Category: Uncategorized

Risk capital in an age of AI and DeFi

In the decade since 2025, algorithmic capital has swallowed much of the private equity opportunity.

In response, “smart private equity” moved downstream to late-stage venture capital. Chasing the remaining alpha, VCs were compressed into the earliest stages.

At the same time, venture’s LPs had long been hungry for a better product with stronger returns and greater transparency.

2027 marked the end of “relationship capital” and sleight-of-hand with proxy metrics. Since the GFC, LPs had tolerated bloated funds, dangerous concentration and slipping performance. The need for a reset was clear.

By 2030, LPs began the transition to pooling their capital in algorithmic treasury DAOs, offering:

• A scalable, automated vehicle that could balance allocation across a range of strategies.
• The transparency of operating on-chain, addressing many of the misaligned private market incentives.
• A decentralised approach, allowing venture capital to reach into every segment of the economy.
• A natively evergreen structure that is less negatively exposed to market cycles and platform shifts.
• A competitive industry with vastly superior returns, where managerial skill earns allocation.

The Role of the DAO

A treasury DAO, in this context, is a pool of LP capital, such as endowments, pension funds, sovereign wealth, fund of funds, or high-net-worth individuals. An LP may choose to manage their own DAO, or join an existing DAO.

Indeed, DAOs are fluid organisations that can merge or split as required. For example, a collective of high-net-worth individuals may form a temporary DAO around a particular opportunity. If a member leaves the DAO, or the DAO is dissolved, the division of outstanding investment economics is handled by the existing smart contracts.

Larger DAOs benefit from access to strategies with more demanding allocations, and the ability to balance the pool of capital across a more diversified set of asset classes. A smaller DAO, on the other hand, may set a much higher hurdle for GP performance but offer above-market compensation contracts.

The DAO automates all financial operations (treasury management, fund transfers, fee payments, report generation, and distributions) through smart contracts.

The Role of Venture Capital

As a result of the market compression, venture capital is now primarily focused on origination. Successful VCs recognise opportunities before they generate enough data to be legible to algorithmic capital. They explore qualitative signals, human psychology and future scenarios in a way that a machine still cannot.

Access to the DAOs is as simple as submitting a ledger of private market transactions. If the ledger demonstrates a good risk-adjusted return in a particular strategy they may be awarded an allocation contract and the associated fee and carry contracts.

Each DAO federates a broad base of independent VCs or small partnerships, effectively solving the top-of-funnel for investment opportunities and fundraising friction for founders, as in every industry, market or region there is likely to be someone on the lookout for opportunities.

Indeed, if their performance is sufficient, it does not matter where a GP lives, works, or went to school. They can participate in the DAO full-time, or part-time. For example:

• In some cases, specialists in fields like biotech or defense can supplement their income with a modest fee stream by surfacing the occasional interesting company.
• Elsewhere, a successful angel investor might convert their track record into a large enough allocation to make investing a full-time career, without jumping through any hoops.

Further downstream, specialist VCs (with deep industry, CFO or internationalisation experience) are still able to find alpha within the later-stage opportunities, shaping successful exits. The highly-automated nature of investment (with investments often made directly by the DAO) at these stages allows them to manage a large book, and focus on signals of overlooked value to pre-empt funding rounds(which, in-turn, the AI learns from).

As each GP’s track record improves, they’re able to compete for incrementally larger allocations at more prestigious DAOs that offer more favourable compensation terms.

Track record is understood through a blend of efficiency (realised IRR), risk (Sharpe ratio) and persistence. This allows a DAO to recognise potential in even relatively amateur investors, as well as downgrading or ejecting underperforming investors.

Emerging Performance

Structural secondaries are core to this model of venture capital, enabled by reduced information friction. The question of how long a startup chooses to remain private is redundant in a market with regular release-valves for liquidity.

This reality aligns with LPs, who want to see faster cash returns. Similarly, VCs that can stay above benchmarks (typically 12-15%, realised IRR) are able to supplement their fee income with a regular drip-feed of carry from liquidity events.

This generally means exiting a significant portion of each position in year 2-4. For example, if a GP sells half of their stake in a company in year 3, at a 3x valuation increase, that’s already a 14.5% realised IRR.

