the 11 labs, for example, or mid journey? I mean, these are…”
Casado: AV unit economics remain on par with Uber despite $100B invested
“The story for autonomous vehicles is, is that even after the industry's put a hundred billion dollars in it, a hundred billion, the unit economics are still, you know, let's call it on par with Uber.”
Casado: Nicira was acquired for $1.2B with under $10M in ARR
“My company was acquired for 1.2 billion dollars. We had, let's call it, you know, less than ten million in ARR, right?”
Casado: Nicira reached a $600M run rate 3.5 years post-acquisition
“The run rate of the three and a half years later, like the run rate was, you know, six hundred million dollars within VMware who acquired the company.”
Casado: Strong venture returns suggest startup entry prices are too low
“The fact that we get the returns we do suggests the prices are too low.”
Polovets: Ownership and entry valuation don't matter for top-performing startups
“Like if you're in like the best company of the year, I don't think like, I don't think ownership matters that much. I don't think like the price matters that much.”
Casado: a16z portfolio includes four $100 billion companies
“Even in the Andreessen portfolio, I was just thinking off the top of my head, we have three companies that are at the 104 companies at the hundred billion dollar mark, right? I mean, there's Stripe, Databricks, Coinbase, OpenAI.”
Casado: The number of decacorns has increased tenfold over ten years
“The amount of, like, Decacorns is order, probably an order of magnitude more than what it was 10 years ago.”
Polovets: In a consensus VC market, cost of capital dictates winners
“In a purely consensus world, like it all just comes down to the cost of capital. Right. And so if my LPs want five X and yours want two X, you could pay two and a half times higher prices and the company's not better. It's just like, oh, like your cost of capi…”
Casado: Public markets favor predictability over innovation, forcing incumbent defensiveness
“If you're in a large public company like I was, you realize that the public markets really care about predictability over innovation for sure. I mean, and so innovation is stifled so much. And in fact, it kind of causes large companies to protect themselves th…”
Casado: Best startups are non-consensus to customers, even if consensus to VCs
“I really believe the best companies themselves are non-consensus to customers. I just think that the investing market is, is, is, is different than that. Like they kind of understand that and therefore a comment on investors being consensus is very different t…”
Torenberg: VC ecosystem will be dominated by massive funds and boutique firms
“I also do still very much believe in the barbell that there will be, you know, these big, you know, sort of massive funds that continue to win and invest in compound value. And also these you know Smaller, focused, concentrated, expert, the boutiques who absol…”
Polovets: Most of Humba's 10-12 unicorns lacked multi-stage seed leads
“I think we've invested in like 10 or 12 unicorns roughly. Maybe like a third of those or quarter of those had a series A investor at seed.”
Polovets: Multi-stage VCs win repeat founders by pricing seed higher
“Where if it is a founder that previously built a business that exited for a hundred million and they're like in the space that they know super well, that's gonna get done at like 40 instead of 20 or 80 instead of 20 post. And chances are it's gonna be a multis…”