The Bifurcation of Capital is Inevitable

What Berkshire Hathaway Teaches Us About the Future of Venture

This is a free weekly newsletter about the art and science of building and investing in tech companies. To receive Investing 101 in your inbox each week, subscribe here:

---

In 2009, a 24-year-old Harvard Business School student named Tracy Britt Cool did the thing every ambitious finance kid is told to do but never actually does. She wrote Warren Buffett a letter. Cold. No connection, no warm intro, no mutual friend to vouch. Just a kid from a Kansas farm who’d decided she needed a mentor, and who did she want? (Checks notes). Oh, just the greatest capital allocator ever, why not?

But in a fairy tale ending only people who read 10-Ks can romanticize, he said yes. She joined Berkshire Hathaway as his financial assistant, and over the next decade she went on to run a string of Berkshire subsidiaries; Pampered Chef, Benjamin Moore, Larson-Juhl, Oriental Trading. She became one of Buffett’s protégés. Felt like an heir-apparent story was writing itself.

But then, in 2020, she left.

She didn’t leave to go run a multi-billion dollar hedge fund. She wasn’t sucked in by the siren song of a SPAC, a crypto exchange, or even a growth-equity megafund. She left, with her partner Brian Humphrey, to start a firm called Kanbrick whose explicit purpose was to buy the businesses Berkshire had gotten too big to buy. In her own framing, she wanted “businesses too small for Berkshire.” We’re talking family-led companies doing $5M to $50M in earnings, the kind of business that simply doesn’t move the needle on a trillion dollar balance sheet.

There’s a critical learning here that will increasingly become relevant to venture capital. The most telling thing about any capital-allocation machine is the set of deals it can no longer afford to do. Tracy Britt Cool’s departure wasn’t anything to do with a lack of success, but the overabundance of it. Berkshire succeeded so completely at one game that it could no longer play the other. And that, in miniature, is the whole story of where venture capital is going.

The Curse of Size & Scale

A couple years ago, I went back and read every single Berkshire Hathaway annual letter. Throughout those letters, there’s an idea that I’ve highlighted probably more than any other. Countless instances where Buffett talked about the cost of being enormous, and he puts it bluntly:

“A fat wallet, however, is the enemy of superior investment results. And Berkshire now has a net worth of $11.9 billion compared to about $22 million when Charlie and I began to manage the company. Though there are as many good businesses as ever, it is useless for us to make purchases that are inconsequential in relation to Berkshire’s capital... We now consider a security for purchase only if we believe we can deploy at least $100 million in it. Given that minimum, Berkshire’s investment universe has shrunk dramatically.“

And that was in 1994! Today that number is a TRILLION dollars. What changes is not the companies that exist or, in the case of venture capital, those getting built. There are “as many good businesses as ever.” What changed is that Berkshire got too big to care about them. When your floor for a single position is $100 million, an exceptional little company doing $8M in earnings is, mechanically, a rounding error. You couldn’t be bothered. The size that makes you powerful is the same size that makes whole categories of opportunity invisible to you.

In my notes on the Berkshire letters, I made the same note in the margins three or four different times, because it kept showing up: the curse of size and scale. And later: Bigger funds move up market exposing a long tail of businesses that they’ll no longer compete for. And later still, the cleanest version of the idea: massive pools of capital will be a category unto itself.

That’s the exact bifurcation in venture capital I’ve written about before in pieces like The Puritans of Venture Capital. As capital compounds (whether through performance or fundraising), it is mechanically forced upmarket, and in moving upmarket it abandons a long tail of perfectly good opportunities that it is now structurally incapable of pursuing. It’s just math! A $40 billion pool of capital and a $100 million pool of capital are not the same animal doing the same job at different scale. They are different species, and they hunt different prey. Or at least... they should be.

I’ve written over, and over, and over, and over again about “Cottage Keepers vs. Capital Agglomerators.” But the cleanest articulation of why it happens isn’t anything I’ve written about. It’s in a Buffett letter from the 1990s, written about a candy company and a railroad.

