Introduction: this wacky world we live in
Venture capital is deeply weird these days. Consider these phenomena:
An extreme bifurcation of companies into those that are hot, and those that are not. Startups that are not hot, but merely “excellent businesses”, seemingly cannot get funded.
A parallel bifurcation of VC firms into multi-stage behemoths on the one hand, and niche strategies (solo cap, seed, frontier, spv-coinvest) on the other. Mid-size, mid-stage firms are a dying breed.
A schizophrenic attitude to traction: either ignored completely, or “no number is high enough”, happening in parallel with some of the fastest revenue ramps in history.
The embrace of king-making and capital-as-a-moat; a pervasive sense of pre-anointed winners and losers.
Companies staying private for longer than ever.
Uniform cap tables, as venture firms co-invest in mega rounds alongside their rivals.
Investors ignoring conflicts of interest with existing portfolio companies.
Massive value placed on pro rata, but pro rata agreements mean nothing; rights are routinely re-traded.
Increasing isolation and solipsism: inside jokes and social codes, shared secrets, industry-owned media franchises, closed networks, disregard for norms and normies.
Horrendously self-destructive PR from the market leaders, especially the frontier labs. They’re going to destroy jobs, disrupt industries, and maybe doom humanity.
None of this makes any sense! Or does it?
In this essay I argue that all of these phenomena are rational responses to the venture capital market becoming more efficient, in a way that is notably distinct from public markets efficiency. I call this the “efficient venture markets hypothesis”. Read on!
Efficiency in public markets
Public markets are mostly “efficient”. This statement is usually interpreted in one of two ways:
The “price-is-right” hypothesis: market prices correctly reflect all known information about the underlying asset.
The “no-persistent-alpha” hypothesis: it’s impossible to consistently beat the market over the long term unless you’re providing a service of some sort (liquidity, insurance, information) or you have a structural edge (monopoly, regulatory capture).
The modern synthesis is that both interpretations are kind of true. Prices are kept close to “fair value” by market participants who buy what’s cheap or sell what’s dear. They are compensated for their information, skill, capital or risk appetite, by precisely the amount required to dissipate the inefficiency. But this compensation doesn’t last; copycats emerge for any successful strategy, making consistent outperformance impossible.
(All of this is well attested in academic research.)
The price is not right ...
At first glance, venture capital does not seem a likely candidate for efficiency. The mechanisms of price discovery in public markets simply don’t work for startups.
Companies are unique and opaque. Terminal value estimates range from zero to the trillions, sometimes for the same company. Financing events are discontinuous and situation-specific. Secondary markets are often illicit and always thin. There’s no good mechanism to short a company or sector; number go up, always. There are no regulators, information standards, or central clearinghouses.
Put these together, and there’s simply no way for market participants to push prices toward their economic expectation. The price is self-evidently not right, and this fact is widely recognized in the industry (“valuation is not value”).1
... but venture is still efficient
In public markets, efficiency manifests as faster convergence to the right price.
Venture is different. In venture, price doesn’t matter. Instead, in venture, efficiency manifests as faster convergence to consensus.
To be clear, venture has always operated on consensus. The genius of modern venture capital is the way it creates alignment among different investors who finance different stages of the startup journey, from inception to IPO. This converts an impossible extrapolation problem into a tractable bridging problem, and thereby unlocks a whole new asset class.2
But this process takes time. Time to hit milestones; time to derisk uncertainty; time to establish consensus on the winners and losers. Efficiency, in the venture sense, means reducing this time. Firms that can compress these horizons beat those who cannot.3
Consensus is endogenous
But here’s the secret. Consensus is endogenous: it exists because we decide it exists. If we can manufacture it, why not skip all the intermediate steps? Why fund-raise in stages, why go through the laborious process of PMF discovery and company-building and market creation? Why run the whole race when you can jump straight to the finish line?
The endgame looks like venture today: the most consensus market in years, if not forever. LPs all herd into the same few mega VC firms; these firms herd into the same few startups; and these startups herd into the same few business models, fast. Consensus is established rapidly, and broken rarely. It’s efficiency, Jim — but not as we know it.
