Venture capital has become a very, very big asset class. Just a couple of weeks ago, Andreessen Horowitz announced a $15 billion in newly committed capital. This is a single fund representing a half dozen products that in aggregate make up 18% of all venture dollars allocated in the U.S. this year. The firm as a whole now manages $90 billion, neck and neck with Sequoia as the largest venture fund in the world.
By my count, together, these two firms manage about 25 times the sum of dollars in total invested in the software industry from 1990 to 2004. The industry as a whole has been totally enthralled with these giant numbers, but perhaps totally uninterested in what these firms actually plan to do with the dollars now that they have them.
To understand why this is a problem, it’s useful to look at the history of what venture capital is and how it came to be.
From roughly 1990 to 2005, you could think of venture capital as the more exuberant, slightly less diligent arm of software private equity, largely funding classical enterprise software. Firms that served as the grandfathers of this industry, firms that still exist today, like Kleiner Perkins and Sequoia, backed companies like Oracle, Siebel, or PeopleSoft that sold installable on-premise software to large enterprises.
The economics of this arrangement were relatively simple. The software itself was a capital expense. It was sold once and installed on customer servers. The bottleneck to growth in this industry was not trillion-dollar product development, but instead relatively linear distribution, specifically hiring direct sales forces that could call CIOs and close six- or seven- or eight-figure contracts to sell the software.
Venture capital, as such, underwrote the sales force. You would raise a round, you would hire reps, you would expand territory, and once the company reached sufficient scale, you would either exit to an acquirer in an adjacent industry or to public markets.
This worked phenomenally well, and as the market got bigger, the companies became more capital-intensive because human distribution became more capital-intensive. The returns were excellent because the margins on installable packaged software are obscene. And this was a relatively straightforward trade. You would put equity dollars in, get sales capacity out, and exit once the equation reached scale.
During the first great Facebook era, we saw the subscription model and the consumer internet expand the scope of possible opportunities without changing the underlying logic. If you think about early SaaS companies like Salesforce, they still needed to build direct sales teams. Consumer internet companies like Facebook still needed to acquire users through paid marketing and scaled operations. But both of them could collect an ongoing, repeating annual revenue from the customer.
Venture companies would become bigger as a result of this, but they would still become bigger as a linear function of venture dollars put into the machine. The internet made the outcomes much, much bigger, so the funds got much, much bigger.
The problem is none of these facts are true of the software industry as we know it today.
Everyone knows this. Ever since we’ve seen the AI trade begin in public markets, growth-stage software companies have traded down to 3-5 times revenue. Compare this to 10 to 15 times multiples of 2021.
The math that justified large checks at the growth stage for largely commodified enterprise software companies just doesn’t make sense anymore. It’s unclear if these companies even needed capital to scale their go-to-markets when the product itself could no longer produce venture returns.
A three to five times multiple in a company at a couple billion dollar scale growing 30, 40% is a great private equity outcome, but it’s simply not a venture outcome, and it doesn’t return a billion-dollar fund, let alone a $15 billion one.
And that is to say nothing of the discontinuity that is the AI market today.
Publicly traded AI names as well as privately traded AI names trade on an entirely different set of assumptions. These are software companies that grow based on capability curves and infrastructure buildouts, and as we’re told, the eventual capture and annexation of all cognitive labor if not including humans themsevles.
These things might be true. They’re certainly not proven today, but they’re not assumptions that most venture funds are equipped to underwrite.
The best private AI companies we see today do not need venture in the traditional sense. They need billions, if not trillions of dollars in infrastructure financing, which they are increasingly sourcing through debt, structured equity, and direct sovereign investment. The venture funds that exist in these rounds and have won allocations to these rounds have done so through brand and relationship, not through their ability to provide or underwrite a differentiated capital product.
Yet, the institutional reality remains unchanged. Fee-able assets must grow. It’s a law of nature. A 2% management fee on a $500 million fund is $10 million a year, but on a $5 billion fund, it’s $100 million.