Don’t Play Gin Rummy

Now, for those of you who aren’t 90+ year old survivors of the Great Depression, Gin Rummy involves discarding your least promising card. That’s all I cared to figure out. Okay, moving on.

When people thing about Buffett, they often talk about him as a stock picker. And he certainly picked stocks. But the structural genius of Berkshire is the structural ability to never have to sell. Here’s how Buffett explains it:

“You should be fully aware of one attitude Charlie and I share that hurts our financial performance: regardless of price, we have no interest at all in selling any good businesses that Berkshire owns ... gin rummy managerial behavior (discard your least promising business at each turn) is not our style. We would rather have our overall results penalized a bit than engage in it.”

Gin rummy is the perfect metaphor, because it names the thing most investors can’t help themselves but get focused on; selling as often as buying. Just one example from this past week was Mohnish Pabrai, a self-described “clone“ of Warren Buffett, explaining why he put 77% of his portfolio into Micron in 2023 and then sold all of it just six months later, before Micron’s stock went up 15x...

I couldn’t find it, but there was a comment on the tweet where someone said, “all he had to do was nothing!” It reminded of a classic quote from Blaise Pascal that is a favorite of Buffett and Munger; “All of humanity’s problems stem from man’s inability to sit quietly in a room alone.”

However, the inclination to sell is, in large part, structural. PE and VC funds have a fixed life (raise, deploy, harvest, return) across seven to ten years. Those funds are contractually obligated to discard. Playing gin rummy is in the formula, whether you like it or not. The fund schedule forces selling. Buffett built a structure that does the opposite. A structure that has become the new dream of every would-be “AI Operator.” A holding company. A structure for, as Buffett describes them, a “buy and never sell” investor.

But here’s the really important functional distinction about buy-and-never-sell. Buffett makes it clear that its not just a preference for the businesses you know over the ones you don’t. He frames it, effectively, as a neutralization of distinction in investment decisions. The decision to retain earnings is the same as evaluating a new investment. It’s not that you default to never sell; it’s that you treat selling with as high a bar as you treat buying. A hedge against sunk cost fallacy.

Take one of Buffett’s more famous hokey investments; See’s Candies. Every dollar a See’s Candies throws off is a dollar Buffett gets to redeploy into the greenest pasture available, anywhere in the empire. He doesn’t have to give it back to LPs. He doesn’t have to time a sale to a fund deadline. He just routes it. He said it most plainly when describing the deal that turned candy money into a railroad:

“We also measure businesses against opportunities available in marketable securities, a comparison most managements don’t make ... Our flexibility in respect to capital allocation has accounted for much of our progress to date. We have been able to take money we earn from, say, See’s Candies or Business Wire ... and use it as part of the stake we needed to buy BNSF.”

See’s Candies operates, he notes elsewhere, on only $25M of net worth while having sent more than $410M in pre-tax profits up to the center “to deploy in whatever way made most sense.” The beauty of the holding company is as a capital-routing engine. And the first law that governs all of it: “what is smart at one price is dumb at another.”

But the same flexibility that lets Berkshire never sell is exactly what drove it upmarket. Once you’re routing tens of billions, you can’t route them into an $8M candy shop. That’s the form of permanent capital, owner-operated, allocated by judgement rather than by a fund clock. Tracy Britt Cool didn’t leave because she had a philosophical disagreement with Berkshire’s scale, it was just acknowledging that she could rebuild at the size Berkshire was forced to abandon.

Venture’s Buy and Never Sell?

Let me turn from Omaha and its intellectual adjacencies to Sand Hill Road now as I think about how these two disparate histories are sort of colliding. Like I said, you’re seeing the exact bifurcation in venture that Buffett’s letter and Tracy Britt Cool lived in real time.