Shortening the liquidity horizon to ~3 years from ~12 years makes understanding GP performance more practical, particularly with an evergreen fund strategy. It also shapes the incentives of downstream investors who are more likely to scrutinise fast-growing investment opportunities for the sustainability of that growth.

Radical Transparency

With all transactions signed with their on-chain identify, each investor (angel or GP) holds immutable and portable deal attribution which may be shared and compared. Thus, performance-based competition is real and important.

“Top VC” lists are based on clear methodology and hard data, and each publisher competes to provide the gold standard of measuring performance across the metrics mentioned above.

VCs that opt-in to sharing their ledger can use their ranking as a public recognition of their ability, feeding into other opportunities outside of access to investable capital.

Venture Firms

While there is no strictly necessary role for venture firms in a world where smart contracts handle a lot of the central functions such as financial operations, compliance and reporting, VCs may still choose to assemble where “the whole is greater than the sum of the part(ner)s”. Most of the time, this manifests as a brand-building exercise to develop inbound interest or a goal of providing more stable results with the aggregate firm performance.

In these cases, informal partnerships may still manage attribution on an individual partner basis, or more formal partnerships may attribute to the firm, with an individual partner as the secondary signature. This way, the firms themselves may compete for ranking (and potentially provide more stable returns) without eliminating the portability and accountability of a partner’s attribution.

Conclusion

In this reimagined model, venture capital works because the structure is built around the desired outcome: managerial skill is focused where it makes the most difference, performance is rewarded with capital, transparency breeds healthy competition, and the broader distribution is able to surface more opportunities and potential outliers.

This represents a new social contract for funding innovation, expanding who gets to invest, who gets funded, and where new category leaders may be found.

(top image: “Webs of At-tent(s)ion”, by Tomás Saraceno)

In 2014, Mark Schaefer coined the term “Content Shock“, to describe the point at which the amount of content being produced outgrew our available time to consume it. The implication being that a lot of content would be written that would be entirely overlooked.

The upshot: increased spend on SEM and social media platforms, and content that was increasingly designed to grab attention.

A similar phenomenon is happening with startups today, as highlighted by Sam Lessin, in his draft letter to LPs.

A summary of the problem:

Investors can only view so many pitches, so they rely on early signals of competence (pitch quality, MVP, early traction) to filter through opportunties efficiently. Not perfect, but it works.

Today, with AI, it’s easier for anyone to develop a high quality pitch, a slick looking MVP, and even some initial revenue. Some of these are companies that would have emerged without AI, but many are just riding the wave and trying to get lucky.

There’s a narrative violation happening right now that a lot of people aren’t talking about. The AI hype is certainly frothy and those frothy rounds are making the news. But the reality is that most AI companies just don’t get funding…

Elizabeth Yin, Co-founder of Hustle Fund

This happens in every cycle:

• Some founders see genuine opportunity to do something important with the new technology.
• Some see an opportunity to exploit investor enthusiasm to raise a load of money.

With AI, the technology driving the enthusiasm also makes it easier to build a startup. Specifically, that it makes it easier to build a startup that passes a typical VC’s filters.

The outcome is a huge amount of time spent chasing dead ends. Which may or may not involve capital deployed.

Screening for Outliers

This isn’t necessarily a problem at the first stage of screening inbound deals, where it’s relatively easy to select for a two basic truths:

• Do we think the problem is important?
• Do we think the founders are credible?

Or, in a pragmatic sense, is the startup on a path that seems likely to unlock vast amounts of economic energy?

For an example of how to scale this process, look no further than Y Combinator. For obvious reasons, that organisation has become very good at screening applications in large volume.

A central element to this process is the single slide format which partners use to review applications. It contains much of the information needed to make the judgement mentioned above, but much more importantly it doesn’t include information that may introduce bias.¹

This isn’t something you should farm out to associates or agents, and neither should you rely on warm intros for signal.

You just have to do the work™.

Recognising Talent

The hard part begins when you turn your attention to a pool of viable opportunities and need to determine which have real potential.

You can make a judgement on whether you think the problem is important, which reflects on the quality of the founders in the “package deal” nature of pitches.

What you can no longer do as easily is test for founder credibility, given the glossy-finish provided by AI tools. This can be broken down in two main categories:

• Do they understand the business model?
  Acquisition costs, unit economics, procurement timelines, cashflow, customer appetite, margins and moats?
• Do they understand the implementation?
  The physics, transaction costs, regulation, R&D, infrastructure requirements and scalability?

