Alphabet is Berkshire Hathaway 2.0
Earlier this week, Alphabet announced Berkshire Hathaway was investing $10bn in a private placement as part of a larger $80bn equity capital raise to expand AI infrastructure and compute. While I have a lot of thoughts on the announcement and what it means for markets, I had just circulated a memo to friends the week before on why “Alphabet is Berkshire Hathaway 2.0.”
Given the news, it felt like a good time to share that memo more broadly. It traces the connections between Berkshire and Alphabet over the past 25 years, covering both direct tactical influences and the subtler strategic ones, and shows how Alphabet has not only learned from, but evolved Berkshire’s model to suit its own needs.
Twitter has been having a field day with Greg, making dozens of memes about gambling now that Buffett has stepped down as CEO—but those takes miss the tight, decades-long connection (and thinking) shared between the two companies.
Behind the paywall is ~5,000 words on initial thoughts on the raise itself: why the raise isn’t what the headline says it is, why Alphabet is issuing equity rather than debt, the underlying benefits that Berkshire and Alphabet are each likely to get out of the deal, and some remaining open questions.
The memo is reprinted below as initially completed.
Memo
Date: May 28, 2026
Re: Alphabet is Berkshire 2.0
Introduction and Executive Summary
In a May 2025 memo on what the market was missing on Google while it traded at 13x earnings, I wrote: “Alphabet is Berkshire Hathaway 2.0”:
Everyone in the investment world talks about finding the next Berkshire. Alphabet is already there—and arguably better positioned. Google’s IPO letter was modeled on Buffett’s shareholder manual; the founders were heavily influenced by Buffett after meeting him. The 2015 restructuring reorganized the company along Berkshire’s holding-company logic, separating the cash-generating core from a portfolio of long-duration bets. But Search is a better core than insurance: it throws off enormous operating income without the balance-sheet risk and regulatory complexity of financial engineering. And the subsidiaries—Waymo, DeepMind, Verily, and others—are science-and-technology bets with the potential for transformative returns, run by their founders inside a structure built to give them autonomy. The Berkshire comparison isn’t a stretch. It’s the stated model.
A number of LPs, investors I admire, and friends have asked me to expand on this idea. At first glance, there seems to be little in common between the two companies beyond the holdco structure: one is a portfolio of compounders and the other is a technology company. But the similarity in structure is just the tip of the iceberg. Alphabet was built, deliberately, on top of the foundations Berkshire laid—at least on the business side.
This memo traces how Alphabet inherited, then evolved, Berkshire’s model:
Lineage: Larry and Sergey meeting and learning from Warren over the years
Capital Structure: dual-class shares, retained earnings, patient capital
Obsessive Mindset: Buffett’s allocative obsession with investing; Page and Brin’s generative obsession with AI
Magic Money Machine: float on the balance sheet (time arbitrage) vs search/ads on the income statement (category creation and expansion)
Same Model, Different Substrates: the same holdco model run on two economies: Berkshire’s physical one (diminishing returns) and Alphabet’s digital one (increasing returns)
Curation and Creation: Berkshire buys established compounders (curates); Google builds, invests, acquires, and amplifies unfinished ones (creates)
Organizational Design and Reputational Moat: a personal moat that may not outlive Buffett vs an institutional one
Cultural Discipline: long-term thinking, fast kills, and three honest objections
Alphabet is Berkshire Hathaway 2.0: Alphabet is carrying on the spirit of Berkshire
These aren’t separate parallels—they’re components of a whole: a decades-long obsession supported by a structure built to protect and amplify it. The dual-class shares protect the long horizon; the money machine funds the obsession; the creation model turns it into product; the institutional moat staffs it. The substrate is the exception—not something the obsession built, but the ground it got to run on, where the math compounds instead of capping out.