The Capital Agglomerators are completing the move upmarket into a pure AUM game. I wrote about this in The Blackstone of Innovation: “any asset manager is in the business of multiplying 2% by as large a number as possible,” and “business building is the marketing, but asset management is the business model.” That was 2022. It’s only gotten crazier since I wrote that. In January 2026, a16z raised a $15B fund, one of the largest venture raises in history if you don’t count stuff like whatever the eff SoftBank Vision Fund was. a16z is now at more than $90B in AUM, neck-and-neck with Sequoia, and captured over 18% of all US venture dollars in 2025.

Ben Horowitz now describes the mission as “ensuring that America wins the next 100 years of technology.” They’ve always avoided the small partnership vibe, but that’s taking on the language of sovereign-scale capital agglomerator that happens to have started in venture. And the gin rummy math is exactly Buffett’s fat-wallet problem. I’ve written before about how, for a megafund, a $1B outcome is “meaningless.” You need exits “north of $50B,” despite the fact that fewer than 50 public tech companies have ever hit that mark. The agglomerators have moved so far upmarket that the universe of deals that can move their needle has, in Buffett’s exact phrase, “shrunk dramatically.”

The Cottage Keepers, what I called the Puritans, are getting rarer, not stronger. I used to frame Benchmark and USV as the durable counterweight, the firms that refused to grow because “your fund size is your strategy.” But, as I wrote about in Venture Capital’s Fourth Turning, I had to mark the moment the most staunch champion broke ranks when, in June 2026, Benchmark raised a $2B vehicle, its first growth fund, the first time it had meaningfully scaled up since 1999. The gap that opened is almost comical. a16z at ~$90B against Benchmark at roughly $4.8B, an 18-to-19x chasm.

The data reinforces the idea that this is the new normal; established mega-funds took 79.4% of all 2024 VC dollars, and in Q1 2026 about 73% of LP capital went to just five firms. The Cottage Keepers of yesteryear are now rarer than unicorns. The middle, as I keep saying, is being spued out.

But here’s where I think the most interesting opportunity exists, and what got me thinking about the long tail of abandoned opportunities by Berkshire. When you think about all the things that are actually changing, whether its AI Roll-Up HoldCos, solo capitalists managing $1B+ of “owned capital” vs. “borrowed” LP capital, I think there is an opportunity for permanent capital amidst startups.

Buffet’s permanent capital famously came from insurance float. I always think about that as a framework for permanent capital; what would be the insurance float today? Mr.Beast has viewership as “float” that he can use for ad dollars, but also for selling his own products. I’m seeing a lot of AI companies that can quickly spin up products at high margin, spitting off cash, but with literally zero defensibility; like a melting ice cone. The concept of “float” unto reinvestment is a really interesting topic for exploration.

The answer of how, however, I don’t have a great answer to. Which new vehicle inherits Berkshire’s permanent-capital freedom, and where does its float come from? Buffett’s edge was never just discipline. Insurance float was a perpetual, near-zero-cost capital base that freed him from ever having to be a forced seller. The AI native HoldCos have Buffett’s temperament (e.g. buy good businesses, never play gin rummy, route the cash flow to the greenest pasture) but most of them don’t yet have his structure. They still answer to investors who, eventually, want their money back. I’m curious about the firm that figures out the modern equivalent of float that allows for durable, patient, doesn’t-have-to-be-returned capital base, with the ability to target the “weirder” long-tail of startups that the Capital Agglomerator’s scale will eventually force them to avoid.

The giants will keep swelling in size. The AI rounds will keep looking like sovereign wealth; the agglomerators will keep climbing upmarket until the only deals left to them are the 5-10 companies on Earth large enough to matter. Fine. Let them. It is, increasingly, not really venture capital anymore The more relentlessly capital concentrates at the top, the larger and more underserved the long tail of opportunities at the bottom becomes. And, as a result, the more valuable it is to be the patient, permanent vehicle that finally shows up for it. LPs, despite their DNA, will often force you to play the game you, increasingly do not want to play.

---

Thanks for reading! To receive Networked Conviction: My Investing Journal, a collection of portfolio updates, Requests for Startups, and investing ideas for paid subscribers, you can sign up below: