Papa Chakravarthy

Just read the Cambridge Associates numbers on US PE and VC. Over 25 years, PE has returned ~12.1% a year. VC ~8.0%. And it’s not just the long view — PE has beaten VC across every horizon, from 1 year to 20 years. The headline is simple: private equity has been the more consistent wealth builder. Venture hasn’t kept up. But that’s only half the story. VC has delivered explosive cycles — years where it looked unbeatable. The problem is what came after. The drawdowns erased a lot of the highs. Meanwhile, PE just kept compounding. Less drama, more predictability. And that’s what shows up when you stretch the lens. So the real question for allocators like us: are we overvaluing optionality in venture, or undervaluing the comfort of compounding in PE? In a world of AI, biotech, and energy transitions, does the risk/reward shift? Not sure yet. But it feels like the next decade may test whether this gap is structural… or just a long cycle waiting to turn. #PrivateEquity #VentureCapital #AlternativeInvestments #FamilyOffice #InstitutionalInvesting #Compounding #InvestmentStrategy #MiddleEastInvesting

97

12 Comments

Only a handful of freshman biotech venture funds have uncorked in recent years. But former Moderna leaders Evan Rachlin and Lorence Kim have been able to collect $325 million for their inaugural fund at their Florida firm Ascenta Capital. They had targeted $350 million but decided to close the first fund early in the summer because their six-company portfolio was gaining momentum. “It’s a tough environment to raise a new fund focused on biotech that doesn’t have the letters ‘A’ and ‘I’ in it, but we did do an AI deal, actually, in biotech [Iambic Therapeutics] that’s going great,” Rachlin said in an interview with Endpoints News. “You do a couple of early deals, got some good momentum, more capital came in, did a couple more and it became this virtuous cycle once we cracked that.” #ascentacapital #adarx #iambic #orsobio #alpha9oncology #odysseytherapeutics #cardurion #biotechventurefunding #biotechvc https://lnkd.in/gQWRf5XS

60

After years working with managers moving funds on-chain, I’ve noticed the same pattern: The hard part isn’t deploying contracts. It’s making the right structural decisions upfront. So we at Enzyme have summarized the 5 questions every on-chain fund needs to answer. These 5 design choices matter more than the strategy itself. Made a high-level breakdown for managers & allocators → just swipe through.

29

5 Comments

I’ve spoken to over 2 dozen MDs at PE firms I can confidently say that the arb of figuring out how to implement Vertical AI at portfolio companies is very real right now It will fundamentally change underwriting for those who can do it predictably and unlock generational returns. Most are aware they need to act. Very few have.

453

64 Comments

We get this question a lot: “Aside from fees and minimums, how’s Frec different from AQR Capital Management for long short direct indexing?” Frec, like BlackRock's Aperio and a few others in the space, has a different philosophy than AQR when it comes to long short portfolio construction. We let customers tilt their portfolios using classic factors (such as value, momentum, etc), whereas AQR has a proprietary “one-size-fits-all” alpha model. At Frec, we believe letting customers choose their own factor tilt is a more personalized approach. For example, some customers may be heavily invested in technology and growth already, and may prefer exposure to value + quality. Others may want exposure to growth because they’ve been conservative with their allocation and missed out on the AI run. Frec curates various factor tilts and historically backtests each across a full economic cycle. The past does not predict future performance, but our simulations provide a good picture of how the tilted portfolios performed relative to benchmarks. TL;DR AQR is a hedge fund, while Frec is a tech investment platform. AQR’s “one-size-fits-all” alpha model has limited scale, so they may be incentivized to increase fees and minimums. Our model is scalable so we are incentivized to broaden access by reducing fees and minimums. And finally, Frec is the only direct-to-consumer offering in the space. Customers can get started with long short in minutes without an advisor.

35

5 Comments

Industry LLMs are here, and Anthropic just announced a big one this week: Claude for Financial Analysis. It’s a platform that ties together sources like PitchBook and S&P to help you analyze markets, conduct research, and make investment decisions. More than a tool, it represents a strategic move to go from general-purpose AI to vertical-specific solutions tailored to industry needs. These launches are significant for #businesses and #startups alike, and partially illustrates why it’s both exciting and difficult to build in this era of AI. Things change so fast. If you’re a user, think about how this might reshape your workflows. If you’re a builder, think about what would differentiate your product in the long run even as the landscape keeps changing. Big Tech will always build platforms, but the startup advantage is deep vertical expertise because even tools like these won’t be right for every company (cost, privacy, specific needs, etc.). This is like when Apple built the iPhone, which gave way to so many category defining apps across social, finance, commerce, health, travel, etc. There are still a lot more apps to be built! [Source Credit: Anthropic website for the visual]

45

In my last post, I discussed the central importance of high-quality, curated data in applying #AI to Event Driven investing. (Refer to my last post here: https://lnkd.in/eyGJKzTA) But even the best data isn’t inherently predictive. It must be transformed into something models can learn from. That’s where FEATURE ENGINEERING comes in. At Versor Investments, feature engineering is the critical bridge between raw data and model-ready inputs. It’s a blend of domain expertise, investment intuition, and statistical techniques — and in #MergerArbitrage, it’s often where much of the real predictive power lies. We employ a variety of techniques to extract useful signals from complex datasets. Dimensionality reduction helps simplify high-dimensional data while preserving meaningful variance and minimizing noise. Natural Language Processing (NLP) allows us to quantify sentiment and language tone extracted from deal press releases, call transcripts, regulatory filings, and merger agreements. These methods help uncover subtle signals embedded in the unstructured aspects of deal flow. The features we engineer are highly tailored to the structure of M&A events. We capture regulatory timelines, estimating how long antitrust or regulatory reviews may take. We incorporate types of required approvals and the complexity of the regulatory environment. We build features based on deal structure — including stock versus cash offers, transaction size, and the strategic rationale behind the merger. Historical analogues also play a role, using past deal characteristics from similar industries and sectors. At the firm level, we incorporate financial health metrics for both acquirers and targets, liquidity conditions, and shareholder concentration. Board composition overlap between the two companies can also be informative in assessing deal dynamics. Importantly, these features don’t just feed a single prediction. They allow us to estimate multiple dimensions of deal risk — the probability of deal closure, likelihood of failure, and potential for competing bids or price revisions. Ultimately, great models start with great features. And great features come from deeply understanding both the data and the domain. In my next post, I’ll share how we train models based on these engineered features — and how we adapt as markets and data evolve. Stay tuned. For more on quantitative investing, subscribe to our blog: https://lnkd.in/eQTV_7hh

24

ICYMI: Venture capital juggernaut Andreessen Horowitz (a16z) pulled a power move at this year's Tech Week — and it's become the conference's most compelling niche drama. Tech Week, whose current iteration was started by a16z in 2022, returned to New York City this month with a full calendar of events just as the city started heating up for summer. The conference's "decentralized" nature means anyone can throw an event and apply to have it added to the official calendar, provided they follow a set of rules. However, one of this year's new rules has drawn criticism from some organizers and attendees: Official tech week events must use Partiful. "The only official events platform this year is Partiful," the 2025 Tech Week guide says. That's left Luma, a competing events platform popular at previous iterations of Tech Week, out in the cold. Versions of Tech Week's events FAQ document from 2023 and 2024 said events could use either Luma or Partiful. And here's what's gotten tongues wagging: a16z is invested in Partiful, but not in Luma. w Sydney Bradley and Samantha Stokes https://lnkd.in/gnGmQnDN

47

7 Comments