The Lineage
In May 2000, right as the Internet bubble began to pop, Ron Conway hosted Warren Buffett as the guest of honor at an SV Angel afterparty. Buffett had hoped to introduce NetJets to the newly rich of Silicon Valley, but Conway was more interested in hosting him than in moving fractional jets. As guest of honor, Buffett gave a speech that essentially amounted to: “You are all going out of business.” The room was the who’s who of the Valley, including Netscape’s Marc Andreessen and Napster’s Shawn Fanning. Also in the room: a young Larry Page and Sergey Brin. Conway recalls them being starstruck by Fanning, who had just been on the covers of Businessweek, Newsweek, Life, and Fortune, and equally taken by Buffett, who stood by the door and shook every hand. Page and Brin went home, read the annual letters, and studied Berkshire.
A year before, KPCB and Sequoia had invested in Google and, as a term of the deal, insisted the founders bring in a “professional CEO”—or the firms could buy back their shares at cost. After a CEO tour that put them in front of Jeff Bezos, Andy Grove, and Steve Jobs, Larry and Sergey decided Larry would be CEO—much to the frustration of a board that thought they weren’t taking company-building seriously. A year after the party, Eric Schmidt joined Google as CEO.
Having lost the CEO role, the founders were concerned about control of their company. They feared an IPO would only exacerbate it. Shortly before their IPO, Schmidt flew Larry and Sergey to Omaha to meet Buffett. Buffett walked them through how dual-class voting could let them go public without giving up control of their company. Google adopted it, and IPO’d with Class A shares carrying one vote and Class B shares carrying ten—Larry and Sergey held 51% of the Class B, which gave them joint control despite owning less than 12% of total shares. Buffett later recalled that when the three of them came to see him, “they were more interested in talking about going public and the mechanics of it.” It would not be the last time their paths crossed.
Before Google’s IPO, Larry and Sergey wrote a shareholder letter titled “’An Owner’s Manual’ for Google’s Shareholders,” footnoting that much of it was inspired by Buffett’s essays and his own “An Owner’s Manual” to Berkshire shareholders. This letter was included in their S-1 filing. Buffett read an early copy and gave his approval, noting he’d be glad to sign it if he were their partner. He did not, however, buy the stock.
In 2011, seven years after the IPO, Page returned as CEO and set about restructuring the company. When they announced the Alphabet restructuring (2015), Schmidt traced the genesis to their pre-IPO visit to Omaha. In his first interview post-Alphabet restructuring, Page revealed that he and Sergey try to talk to Warren and Charlie as much as they can.
Then the loop came full circle. Berkshire initiated a position in Alphabet in the third quarter of 2025—one of Buffett’s last major moves before he stepped down as CEO at the end of the year. Under new CEO Greg Abel, championed by Ted Weschler, Berkshire more than tripled their stake in the first quarter of 2026, to ~$17 billion, making Alphabet a core holding.
Buffett had reflected in 2019 that Berkshire should have bought Google earlier, calling it a mistake. It took him another half-decade to act, but he made it one of his last big purchases as CEO. The same “too late” was said when he bought Apple in 2016 as a multi-hundred-billion-dollar company, which has become the biggest gainer in Berkshire’s history. Late, for Buffett, does not mean wrong.
The Capital Structure
Berkshire’s influence on Alphabet shows up most directly in the financial architecture of the company itself.
Start with the holdco structure. After Google restructured into Alphabet, Schmidt traced the genesis to a single meeting in Omaha, where he, Larry, and Sergey had visited Buffett and come away struck by the scalability of his model. He described the restructuring as an attempt to build “a holding company that looks like Berkshire Hathaway out of an existing large company.” But the two arrived at that shape from opposite directions: Buffett started with a failing textile mill, bolted an insurance company onto it, and spent six decades compounding the float into a conglomerate, whereas Alphabet started as one of the most valuable companies on earth and restructured itself, in a single stroke, around a search engine that was already throwing off cash.