Alpha? We don’t need no stinkin’ alpha!
There’s a price to pay for this efficiency, and it’s steep.
Consensus decisions lead to mediocre outcomes.
If every VC chases the same investments, by definition they will all have the same returns.
This is deeply weird. What could possibly possess professional investors, that they no longer care about outperforming their competition? Why is consensus so attractive?
The answer, as always, is AI.
AI changes the math
For most of the last two decades, tech returns have been dominated by software. And software is a great, great business model: asset light, recurring revenue, high gross margins, customer upsells, negative working capital, strong FCF.
But what’s great for the business is not so great for financiers. SaaS businesses simply don’t need that much capital, and that means there’s an upper limit to the size of the venture industry that funds them. If you’re a VC competing for this limited pool of investment dollars, relative performance matters a lot.
AI ... is not like that. AI absolutely incinerates capital. There’s huge capex for chips, data centres, and power generation; high COGS because of inference compute; fast-depreciating R&D in the form of training runs; generally lower gross margins; expensive data and talent; and no apparent upper limit to any of these, because of the bitter lesson.
This requires investment. It requires a LOT of investment.
The inbound trade
AI’s insatiable appetite for capital is fantastic news. I say this unironically: it’s manna from heaven for the venture industry. Because this money has to come from somewhere. More specifically, it has to come from outside the traditional sources of venture dollars — the measly 5% of every diversified portfolio, earmarked for venture by traditional allocators — it has to come from net-new investors. And VCs stand ready to intermediate it.
If you’re a scaled VC, nothing is more important than inserting yourself into this flow.
And consensus is the tool you use.
Consensus is how you make the tech industry as a whole “legible to capital”. Consensus removes friction. Consensus makes it easy for outside money to come in. If every GP has essentially the same portfolio and charges the same fees for the same performance, then no LP has to make hard choices or take career risk. Align investors; mobilize capital; select winners; deploy funds; don’t waste time; make it easy.
Consensus lubricates the flow of non-tech capital into the tech industry.4
AI beta is global alpha
Why would any rational LP agree to this? Why wouldn’t they seek outperformance, instead of piling into median portfolios?
It has to do with the global opportunity set. Over the last few years, the AI trade has been so lucrative, relative under-performance or out-performance within that trade almost doesn’t matter. Median portfolios within the AI bucket utterly dominate even the best portfolios outside of it.
OpenAI and Anthropic lead the way, of course, but firms like Databricks, Cursor, xAI + SPCX, Scale, Perplexity, Cognition, DeepSeek, ElevenLabs, Sierra, Modal and dozens more have created immense amounts of customer and shareholder value.
And so for a global investor, what matters is getting exposure to the private AI space. In fact, getting a consensus portfolio within that space is probably a feature, not a bug: you really want to index the market, not make risky prop bets within it. AI beta is global alpha.
Venture weirdness, explained
Let’s recap our theory here. I claim that:
AI requires unprecedented amounts of capital
This capital must come from outside the tech industry
VCs want to intermediate this inbound flow of net-new capital
Consensus is their mechanism of choice
Consensus trades off performance in service of efficiency
Investors want this because they get to index the tech sector
The industry as a whole has oriented itself around consensus
Does this theory have explanatory power? Well, let’s take a look at the assorted weirdnesses I listed at the start of this essay:
An extreme bifurcation of startups
This one is pretty clear. Hot = part of the consensus. If you’re not part of the consensus, you’re not part of the flow, and you won’t get funded, by net-new investors or by the insiders who anticipate them.
Note btw that “hot-or-not” is a self-fulfilling prophecy in a consensus world. Hot startups have a knack for self-promotion on top of (usually) a decent business; both of those are flywheels in a market that rewards heat.
A similar bifurcation of VC firms
If the goal of venture today is to construct consensus portfolios for net-new investors, there are only two roles that matter: hunters and gatherers. Hunters seek to discover consensus targets, while gatherers seek to deploy consensus dollars. In the past, we’d have called them “risk capital” and “growth capital”, respectively. And there’s nothing in between - you can’t be “kind of consensus”, the very concept is incoherent. Goodbye middle class.