Gulfstreams are not getting cheaper, and the incentives are obvious.
We must grow fee-able assets, and those fee-able assets must produce a 30% gross IRR to justify the illiquidity premium over public markets. We don’t change those targets. We just tweak the machine until it starts to make sense. That is, we financialize the life out of every single product until that return target is reached.
My friend Jeremy Giffon has a useful frame here.
The idea that it’s hard to beat the market is mostly trotted out by money managers. What they really mean is that it’s hard to do when you manage other people’s money, because it’s difficult to get paid for doing the obvious thing. I think it’s substantially easier than most
Show more
Consider what the obvious thing might be in today’s market.
It might be to sit on cash and wait for better companies to come along. It might be to concentrate into a tiny number of positions where you have genuine conviction, putting huge chunks of your fund into each one. It might be to buy secondaries at 50 cents on the dollar rather than paying full price for primary. It might be to not do any overpriced Series Bs at all, even if the companies are excellent, because the entry price makes the return profile unattractive at the fund size that you’re now at. It might even be to return capital to LPs and let them know that the opportunity set does not justify deployment at its current pace.
None of this is possible for institutional venture capital.
LPs expect capital to be called and deployed on a set schedule that is outlined at the start of the fund, regardless of market realities. Concentration violates diversification mandates in LPAs. Secondaries are a different product that require a different team and a different fee structure. Passing on consensus deals damages reputations and LP perception and re-up rates. Returning capital is simply an admission of failure that stops your next fundraise before it starts.
The structure of the LP-GP relationship makes any of this optimal behavior impossible. You are simply paid to be busy. You are simply paid to deploy. You are simply paid to have an opinion on every deal and be present at every single deal that matters. You are not paid to wait, and you are certainly not paid to return capital.
It’s a trite example, but this is the Warren Buffett problem.
The entire structure of Berkshire Hathaway, regardless of what Warren actually says, is designed to make institutional scale capital behave like you would a personal account. Permanent capital, no redemptions, no pacing requirements, the ability to do nothing for years, and to act decisively with huge chunks of the balance sheet when an interesting opportunity arrives.
Buffett’s famous line that his holding period is forever might make you physically ill when you hear it for the 44th time in a Columbia Missouri conference center, but it is correct. He can hold forever because no one can force him to sell. He can do nothing forever because no one can fire him for it. That is, at least until he fired himself.
99% of venture capitalists managing other people’s money under 10-year fund structures with neat deployment periods, neat LPAs, and neat harvest periods don’t have this luxury. They must invest. They must be busy. They must deploy.
This, self-evidently, is a doom loop. You raise money you can’t deploy well under constraints that prevent you from doing the obvious thing, so you do the non-obvious thing instead, and then you invent reasons why the non-obvious thing will eventually work out.
The immediate problem for the industry is there are no longer terminal buyers.
Venture has always been a liquidity arbitrage. You accept illiquidity for the possibility of outsized returns, and the returns are realized when the company becomes liquid, either through an IPO or an acquisition. The entire model depends on the existence of a functioning exit market.
For most of the last 40 years, that market existed. The IPO window opened and closed with economic cycles, but when it was open, you could get companies out. Strategic acquirers paid premium prices for high-quality growth assets. The music occasionally paused, but it always would come back on.
The IPO window is now fundamentally closed.
There are exemptions, and I suspect OpenAI, Anthropic, and SpaceX will get out for incredible prices. But it is closed in the sense that matters for institutional venture. The median growth-stage venture-backed company cannot go public at a valuation that returns capital to the late-stage investors.
The companies that can go public are an ultra-narrow subset. The best AI names, sterling Madison Avenue quality properties, a handful of private darlings that have defined their categories for decades with exceptional growth, emergence, and mystique, and occasionally a meme or market-driven anomaly. Perhaps a rocket company or a defense company when a weak China feels particularly aggresive.
For the vast majority of the portfolio, the solid companies, the good and not-so-great outcomes, there is no path.