Previously, you could get a good understanding of this from a well-developed pitch deck and an MVP. Today, AI can fill in a lot of those answers for otherwise incompetent or poorly-motivated founders.

One clue for how this problem may be addressed lies in Sam’s letter to LPs:

A generation ago to make an investment VCs would say ‘where is your detailed business plan’. The business plan as an artifact served two purposes for investors; (1) it allowed you to ideally actually know / get up to speed on the business, (2) the form of the business plan and the thinking inside it told you something about how smart / talented / diligent the team was.  It was easy for investors to process relative volume of inbound because they could read a few pages of a bad business plan (just like a bad script) and say ‘pass’ vs. ‘oh this person is smart’.

Sam Lessin, GP at Slow

Sam suggests that business plans fell out of favour because (similar to today’s problem) it became easier to build a prototype and put together a nice deck. I’d take that a step further, with some observations learned via Equidam’s 14 years in the business of valuation and fundraising:

The prolonged zero interest-rate period and the surge of SaaS solutions destroyed financial literacy in venture capital. It reduced much of the logic to a simple formula:

CAC: $20
ARR/customer: $4
Net cash: -$16
Multiple: 20x
$1 Invested = $4 in (gross) NAV  

For over a decade, all many VCs cared about was how soon cash-incinerating SaaS growth engines could convert capital calls into NAV inflation and enable subsequent funds.

The upshot was that investors, founders, advisors, and accelerators all forgot the basics of finance and economics. Cash flow doesn’t matter that much for SaaS… Projections are just extensions of the typical growth logic, and don’t serve a useful purpose… Theyr’e all just the same assumptions… The atrophy accelerated as capital flowed more freely, with fewer questions asked.

Another (infuriating) consequence was that in a decade of unprecidented investment in venture capital, most of the money was shovelled into recursive NAV inflation, and very little progress was made on important problems in biotech, energy, infrastructure etc.

By way of example, Equidam would have had a MUCH easier time over the last 15 years if it conjured up some bullshit private market multiples to value SaaS companies. Instead, with the mission to help get funding to the best opportunities, it has stuck by a principled and rigorous approach to valuation.

Particularly, this includes using projections properly, and consequently most of what Equidam does is actually teaching founders how to do projections well, and teaching investors how to parse them.

Fortunately, this too can be reduced down to a simple-ish formula:

• Is the founder’s pitch coherent and credible?
• If so, where do you expect the company to be, financially, in 3-5 years?
• What is the implied value at that point?
• Given typical success rates, anticipated dilution, and time to exit, what is the value today?

Essentially, the difference is that this approach looks at the specific future of the company in order to understand value, rather than applying a generic multiple to past performance. This is a much more appropriate lens when you’re looking for outliers.

The main criticism from ZIRPers is that the future is too uncertain, and projections are too unreliable. This earns them a blank stare and a strong cup of coffee.

The entire purpose of venture capital is to make well informed, rational judgements about the future. If you aren’t comfortable investing based on assumptions, then one of two things is true:

1. You should not be working in venture capital.
2. You don’t understand the assumptions.

(Though perhaps the second point also points to the first.)

This leads to one of the most misunderstood points about valuation, which I’ll try and use to pivot this article towards a conclusion:

Valuation is not intended to determine the right price.

It is an exercise which, when done properly, helps both sides of a transaction understand the assumptions and align their expectations. From that scenario, you get an indication of value which can inform the price you agree on, but the value is in the process itself.

Thus, the key here, and what has been missing from venture capital for more than a decade now, is a return to financial literacy, rather than financialisation.

Sit down with founders. Talk through their strategy together, using both the deck and the financial model as a guide. Connect it with questions about the business and technical implementation. Their answers, guided by your questions, should slot together to create a coherent and credible picture of the future.

Done well, this should give you a huge amount of insight into the founders, and their company.

Indeed, it turns out that founders who go through this process actually have a significantly easier time raising capital. This is one of the many reasons I’m so irate when people tell founders to “let the market price the round”. The process of valuation is so valuable, so useful, and so important to innovation.