A schizophrenic attitude to traction
Consensus is built on many things; traction is just one of them. Traction by itself is neither necessary nor sufficient to establish consensus. Indeed, sometimes the consensus comes first, and the metrics follow. So if you’re trying to build a causal model with traction as the independent variable, you’re probably doomed to failure.
The embrace of king-making and capital-as-a-moat
If consensus is endogenous, and we’re going to manufacture it anyway, why not just buy it? Spend enough and you might scare away rivals; you’ll certainly scare away their potential funders. A dangerous game, but the big funds these days are willing and able to play it.
Companies staying private for longer than ever
This is an interesting one. If consensus is the goal, then the biggest danger to a scaled company is falling out of consensus. And going public risks precisely that outcome. Now there’s regulatory scrutiny, and short sellers, and reporting requirements, and analyst buy/sell ratings. Who needs that noise?
Far better to control your cap table, and fill it with folks who are motivated to maintain your consensus position, while using secondaries to generate liquidity for those who want out. (And if you play the consensus game right, there will be an endless stream of buyers for said secondaries.)
Uniform cap tables, as venture firms co-invest
There are fewer consensus companies than there are VC firms, so co-invest is rife. This also speaks to personal incentives: the biggest career risk today is not “losing money”, it’s “missing consensus names” in your portfolio. Buying hot-co at Series J is the move.
(This also explains vibe investing and the (seemingly shocking) lack of diligence that goes into many deals.)
Investors ignoring conflicts of interest
What if you invest in a name, and it turns out to not be the consensus name in its space? In an earlier era of VC, you’d shrug and write it off - but you wouldn’t invest in a competitor, the reputational damage would be too great. Today, it’s like, whatever. Everybody — including, critically, the founders! — knows that mega firms HAVE to be in the big names; it’s nothing personal. Once founder stigma vanishes, all bets are off; all behaviour is permissible.
Massive value, but pro rata agreements mean nothing
Venture has always had an honour culture, not a legal one. This is partly because of its size: tech has historically been small, tightly interlinked, and a “repeated game” (California’s non-compete law helps here). And partly because of the nature of consensus, which is by definition implicit and endogenous. The honour culture is why investor vs startup lawsuits are so rare, but it’s also why agreements carry very little enforcement weight.
There’s another effect at play here. Late-stage investors have more power today than ever before. And it’s not just due to the sheer size of their funds; dollars are fungible! More important is that late-stage cheques bestow the seal of consensus, which is worth more than any dollar amount. This is why rights get re-traded, and founders and earlies go along with it.
(Dual-priced rounds are another illustration of the same phenomenon.)
Increasing isolation and solipsism
If you’re intermediating net-new capital entering tech for the first time, you have to justify your gatekeeper status, your privileged access. This means highlighting the “otherness” of the industry. The best founders are brilliant and unique and don’t play by the rules and hang out in weird communities and think differently and don’t even deign to talk to squares and normies and the insufficiently AI-pilled: does this sound familiar? Of course you need Big VC to gain access to them.
And this replicates: founders play obscure and hard-to-get with VCs, cracked engineers play obscure and hard-to-get with founders, and fractally down the line. The SF bubble is an intentional construct!
Horrendously self-destructive PR from the labs
This is another interesting one. We’re seeing net-new inflows into private tech. This pays off if — and only if — tech’s TAM expands dramatically. So the labs have to claim that they’re going to take over the world; disrupt every industry other than tech; replace every worker; transform every sector. It’s bad PR, but the alternative is worse. If they admit that tech cannot, in fact, grow to be a significantly larger share of the economy, that implies lower capital reallocation from other parts of the economy to tech. And that kills the AI funding trade.
I think we’re doing pretty well on the explanatory front so far!
Further implications
So great, we have a theory of the market. How does that change our behaviour? Some obvious implications:
For investors: pick your lane. The factory-farming, assembly-line version of venture is dead (for now), so decide: do you want to be a hunter, or a gatherer. When even USV and Benchmark can’t hold the line, ask yourself: can you?