The M&A market also cannot absorb this backlog of companies.
In the model that existed before the last few years, strategic acquirers provided a second clear path to liquidity. Large technology companies bought smaller ones to acquire their technology, their talent, or their market.
This market has meaningfully contracted for reasons that are both structural and regulatory, but certainly not cyclical. Antitrust enforcement has made large acquisitions risky, slow, and often impossible. The major platform companies are under sustained regulatory pressure and have pulled back from aggressive M&A entirely. The private equity firms that once provided a buyout path for mature software companies are themselves facing deployment pressure and have become more selective.
The buyers who remain are disciplined on price and disciplined on asset quality. They’re not going to pay 2021 valuations for 2025 assets.
So the companies sit still. They can’t go public. They can’t be acquired at reasonable prices. They continue to operate, even profitably, generating real value. But the value cannot be converted to cash and returned to limited partners under a traditional venture model.
The venture funds that own these companies mark them the same quarter to quarter, but those marks are not distributions. The LPs need liquidity, the GPs can’t provide it, so the funds extend and pretend. The DPI figures remain stubbornly and consistently low.
If you talked to venture capitalists today, you wouldn’t hear about this reality.
In the absence of real liquidity, we invent eschatological narratives about when and how we’re going to get paid. The language that we’re now using to describe our exit theses has taken the structure of apocalyptic prophecy straight out of Revelation.
There’s going to be a coming moment, a rapture in our current order when the faithful will be rewarded. The details vary, but the form is consistent. It’s millenarianism in a technical sense, a clear and imminent belief in a transformative event that will resolve all present issues and contradictions.
The American Dynamism thesis is in fact a millenarianism based around an apocalypse that involves China invading Taiwan.
The thesis runs as follows. We are in a great power competition. That great power competition will force the U.S. government to procure aggressively from domestic technology companies at the start of a third world war with China, Russia, or insert your great near-peer adversary. Defense budgets will balloon to an even bigger part of GDP. Regulatory barriers will fall. Pete Hegseth himself will procure anduril drones for every single American. The industrial base will be rebuilt with venture-backed companies as their only and sole suppliers.
And when this happens, when this great power conflict enters its hot phase and China moves on the island of Taiwan, the companies in this portfolio will become strategic assets, acquirable at premium prices or IPO-able to every single American that has just been given a war dividend.
The timeline of all of this is exceedingly vague, as is the mechanism, but the conviction is absolute.
You see the same thesis happening in AI.
One wonders how many foundation model companies the public is actually willing to own. But the thesis does not rely on this. Simply once AGI is achieved, the AI will just continuously make money until it buys galaxies, or certainly at least buys all the property in the San Francisco Bay Area.
Bay area employees of OpenAI, Anthropic, and Perplexity are using secondary share sales to fund home purchases, fueling a rebound in the local housing market. The map below shows the hottest Bay area ZIP codes are those with the highest proportion of tech workers.
In fact, this AGI will generate such economic value that the current valuation frameworks we use to assess software are beyond meaningless. The companies that are building towards AGI, or at the very least adjacent to AGI, or maybe even enabled by AGI, or even simply present in the sentence next to AGI, will simply not just be worth billions but will be worth trillions. The returns will be so vast that every American will own a Gulfstream, or at least every OpenAI employee will own five.
The holding period is indefinite but the payoff is infinite.
In this framing, the absence of current liquidity is not a problem but a feature. We’re early and the doubters will be proved wrong and the believers will inherit the earth, or at the very least buy it.
The issue is these are not coherent investment theses. These are articles of faith.
The structure is identical to millenarian movements that have recurred throughout history: a present period of suffering, a future moment of deliverance, and a community of believers who maintain conviction in the face of disconfirming evidence.
The individual assets in these portfolios may be entirely coherent. I for sure certainly wish I owned more Anduril or more OpenAI or certainly more Anthropic at the very least. The companies themselves may be real, the technology may be real and valuable, and the founders may be exceptional and often are. But the liquidity theses, and certainly the claims about how and when these positions will convert to cash, are simply not real.
