Aaron Guo

Co-Founder @ Prosights | YC W24 | Previously at GS / Permira

[Cursor @cursor_ai]

Origin, our code hosting platform, is now live.

It's fast, easy to use, and deeply integrated with Cursor.

Get started by syncing your repos from GitHub.

[Aaron Guo @AaronGu86858224]

@OpenAI and @Anthropic are actively diligencing PE portfolio companies for acquisition.

Corporate finance class historically has taught two classes of buyers – the sponsor and the strategic.

Today there's a new buyer in town – the foundation model companies.

They're underwriting a different game than IRR or synergies – they are underwriting the value of the data.

Their diligence comes down to three questions:

1) Replacement cost. What would @Mercor or @Surge AI charge to recreate the dataset, and how many months would it take?

2) Model lift. How much does the data improve performance on audit, tax, or claims? What revenue or research speed can that improvement support? 

3) Rights and provenance. Do customer contracts permit the buyer to use the data for training and evaluation? Can every record be traced to a lawful source?

A flat or shrinking services business with clean rights and hard-to-recreate data can be worth more to an AI buyer than its cash-flow might suggest.

The math has now arrived in private equity. Many of the largest PE funds have taken inbounds from corp dev teams at these labs about specific portfolio companies. They’ve looked at accounting firms, IT services businesses, and specialist BPOs whose systems hold years of expert-reviewed work product. 

For a fund holding one of these assets, that opens a new exit: a solid outcome priced on the data alone.

The reputational cost is harder to price. Sell a portco to a lab, and management teams across the vertical will remember which fund sold the industry's playbook.

It might still be the right call – but you only get to make it once per vertical.

[Aaron Guo @AaronGu86858224]

A partner at @KKR followed the same @Thoma Bravo-owned software company for eight years, waiting for it to come to market.

It finally did – and his team got a first look. No bankers, no other sponsors.

98% gross revenue retention.
115%+ net revenue retention.
Mission-critical software.

A friend of mine worked alongside the deal team. He called it one of the best businesses he'd seen in his career.

It went to investment committee.

The reaction: "Are you kidding me? You want us to buy this?"

The issue was not price. Not rates.

It died on one question: what does AI do to this business over a five-year hold?

They had a preliminary point of view – a few customer calls, a slide on where
AI helps and where it could hurt – but nothing meaty anyone could underwrite.

So the IC defaulted to no. Maybe they're right – nobody will know for years.

This isn't a one-off. I talk to 30+ PE investors a week – this is the shape of many IC discussions right now.

A friend at a large tech fund hasn't worked on a live deal since last year.

Neither has anyone on his team. They're working 40-hour weeks – in private equity – and the VPs around him are interviewing for hedge funds / AI companies.

Everyone expected AI risk to show up in valuations – knock off a few turns of EBITDA and move on.

But surprisingly, that's not what's happening. Many of these deals aren't getting repriced – they're just not getting done because people don’t know how to price them.

PE spent forty years refining its checklist for the perfect asset: strong retention, mission criticality, pricing power, low capex intensity, etc.

You can wait eight years for the perfect company – and by the time it shows up, the definition of perfect has shifted.

This company checked every box. But the boxes just don't answer the only question the room cares about anymore – AI disruption risk.

However, the question isn't unanswerable. It's just not answerable by the deal team alone.

What AI does to a business over five years depends on what the owner does with it over those five years.

A firm with a real operating bench can write their own destiny. A firm without one has to guess it — and nobody is willing to bet their career on a guess.

The best funds in this vintage won't have differentiated analyses on the financials.

They'll have better operating benches who turn AI uncertainty into value creation opportunity.

[Aaron Guo @AaronGu86858224]

Several of the most well-known AI startups are raising right now — across legal, robotics, voice, and infrastructure AI, among others.

This weekend, I caught up with a friend who's a partner at a well-known crossover fund in the valley. He's diligencing these deals.

And he and every peer he's compared notes with have landed in the same place on the numbers: at today's entry prices, each company has to be worth 5x+ its current valuation for the fund to return a 3x MoM after dilution.

None of them feel comfortable underwriting that — especially when the valuations are already in the multi-billions with 50x+ ARR price tags, and public comps are trading at a fraction of that (even on a growth-adjusted basis).

But he's submitting a term sheet anyway.

So I asked him, "Why would you do this deal if you don't think you can get a 3x MoM?"

His answer, more or less word for word:

"I absolutely need to do this deal. If I do it, I'm on the board next to some of the most storied investors in venture. I build the relationship with the founder — and when his friends start companies, I'm the first call."

He kept going: "If it blows up, I got it wrong alongside every tier-1 firm on the cap table. Nobody looks dumb losing money in that crowd. If I ever need to leave my fund, the partners I meet on that board are my next job. And if it actually works? I'm the one who led it."

The fund absorbs the downside, but the individual investor captures most of the career upside.

After all, if he ever moves to another fund, the partners there won't call his old firm for references — they'll call the founders he backed.

So I asked one more question: "Would you coinvest your own money into the round?"

"Absolutely not."

However, the fact of the matter is that, given his incentives, every step of his thinking is rational.

And that's when the valuation finally made sense to me. The round isn't priced off future cash flows. It's priced off board seats, founder relationships, and landing spots — and the others on the cap table are bidding for the same things.

He's not overpaying for the company. He's paying fair value for his career. The two just happen to be denominated in the same round.