For founders: play the game. They say that with AI codegen, product is easy and distribution is everything5. Well, the same goes for your company. Your startup is a product for investors; even a great product (company) still needs distribution. You can’t just slide into consensus, you must actively foment it.
For LPs: understand the game being played. You’re paying 2+20 for an index product; that’s great as long as it outperforms your baseline, but not so fun if the market turns.
For angels: are you in the flow? If not, beware. Adverse selection has never been more powerful.
For operators: understand the trade-offs. You can join a consensus company; the money is great, the momentum strong, the pressure intense, the learning steep. Or you can stay off the beaten track: it’s not necessarily “safer”, but is probably less stressful (also less exciting).
None of this is new; but it’s worth seeing the world clearly.
And wrapping up
This brings us to the end of another Pivotal essay. I hope you enjoyed it! Thank you, as always, for reading.
I have so much more I want to say on this topic, but I’m trying (not entirely successfully) to keep my word count low and my cadence high. 2,000 words every few weeks beats 10,000 words once a year: on this front, I too am moving towards consensus and efficiency.
Toronto, October 2026
Further reading
This essay is the third of a trilogy on modern venture capital. The previous two were 2022’s Minsky Moments in Venture Capital, and 2025’s Making Markets in Time. Both proved very popular with Pivotal readers; if you liked today’s essay, you might like those as well.
If you’re a finance professional – specifically, if you’re a startup finance lead or fractional CFO, or if you work at a family office or VC finance team - you might want to read about what I’m doing outside of Pivotal, and join the waitlist!
Never miss a post!
Pivotal is a free newsletter of long-form essays on finance, AI, data, startups and tech. It’s written by Abraham Thomas, the founder of Windkey. I write infrequently and in depth; I like to think my pieces are worth writing, and worth reading.
Subscribe to Pivotal and share it with your friends and colleagues: especially those who work in finance. Encourage what you want to see more of in the world.
What about the no-persistent-alpha argument? This one is less clear. VC is one of the few asset classes where outperformance persists; the same firms stay at the top of the ladder for decades. This is due to a well-known feedback loop: the top investors attract the best founders; the best founders drive portfolio performance; and portfolio performance defines the top investors.
I wrote about this at length in Making Markets in Time.
This idea, that venture capital risk-reward is fundamentally time-based and not price-based, was the inspiration for Minsky Moments in Venture Capital.
Just to give a sense of scale here: Anthropic’s last raise was $65B in May 2026. That’s as much as the entire “traditional” US VC industry raised in 2025. And the Anthropic round was led by venture/growth/crossover firms: Sequoia, Greenoaks, Dragoneer, Altimeter, D1, DST, Coatue, Lightspeed, General Catalyst, Insight, ICONIQ. All the usual suspects, in fact, rounded up in an up round.
I once described the business model of the “venture majors” — the mega investors that dominate VC today — as not investing, but market-making. This is still true, and now they’ve found the perfect flow to intermediate.
“They” are not completely right, but they’re not completely wrong either.


Separately, this article is dead-on. I left VC for the reasons you described. 10 years ago you could do well if you were a good picker of oddball companies with potential, and we did. Now you have to operate entirely within the consensus. My strategy of building a network of offbeat characters with peculiar deal access in different markets around the U.S., and intentionally steering away from the handful of kingmakers in the Bay Area, became maladaptive in the new landscape.
In many ways I feel blessed that I got out when it stopped being fun. Being stuck playing a game I sucked at for the rest of my career would have been a nightmare.
> “SaaS businesses simply don’t need that much capital, and that means there’s an upper limit to the size of the venture industry that funds them.”
Yes compared to AI. But I’d argue that SaaS absolutely incinerates capital compared to the prior era of on-prem license software. I’ve long suspected that the move to subscription models was aided along by investors who wanted to own more of these companies on the way up, so they pushed a business model where financing CAC against a very long time horizon LTV made rational sense for management.