We have no idea how we will get liquid on any of these names.
This is what happens when you can’t do the obvious thing. You invent reasons to convince yourself why the non-obvious thing will eventually work. You construct narratives that defer the day of reckoning. You find communities of like-minded believers who reinforce the faith. And hell, you might even produce some podcasts along the way.
Where else have we seen this before?
This transformation is certainly not the first. Private equity faced a similar version of this problem in the late 1990s and early 2000s. The leveraged buyout model, the LBO, that had produced spectacular returns from the 1980s began to mature. Competition for deals intensified. Returns compressed as entry prices expanded. It’s a very similar phenomenon to what we’re seeing in the accounting roll-up and other spaces right now.
The funds grew larger, but the opportunity set certainly did not grow proportionally. The exit market became more crowded and less forgiving.
The industry faced a choice: remain a niche asset class producing diminishing returns for a small number of institutional investors, or transform itself.
Over the following two decades, the industry reinvented itself.
Liquidity, which previously was a point-in-time event, the exit via IPO or the sale to a strategic, became a spectrum of products distributed across the entire lifecycle of an investment.
Continuation vehicles allowed GPs to roll winning investments into new structures rather than selling them prematurely. NAV financing facilities let funds borrow against portfolio value to return capital to LPs without forcing premature exits. Secondary markets matured until every brokerage had complete coverage, which allowed LPs to sell their positions to other investors when they needed liquidity.
Hybrid structures proliferated: preferred equity structures, debt, minority recapitalizations, dividend recaps. The industry developed a toolkit for manufacturing liquidity across the entire lifecycle of a company.
This shift, accordingly, required a new LP base.
Traditional institutional investors in private equity funds, the endowments, the pension funds, the sovereign wealth funds, had specific requirements about fund structures, return targets, and liquidity windows. Meeting those requirements while also solving the liquidity problem required capital with different constraints.
Private equity found that capital in the retail and ultra-high-net-worth channels.
For example, Blackstone’s BREIT, a semi-liquid real estate vehicle accessible to individual investors through wealth management platforms, became a common template. By 2023, Blackstone was raising more capital from individual investors than institutions.
The democratization of alternatives was by no means a philanthropic project to invite the mainstream into hoity-toity investment opportunities, but a structural solution to an existing liquidity problem similar to what we see in venture today.
Venture capital must necessarily undergo the same transformation.
The forces that drove the transformation in private equity are similar, if not identical, to the ones driving the transformation in venture capital. Oversized funds, compressed returns, broken exit markets, and the institutional imperative to continue raising more and more high-fee capital regardless of the opportunity set.
The solution will also be identical and similarly profitable. Liquidity will move from a point-in-time event to a spectrum across the lifecycle of the company, and a new LP base with different constraints and different expectations will fill that market and make a lot of money.
The outlines of this are somewhat visible today in premature form.
Secondary markets for venture positions have grown substantially. Continuation vehicles, while still tastefully exotic, are becoming common. Many firms are experimenting with evergreen structures that eliminate the artificial timelines of a traditional fund cycle. Many others are building relationships with, or even buying, family offices or RIAs who have much longer time horizons and fewer liquidity requirements.
Pieces of the infrastructure are being constructed: the transfer agents, the compliance frameworks, the distribution platforms to move venture exposure into the wealth management channel at scale.
Some firms certainly understand the transformation that’s about to occur, but fewer are taking seriously what this will actually require.
The most common gesture to make along these lines is to attempt to build an incubation or origination platform, to manufacture proprietary deal flow by originating companies that you allow your partners to invest in at lower valuations before selling them to retail.
General Catalyst has built a lot in this direction. The logic is not ahistorical, borrowed from private equity, where firms like 3G and JAB built origination capabilities that gave them access to deals outside of competitive auction processes and enriched their partners first and their LPs second.
If you can create companies rather than compete for them, you escape many of the competitive dynamics that compress returns and increase pricing efficiency.
The problem is that to date, many of these efforts have been structurally incoherent.
They want proprietary deal flow, but they also want to participate in hot rounds. They want the optionality of flexible capital, but they’re operating within funds that still demand aggressive deployment periods. They’re trying to manufacture value and mark that value to the same aggressive multiples that their peers would see in competitive deals, creating inflated NAVs that ultimately turn into deep value traps.
You certainly can’t have it both ways.
The returns available to patient, flexible, and unconflicted capital are certainly not available to capital that is impatient, constrained, and conflicted. The incubation platforms are attempts to arbitrage this, to get the benefits of one structure while operating within another. The market will certainly not allow this arbitrage to persist.
Either you are truly outside the consensus, willing to be alone for years, building positions that no one wants at prices that reflect genuine value creation before others see it, like Sutter Hill and many others, or you’re inside the flow, playing the allocation games and the relationship games, capturing the returns that the flow produces.
The firms trying to straddle will forever be forced to choose or die straddling.
The winners of this transformation will certainly not look like venture funds.
They will be massive, patient, product-agnostic pools of unconflicted capital. They will have no deployment clocks forcing them to invest when opportunities are scarce or expensive. They will have no mandates restricting them from buying debt or secondaries or whole companies or doing nothing at all. They will have no LP politics forcing premature exits or performative activity, and certainly not performative podcasting.
They will be able to hold forever or sell tomorrow. They will be able to concentrate into a handful of positions or spread across hundreds. They will be able to wait.
It’s the structure of great family offices, the ones that have compounded for generations while the fund managers cycle through boom and bust. It’s the same structure the sovereign wealth funds use to take 50-year views on capital deployment. But it’s certainly not the structure of venture capital today.
That capital exists on earth, certainly in the wealth management channel, and certainly in a set of progressive sovereigns and endowments that are allowing direct-to-deal and direct-to-solo-manager deployment. But accessing it requires new vehicles, new compliance frameworks, and new distribution relationships.
The firms that build this infrastructure and lock into it early will have compounding advantage over time.
What does this mean for venture capital as we know it?
It means venture will survive at the margins. There will still be funds that raise from institutions, deploy into early-stage companies, and exit through IPOs and acquisitions. That model is not dead. It just will not be the common one, or certainly the bulk of where dollars are going.
The largest pools of capital, the most advantaged structures, and the highest-returning strategies will increasingly sit outside the venture framework. They will look more like Berkshire and more like Blackstone and more like the generational family offices and sovereigns that have always understood that permanent capital with long durations and deep conviction is the only scarce product in the market.
The firms that recognize this early will hopefully adapt. They’ll raise different structures, they’ll cultivate different LP relationships, they’ll develop different and differentiated capabilities. They’ll become something other than venture capitalists. Maybe we can even call them modern deal guys. They’ll retain the networks and pattern recognition that made them successful in the old model.
The ones who do not recognize it will certainly continue to raise funds, continue to deploy capital, continue to mark portfolios at ambitious valuations.
They will be surprised when the music stops, though.
This transformation will not be obvious. It won’t arrive as a crisis with a clear beginning and end. It will happen gradually and slowly, as these things often do.
Many of the best players have already been working this strategy. There are many that I can’t name out of fear for my safety that are already managing very large pools of capital over long durations and building incredibly successful companies, including some of those that can save you 5% on your corporate expenses.
But we’re only just getting started.
Comparing AI venture capital to a millenarian movement is hilarious and unfortunately accurate
All that money gotta go somewhere!
> The immediate problem for the industry is there are no longer terminal buyers.
This mirrors Michael B. Green’s index fund doom loop argument: index funds need active (discretionary) investors to provide the liquidity to deploy inflows and fund redemptions. If index funds reach a critical percentage of the total market (83% accordingly to Green, we’re at 50-55% now), there simply won’t be enough liquidity and the market will freeze. IPO’s are by definition excluded from index funds, which further limits terminal buyers for VC exits.