The dual-class shares came from Buffett too. By Brin’s own account, Google’s A/B structure was patterned after Berkshire’s, and by one investor’s telling, Page and Brin only agreed to go public after Buffett walked them through how dual-class voting could let them sell shares without surrendering control. Google built this into their IPO with Class A shares that held one vote and Class B shares that held 10 votes. The founders held 51% of the B shares, which ensured they had joint control despite holding less than 12% of the equity. But there’s nuance here that goes largely unappreciated: Buffett had created Berkshire’s B shares for liquidity; Page and Brin adopted the mechanism to maintain control—and their model has proliferated across Silicon Valley.
But it’s not just in their structure—it’s also in their capital allocation communication.
Neither company offers quarterly guidance. In their S-1 shareholder letter, Page and Brin directly quote Buffett: “In Warren Buffett’s words, ‘We won’t ‘smooth’ quarterly or annual results: If earnings figures are lumpy when they reach headquarters, they will be lumpy when they reach you.’”
Neither company pays a meaningful dividend. Berkshire has paid one in exactly 60 years—a $0.10 distribution in 1967 that Buffett calls a “terrible mistake” to the point that he jokes he was in the men’s room and the directors voted while he was gone. Alphabet paid its first dividend in 2024—a symbolic $0.20 for 20 years as a public company. The logic is the same: earnings retained inside the company can be moved between businesses and redeployed without double taxation.
Both companies have historically hoarded cash and opportunistically deployed it. Berkshire held over $31bn before the GFC, then put $5bn into Goldman preferreds and $3bn into GE as the crisis hit; today it sits on ~$400bn. Google long carried ~$100bn on its balance sheet—a cash pile large enough that critics pointed to it as proof Google had stopped innovating. But that same pile is what let Alphabet acquire and invest without anxiety—and, now, commit $180-190bn of capex to AI.
Every one of these choices serves the same end: control. They are free from the whims of Mr. Market and can allocate capital as they see fit, optimized for the long-term—even (especially) if it means short-term pain.
Obsessive Mindset
Buffett, Page, and Brin are known for being monastically obsessed with their intellectual pursuits over multi-decade periods. By the time he was 11, Buffett had read every single book about investing in the Omaha Public Library, and purchased his first stock. As he has famously said, he “tap dances to work.” He wouldn’t have been able to invest for 80+ years otherwise. And it’s not just long-term investing—he’s chased returns everywhere from buying up Korean small-caps to trading options in his PA to making some of the world’s largest acquisitions.
Page has been obsessed with AI for as long as he can remember. When he was 12, he read Nikola Tesla’s autobiography and found himself crying and vowing that inventing wasn’t enough—he had to commercialize his inventions to ensure that they actually changed the world. His father was an AI professor. Many of his first hires were AI/ML PhDs. In his very first public interview in 2000, he said: “Artificial intelligence would be the ultimate version of Google.”
He has repeated this message in every single interview he has given to this day. Sergey has as well. They don’t give as many interviews as other CEOs, but listen to any interview with either and you will hear about AI. People cite Bezos as a visionary for always repeating “customer obsession,” but Page and Brin have talked about AI more and more consistently than Bezos on customer obsession.
Sergey, who went into “retirement,” actually returned to Google and has been working in office nearly daily with the AI team. This was prompted by an OpenAI engineer who asked him: “What are you doing? This is the greatest transformative moment in computer science ever, completely.”
If Berkshire was the vehicle for Buffett to invest, Google was the vehicle for Page and Brin to build AI. The logic was always the same: when they ran out of something, they built it. They ran out of data, so they made more—Gmail (2004), Maps (2005), YouTube (2006). They needed mobile distribution, so they bought Android (2005); somewhere to put it all, so they launched Cloud (2008); moonshots, so they built X (2010) and spun out Brain (2011). Then the AI stack: TPUs to run it (2013), DeepMind to push it (2014), and the Transformer paper (2017)—which, by the way, came from Brain, not DeepMind. Meena (2020) became LaMDA (2021), arguably the first genuinely conversational LLM, before ChatGPT (2022). Over 20 years, Google became the most vertically integrated company in AI.