So, you’re going to see a lot more startup pitches in your inbox over the next couple of years. It would be a grave mistake to ignore them, so you need to figure out how to embrace this process.

Step 1: Learn from Y Combinator’s process for screening deals. Particularly, learn what not to look at in order to reduce the vectors for bias in your process. Limit as much as possible.²

Step 2: Find a way to dig into this “complete package” of strategy, business model and technical implementation in a way that establishes founder credibility and can’t be faked with AI.

It doesn’t have to be all at once. You can start by asking them to send you a video talking through their financial model and deck to explain the underlying strategy and how it connects to revenue/profit targets and milestones. Eventually, though, it should probably be in-person, where hopefully you can get a sense for their passion and tenacity.

(header image: ‘Deluge’ by Ivan Aivazovsky)

1. I can see why YC includes traction for their objectives, but I’d argue it’s perhaps a negative to include for a seed fund. i.e. it may bias your review towards companies with traction, rather than the best companies. [↩]
2. To be honest, I don’t even like that the YC template includes the logo. It seems like unnecessary noise which may colour opinions early on. [↩]

I had a meeting once with Howard Marks, who I had wanted to meet for a long time. He’s a famous bond investor that does a lot of writing. For 15 minutes he asked me questions about the venture industry, a lot of structural questions, and I answered him as best I could. 

He said, “That’s a really shitty industry”. I said, “Why do you say that, what do you mean?”, and he said “Cyclical collapse is built into the structure.” 

Bill Gurley on his conversation with Howard Marks, All In Summit, May 2022

---

‘2024’ could have been a short chapter in the history of venture capital. Another post-boom year of declining deal activity, recovering valuations, recaps and shutdowns. We left it with the public markets looking healthy and hope that IPOs might reappear on the horizon. All good reasons to flip the page to 2025. 

Before you move on from 2024 entirely, consider the boiling frog. Venture capital has been quietly outgrowing traditional definitions over the last 15 years. Last year, as AI drove heat into a cool market, the cracks started to appear. 

The venture capital industry has historically worked as a relay race where investors at one stage bring in the next stage of investors to fund and support companies as they scale. Right now, it feels like AI investing breaks this model in a few important ways.

Charles Hudson, Managing Partner at Precursor Ventures

This isn’t just about inflated valuations, overcapitalisation and ZIRP. It’s about the effort of a few firms to outgrow the boom-and-bust nature of venture capital through a new model for investing. 

---

Consider the well documented rivalry between Andreessen Horowitz and Benchmark, with their opposing views on capital abundance in venture. 

Bill Gurley declined my requests for comment, but he has publicly bemoaned all the money that firms such as a16z are pumping into the system at a time when he and many other V.C.s worry that the tech sector is experiencing another bubble. So many investors from outside the Valley want in on the startup world that valuations have been soaring: last year, thirty-eight U.S. startups received billion-dollar valuations, twenty-three more than in 2013. Many V.C.s have told their companies to raise as much money as possible now, to have a buffer against a crash.

“The Mind of Marc Andreessen“, by Tad Friend

Benchmark has set the standard for “traditional” venture capital, managing small and efficient funds, close relationships with founders, and exemplary returns. In contrast, Andreessen Horowitz represents the era of ‘capital of a competitive advantage’; a confluence of hyper-scalable SaaS, digital growth channels, and historically low interest rates. Not merely a competing venture strategy, but the emergence of a fundamentally new product that trades efficiency for scale.

The divergence of Andreessen Horowitz (and a few similar firms) from the traditional playbook has created some confusion, as greener GPs attempted to emulate aspects of both. Lacking either the AUM of Andreessen Horowitz or the discipline of Benchmark, they inevitably hit the slippery slope of badly managed risk. This can also be seen in the muddled logic on topics like entry multiples, valuation and concentration, where the ‘common wisdom’ has often made little practical sense. 

Fortunately, there is light at the end of the tunnel: As this new product is better understood by LPs, GPs and founders, the chaos is better controlled: LPs will have more realistic return expectations, benchmarks and timelines. GPs will be able to identify coherent strategies and compatible advice. Founders will have a clearer choice between traditional venture capital, and the new model of “venture bank”, and the expectations associated with each.

Too Big to Fail

In 2009, with the launch of their first fund, Ben Horowitz and Marc Andreessen slipped into the role of VC brilliantly. They had strong opinions about capital efficiency and portfolio structure, a level of sophistication that surpasses many of today’s GPs.

Despite the success of that first outing, capturing huge outcomes like Skype, the market soon presented a new opportunity. In 2010, rebounding from the GFC, incumbent firms were raising funds in excess of $1B—a throwback to the dotcom exuberance that Benchmark had criticised. Institutional money was once again on the hunt for new opportunities and software had begun ‘eating the world’.

As entrepreneurs (another contrast to Gurley) Horowitz and Andreessen felt no loyalty to the traditional playbook. They saw an opportunity to innovate; the chance to build a scalable model for venture capital and take a dominant position. So, they invested heavily in media to build a ‘household name’ brand. They sought network effects by launching scouts programs and accelerators. Platform teams were assembled to handle their growing portfolio. Rather than an investment team, they built something more like a sales force. Essentially, they exploited a “boom loop”: raising money to maximise exposure, manufacture category winners, distribute healthy markups and raise more capital. In this environment, venture capital became a cutthroat zero-sum game with very little discipline. 

Some of the VCs who funded predation succeeded spectacularly, and the basic incentives of venture investing that tempt VCs to employ this strategy persist. The goal of antitrust law is to push businesses away from socially costly anticompetitive behavior and towards developing socially valuable efficiencies and innovations. We think that Silicon Valley could use a nudge in that direction.

“Venture Predation“, by Matthew T. Wansley & Samuel N. Weinstein

Jump to mid-2022, and the post-ZIRP hangover experienced by the venture market. Almost everyone had gotten caught-up during the decade of near-zero interest rates capped off by pandemic-driven irrationality (except Gurley, who had stepped back from Benchmark in 2020). The critics of the Andreessen Horowitz model began to emerge, pointing to their inflated check sizes and irrational pursuit of “Web3.0”. Now, like everyone else, Andreessen Horowitz would surely suffer from falling returns and backlash from LPs? 

It happens that Andreessen Horowitz raised more than $14B in new funds during the first half of 2022, close to the combined total of their funds raised in the preceding three years. They were well prepared for the fundraising winter of 2023, a year which many large firms used to rebalance their portfolios through secondary transactions—playing on the extreme heat around companies like OpenAI, SpaceX and Anduril. 

In 2024, as the slump continued (down roughly 50% from the funds raised by VCs in 2020 and 2021) Andreessen Horowitz came back and raised another $7B. In fact, the 10 largest venture banks raised more than half of all venture dollars last year. The firms that had contributed most to the collapse were suffering from it the least. How could this be? The outcome frustrated many smaller GPs who had been squeezed out of deals and were having a harder time closing funds. 

Plans Measured in Decades

If you want to build a venture bank, you need billions of dollars on a regular basis, and it needs to be truly patient capital. Not high-net-worths, not family offices, perhaps only the largest pension funds and endowments. Most of all, you need sovereign wealth. 

This transition away from venture capital’s traditional LP base was the central gambit of venture banks. With performance stabilised by the scale of their AUM, the ability to extend into other assets and an internationally recognised brand, they are able to court some of the world’s largest pools of capital. Rather than hunting alpha, they farm ‘innovation beta’—indexing across as much of the fast-growing tech market as possible, with exposure to every hot theme, at every stage, both private, public and tokenised. 

A lot of the detail is lost when you’re operating at that altitude. Valuation, consensus, discipline, markets… they all matter less if you’ve escaped the typical venture capital fund cycles. Liquidity can be delivered through the secondary market, continuation funds, distributions of public stock or sales of crypto holdings. If it’s a tough market for IPOs, you have the scope to wait it out. When there’s a ‘liquidity window’, like 2021, you can jettison as much as possible—as quickly as possible. 

If success in venture capital is primarily about being in the right place, at the right time, the proposition of venture banks is to be everywhere, all the time.

Our results suggest that the early success of VC firms depends almost entirely on having been “in the right place at the right time”—that is, investing in industries and in regions that did particularly well in a given year.

“The persistent effect of initial success“, by Ramana Nanda, Sampsa Samila and Olav Sorenson

There is also opportunity in this bifurcation for GPs at traditional firms. While venture banks smother consensus themes, they do so on behalf of this new class of supersized LP. For everyone else, there should be less competition in the traditional LP base—once they recover their enthusiasm for the asset class. 

The shift obliges smaller firms to focus on non-consensus themes—which they should have been doing all along. Access to ‘hot deals’ is no longer a core proposition for LPs, as it becomes clearer that the juice isn’t worth the squeeze. Instead, GPs should focus on evergreen investing theory. This includes portfolios structured to optimise for outliers and manage risk and a sourcing process that minimises bias, or what Joe Milam calls “Process Alpha”.

The conventional venture model relied on the power law to rationalize overly concentrated portfolios and “hot deal chasing” when basic portfolio management theory emphasizes (non-systematic) risk management through diversification.

“How to Construct and Manage Optimized Venture Portfolios“, by Joe Milam

Essentially, venture capital goes back to being a positive-sum game that effectively finances innovation, and looks less like a ponzi scheme.

A number of other positive consequences could emerge from this correction: 

In moving away from consensus themes, VCs will be forced to develop an approach to understanding the value in novel propositions.

So much funding in the valley has rushed towards the consumer, at the expense of what we would argue are more significant projects.

Nicholas Zamiska, Office of the CEO, Palantir

Some of that shift has been cultural, but much is simply that the pervasive practice of pricing with ARR multiples has favoured well-understood segments over the last decade. Deep tech and hardware suffered because VCs simply weren’t sure how to put a price tag on those companies. 

Calculating or qualifying potential valuation using the simplistic and crude tool of a revenue multiple (also known as the price/revenue or price/sales ratio) was quite trendy back during the Internet bubble of the late 1990s. Perhaps it is not peculiar that our good friend the price/revenue ratio is back in vogue. But investors and analysts beware; this is a remarkably dangerous technique, because all revenues are not created equal.

Bill Gurley, GP at Benchmark

As VCs develop a keener and more independent sense for value, the exit pipeline begins to look a lot healthier. Metrics that look at financial health in addition to sheer growth of revenue will more effectively align companies with the expectations of both public market investors and prospective acquirers. Like the old days of venture, much of growth and prosperity will be available for mass market retail participation — which is great for the ‘soft power’ of tech. 

Some of the world’s most highly valued public tech companies entered the public markets with quite modest valuations, at least by today’s standards. Microsoft, Amazon, Oracle and Cisco all debuted with market caps south of $1 billion. Of those, only Microsoft topped $500 million. This translated to relatively modest gains for their private market investors, compared to the massive value appreciation they have all experienced post-IPO.

“Who’s reaping the gains from the rise of unicorns?“, by Adley Bowden

On the other hand, venture banks who have developed monster companies like OpenAI, SpaceX and Anduril, will be able to keep tapping into this expanded pool of private market capital via secondaries. Where companies are strategically sensitive, particularly in the realms of AI or defense, this may be seen as preferable to public company disclosure requirements or the threat from activist shareholders. 

Secondary markets also represent an opportunity for traditional venture capital. As managers pick up better standards for valuation, providing greater transparency and explainability, activity is bound to increase. This means more options for liquidity, more consistent buy-sell tension to keep prices under control, and participation from a more diverse base of investors. Finally, it also offers greater cap table flexibility to founders and early investors.

Let’s be realistic here; you’re better off in the fullness of time if certain players are in your cap table, and not a seed fund. 

Mike Maples, Jr, Co-Founding Partner at Floodgate

Clarity Precedes Success

2025 should mark the end of confusion about venture banks and their role in the ecosystem. After a decade of increasingly confused standards and benchmarks for traditional VC, the two products have been neatly split by the wedge of AI.

This is an opportunity for rehabilitation, as the confusion around the change has created more issues than the change itself. Both sides are now able to lean into their strengths, consensus vs non-consensus, alpha vs beta, and deliver the commensurate returns to LPs with matched expectations. 

Make no mistake, 2025 will be a crucible for venture capital. The divergence of exits from public market performance is worse than ever, and confidence is low. Despite this, there seems to be a resurgence of smaller firms raising capital—matching historical patterns where incumbents lead the rest of the market by a year. 

Critically, how will smaller GPs execute in the new environment? Will they learn the lessons from the last five years and play to their strengths? Or will they keep trying to span the chasm of small fund risk tolerance and large fund risk appetite?