And their obsession has paid dividends many times over. Demis Hassabis, after deciding to sell Deepmind, chose Google over Facebook, despite being offered less money, because he believed Larry truly saw and believed in AI whereas Mark Zuckerberg was just as excited about all new technologies (AR, VR, 3D printing) as he was about AI.
Buffett’s obsession is allocative: deploying capital across equities, options, micro-caps, treasuries, acquisitions. Larry’s is generative: building AI. It’s the same psychological structure with different intellectual content. In both cases a perpetual cash machine funds the obsession and lets them run it to its limit—though the two machines work in very different ways, which is the next section.
Oh, and don’t forget—Larry and Sergey are only 52 and 53. Buffett accumulated over 99% of his wealth after his 50th birthday, and over 95% after his 60th birthday. Larry and Sergey still have a long ways to go. And they have already passed the torch to Sundar to take on the political, public-facing role so that they can focus on what they’re interested in. For Larry, who knows. For Sergey, it’s back in the trenches coding and trying to understand and scale AI.
The Magic Money Machine
Both Berkshire and Google have a magic money machine—a source of structurally cheap capital that funds their compounding. Both have built fortress balance sheets: Berkshire regularly issues yen-denominated bonds (including 30-year tranches), and Google just issued a century bond. Neither “needs” the cash; both are happy to lock in permanent capital. But there’s a key difference between their machines: Berkshire’s lives on the balance sheet; Google’s on the income statement. Berkshire’s financial; Google’s operational. Berkshire’s arbitrage; Google’s creation.
Berkshire’s machine is insurance float: its insurers collect premiums today, pay claims years, sometimes decades, later, and invest the spread in between. It’s access to other people’s money at structurally low cost, but it borrows against future claims, which is why the discipline of underwriting can never relax. Google’s machine is search and ads, and it doesn’t borrow against anything; it creates surplus outright. Google Search and other advertising generated $60.4 billion in the first quarter of 2026 alone, up 19% YoY—revenue with no claims attached.
The scale shows up as margin, not just dollars. The tell isn’t the size of the surplus but how cleanly it falls through. Alphabet’s consolidated operating margin reached 36.1% in Q1 2026, on $39.7bn of operating income. And that’s the blended number, dragged down by the money-losing bets. The core Services business ran at 45%. Compare that to insurance, which is engineered to make almost nothing on underwriting and earns its keep on the float instead. Berkshire collects the spread; Google collects the spread and the margin. And the gap widens with scale. Insurance demands constant underwriting vigilance as it grows (every new premium is new risk to price) while Google’s margins expand as it gets bigger. The ponds are different too: Berkshire fishes in insurance while Google fishes in global commerce—an inherently larger market that is also growing faster.
Google’s money is free in a way float can never be: float has to stay liquid and claims-paying, so it can only fund marketable securities and cash-generative acquisitions. It can’t bankroll a decade of zero-revenue R&D (this is a key reason Buffett missed Amazon). Google’s ad money carries no such constraint. It funded DeepMind for a decade with no revenue. It built TPUs for internal use with nothing to sell. Gmail offered so much storage at launch people thought it was a joke.
Both machines compound because the surplus is reinvested, not distributed—but they reinvest in different things. Berkshire’s funds acquisitions of cash-generating businesses; Google’s funds the creation of new, often seemingly crazy ones through long-horizon R&D. (Recall that neither pays a meaningful dividend: retained earnings move between businesses and get redeployed without being taxed twice.)
The proof that Google’s surplus is generative is where it goes—straight back into the ground. Alphabet guided to $180–190bn of capex in 2026, roughly double the $90bn of 2025 and a ~6x jump in four years, funded substantially out of operating cash flow rather than debt. Cloud is already turning: operating income tripled to $6.6bn last quarter at a 32.9% margin, up from 17.8% a year earlier. Float, by contrast, has to sit in liquid securities waiting for claims.
One machine compounds by feeding itself; the other arbitrages a spread it can never fully spend. Both are magic, but they are structurally different. Google’s is the kind that can fund a 25-year bet on AI without ever asking permission from a claims department.
Same Model, Different Substrates
Different as they are, Berkshire and Alphabet both run on a magic money machine. Just as important, though, is the substrate each one operates on. Berkshire operates in the physical economy; Alphabet in the digital one. They’re two structurally different things. In the physical world, size and growth are bounded by physics. In the digital world, they aren’t.
This isn’t a novel idea—here’s Brian Arthur in 1996: “Modern economies have split into two interrelated worlds of business corresponding to the two types of returns. The two worlds have different economics…They call for different understandings.”
The core of his paper is the two types of returns: 1) The physical economy runs on diminishing returns—businesses hit limits in capacity, geography, market size, and settle toward equilibrium; 2) The digital economy runs on the opposite: with zero marginal cost and global reach, the one that gets ahead tends to get further ahead. Berkshire’s book (insurance, rail, energy, retail) sits almost entirely in the first world; Alphabet operates in the second. Same structure, different substrates. Those static properties permit a dynamic loop—and AI is accelerating it in real time.
The first version of the digital economy was the internet, which connected the entire world. All of a sudden, you could graph the human population. Each human was a node, and could be connected to every other node. They didn’t actually know each other, or even interact with each other—but they could. “Could” is the key word. And the value of a network grows with the connections between nodes, not the number of nodes, so each new person added value superlinearly. Connecting all humans was the first exponential, but it had a built-in ceiling: the number of humans.
AI agents leapfrog that ceiling. The internet was bounded by the number of humans who could do economically valuable work, which agents can now do too. So each new agent is a new node—and because agents can be spun up and transact with each other (and with humans) at machine speed, they add nodes and connections faster than humans ever could. The internet had bots, but they were narrow. Now the graph scales with the number of economically valuable tasks, not the number of workers—and there’s no human ceiling on how many of those there are.
Back to Arthur: increasing returns are the engine: more nodes -> more connections -> more value -> more nodes. The internet connected the world—AI expands it.
This is easy to miss, because humans don’t inherently understand exponentials (see the IPO memo; a full one on this soon). As a result, the growth physics gets systematically underestimated.
Berkshire’s world doesn’t work this way. A railroad gets no network effects from more track; insurance doesn’t compound on connections. This isn’t an execution gap—the loop simply isn’t available in that domain.
The irony, though, is that digital’s expansion is now paid for in the most physical terms imaginable: power, fabs, and land. This is why Alphabet’s capex makes it look more like Berkshire’s rail-and-energy book than asset-light software. And its frontier pushes the other way: Waymo and robotics are AI reaching out of the data center and into the physical world it supposedly transcended.
Curation and Creation
Tech observers like to say Google can’t innovate—Peter Thiel famously said it to Eric Schmidt’s face, pointing to Google’s cash pile as proof. The critique is meant to diminish Google, but it diminishes the critic instead, in three ways: 1) acquiring is hard, and it’s what Berkshire is celebrated for (”Google bought it” places Google in the Berkshire tradition, not beneath it), 2) scaling what you acquire is even harder (Google’s compute, data, and distribution turn acquisitions into things the original founders couldn’t have built alone), and 3) Google has innovated, continuously, for 25 years (Search, Ads, Gmail, TPUs, the Transformer paper that started the modern AI era). Berkshire does the first; Google does all three.
Acquiring is easy to recognize—Berkshire does it. And even critics of Google’s technology call them one of the best M&A teams in tech. But what they miss are 2 and 3: scaling and innovation.
Berkshire works the first two columns—Alphabet works all three.
The minority stakes column is the most Berkshire-like: patient capital allocation, little operational engagement. Google’s venture portfolio is similar in muscle to Berkshire’s public equities activities (Apple, Coca Cola, BoA)—just earlier.
The Acquired + Amplified column is where the separation begins. When Berkshire acquired GEICO, GEICO got Berkshire’s balance sheet and direct-response advertising scale and grew from a niche auto insurer into the second-largest in the US. When Berkshire acquired BNSF, BNSF got capex patience no public railroad could match. See’s got pricing-power discipline and a long leash on margin expansion. Buffett buys, then “delegates to abdication.” The holdco structure amplifies each company. However, there aren’t direct network effects. GEICO doesn’t make BNSF more valuable. Neither does See’s. Each subsidiary gets Berkshire’s capital and Buffett’s allocation discipline, but they don’t get each other.
Alphabet takes amplification a step further. Search required servers that scaled YouTube. YouTube’s videos trained VEO. Crawling for Search laid the foundation for LLM training data. AI improved Search. The pieces compound on one another, not just on the parent. Berkshire amplifies vertically; Alphabet amplifies horizontally. And the horizontal compounding is in service of one obsession: AI. Every component of the AI stack (data, compute, talent, algorithms) feeds and is fed by nearly every other Alphabet business.
This active curation isn’t unheard of: Henry Singleton’s Teledyne is the closest historical analog, (although Thorndike files it as a capital-allocation story when it’s really a creation-and-amplification one. A different memo). Elon is doing something similar with SpaceX. Another different memo.
That obsession explains the third column: Google has always worked on a lot of things internally (search, maps, mobile, ads, video, AI) and has no ego about how each one gets built. If it sees talented people or great technology further along on a problem it’s working on, it has no qualms acquiring them. They care about the end result—not credit for starting. And if no one is further along, Google builds it. Same psychology, different output depending on what the world has on offer.
Take AdSense: Google had launched its own content-targeted ads program in March 2003 and acquired Applied Semantics weeks later, adopting their more advanced technology and even the name. The combined product launched as AdSense in June. YouTube was a 20-month-old startup, but it was already the leader in video—Google Video was struggling to get traction. Android was basically just Andy Rubin exploring different ideas (similar to AI neolabs today). DeepMind was the most promising AI lab in the world; it eventually merged with Google Brain. Each acquisition was a thesis Google already had and was executing on.
Waymo was placed in “Acquired + Amplified” but it deserves a footnote of its own. The 510 Systems acquisition wasn’t a clean outside-in purchase. Anthony Levandowski incorporated 510 in 2007, reportedly a month after he joined Google as a SWE to work on Street View. Google was 510’s only customer for its first 18 months, buying its mapping camera systems through a manufacturing intermediary. When Discovery Channel asked Levandowski to build a self-driving pizza-delivery car in 2008, Google declined to be involved and asked that it be clearly branded a side project. Levandowski spun up a second entity, Anthony’s Robots, to do the work. Google then acquired both companies in 2011 to bring the autonomous-vehicle program fully in-house.
The pattern is hard to miss: Google was already running speculative, high-uncertainty work through external structures it could later absorb—the holdco logic, operating ad hoc, years before it was formalized. Google X (2010) gave that pattern a name and a place; Alphabet (2015) gave it an org chart. The shape was there in Google’s behavior before it was on the org chart.
And when no one is further along, Google builds. The “Created from zero” column is the approach Berkshire doesn’t run. They acquire businesses—they don’t build them from scratch. There’s a structural reason for this: float can’t fund a decade of zero-revenue R&D. Google’s ad money can. Search, AdWords, Gmail, Chrome, Cloud, Brain, the Transformer, TPUs—each was built because no one else was. Or could. The same no-ego posture that drives the acquisitions drives the creations: Google cares that something exists; they don’t care about who gets the credit.
Under one roof, the founders and board can direct funding into any one of these businesses, all underwritten by the magic money machine. The capital flexibility shows up in another contrast: margin of safety. Margin of safety is one of the three pillars of value investing as espoused by Buffett, and one of the central ones he took from his mentor Ben Graham. But to Buffett, margin of safety is a financial term: buy a dollar of value for sixty cents and any valuation errors is absorbed. He’s told people I know that valuation is both his diligence and his risk management. He credits Charlie for evolving his thinking from cigar butts to quality businesses, but still thinks of himself as 85% Graham and 15% Fisher.
Google invented a different mechanism. It’s less valuation-sensitive, but not insensitive (it offered less than Facebook for DeepMind). It can supercharge what it buys. Google’s compute, data, and distribution can 10x an acquisition. Its margin of safety lives in the value it can add, which is more durable than a financial cushion because it doesn’t depend on Mr. Market being moody.
The standard rebuttal is that Google can’t ship: the Google Graveyard lists 280+ killed products. But that’s discipline, not failure. If Google doesn’t see the potential, it kills the project. Buffett did the same thing: he shut Berkshire’s textile mill, refused to chase 1990s tech, and dumped the airlines in 2020. The best holding period is forever, but that doesn’t oblige you to hold a mistake forever.
And on its core Search, Google has shipped continuously for 25 years—the only company that has. PageRank -> personalized results -> Universal Search -> Knowledge Graph -> BERT -> MUM -> AI Overviews -> AI Mode, with many steps in between. Every major shift in how people find information online for a quarter century was Google’s.
Another common critique is that Google hasn’t produced any outstanding founders (aside from Colin Huang (Pinduoduo) and Kevin Systrom (Instagram) who they call “one-offs”). This critique misses the point. Set aside that most of Google’s early employees stayed for years past the IPO: Susan Wojcicki, Salar Kamangar, Jeff Dean. The real evidence is the AI diaspora: the authors of the Transformer paper alone seeded a generation of multibillion-dollar labs; Dario Amodei went from Google Brain to OpenAI to Anthropic; Ilya Sutskever went from Brain to OpenAI to SSI.
Critics judging Google by its dead products instead of its living founders are measuring the wrong inputs. Experimentation is where the successes come from.
Three modes, one circle of competence. Berkshire works the first two; Alphabet works all three. The third is where the playbook diverges—and where the next quarter-century of compounding lives.
Organizational Design and the Reputational Moat
As holdcos, Berkshire and Alphabet share a structural spine: both decentralize operations and centralize capital allocation. But the founders differ in their proximity to control: Buffett ran Berkshire as CEO until he was 95, whereas Page and Brin gave up their operating titles years ago, choosing to operate as individual contributors and steer the company as owners and board members. Both are “homes of choice” with institutional depth that survives any single departure—Berkshire for family-business sellers and Google for technical talent. And both often win deals on identity over than price: DeepMind picked Google because Page understood AI; YouTube because Google understood networks and scaling.
But the moats are sourced differently, and that difference is this section’s argument.
Start with capital allocation. At Berkshire, the operating companies don’t keep their own earnings. GEICO’s underwriting profit, BNSF’s freight cash, and the utilities’ returns all flow up to Omaha, where Buffett decides how to allocate them. The subsidiaries run themselves operationally, but they don’t control their cash. Alphabet runs the same play, but on a different surface. Search, YouTube, and the ad stack throw off the cash; the company moves it (internally, without asking the market) into Cloud’s buildout, DeepMind, moonshots, and now $180–190 billion of AI capex.
In both cases the businesses are decentralized but the capital is not: one desk decides where it goes. That single fact is what makes the money machine a machine and the creation model possible—surplus generated in one place, redeployed into another, with no dividend, no tax, and no banker in between. But there’s another core difference in how their machines are powered.
Berkshire’s deal flow runs through Buffett’s personal reputation, which he has cultivated deliberately and guards above all else. He told Salomon employees: “Lose money for the firm and I will be understanding. Lose a shred of reputation and I will be ruthless.” That reputation is also why he runs, by choice, well below full capacity—he needs cash and credibility to be the backstop of the American economy. It allowed him too rescue Salomon in 1991 and Goldman in 2008. If he was always fully invested, he wouldn’t be Buffett.
Google’s moat is institutional. It doesn’t rely on Larry and Sergey (who are largely absent from public view), or even Sundar. Search kicked off the product flywheel, which has generated its own deal flow ever since. People use Search daily, which gives them brand awareness, and the magic money machine allows Google to pay up for talent, give them autonomy, and focus on the work. Developers build on Android because it has distribution; researchers join because they have compute and data. The reputation is the company’s, not the founders’—and it’s sourced from operational dominance, not founder charisma.
This institutional depth is where the gap is widest. Buffett prizes businesses “so good a ham sandwich can run them,” but he’s still the scarce input at the top of Berkshire. Google has engineered away that single-point dependence in its core research.
Any star joining or leaving OpenAI, Anthropic, or Meta is front-page news, because those labs’ research capacity is concentrated enough that a top 10 departure moves the needle. Gemini itself has over 3,000 contributing authors; a competitor poaching a hundred would barely register. DeepMind’s bench runs to hundreds of senior researchers, any of whom could lead elsewhere—and many have left to found or run the very labs now competing with Google. That reads as leakage until you notice Gemini keeps shipping anyway: the leavers took their talent but not the compute, the data, or the distribution, and the bench refilled behind them. It’s a sign of depth, not fragility. AlphaGo, AlphaFold, AlphaZero, AlphaStar, Gemini, the robotics work—these are outputs of a culture, not of any individual who could be hired away. Berkshire’s reputational moat is a person; Google’s is a system. Systems are harder to poach.
And this isn’t hypothetical anymore. Berkshire just ran the experiment. And notice it was never a company of one. It was a duo: Buffett and Munger. Just like Page and Brin. Munger died in 2023; Buffett handed the CEO role to Greg Abel at the end of 2025 and stayed on as chairman. Notice that Berkshire didn’t distribute Buffett’s job—it found one person to hold all of it. Abel runs operations and has final authority over capital allocation (the whole concentrated role) because the model requires a single allocator at the top. The only real question was who. The dependence didn’t disappear; it changed hands.
Alphabet did things differently: Sundar Pichai was named Google’s CEO in 2015 and Alphabet’s CEO in 2019. He took on the operating and public-facing role. The research core was built to run without a founder at the helm, and Larry and Sergey stepped back to steer as owners. Berkshire’s succession plan was to fill the seat; Alphabet’s was to make sure the seat was never a single point of failure in the first place.
There’s an apparent tension here: the obsession argument says Google compounds because Page and Brin never stopped caring about AI. This section says the moat is institutional, not personal—that the founders can leave the stage and the flywheel keeps turning. Both can’t be the engine.
In reality, the two claims are the same, albeit at different points in time. The founders’ obsession isn’t something shared every morning. It was encoded into the institution years ago, and now runs without them. It’s why the company hired AI and ML PhDs among its first employees, why it tolerated a decade of unprofitable DeepMind, why it built TPUs no one would buy, why thousands of researchers now work on Gemini. None of these were daily decisions by Page, but all of them are the predictable output of an organization built by someone for whom AI was the point.
The payoff: remove Larry and Sergey today and the thing their obsession built is still standing—the core components (compute, data, talent, distribution, the model stack) and the network effects binding them, each piece making every other more valuable. The compounding is structural now; it doesn’t route through a person. It’s the thing Jobs said he was proudest of: not any single product, but the company that could keep making them after him. They didn’t just build AI, they built the machine that builds AI. They could step back because the obsession no longer needed their presence to express itself.
Buffett never wanted that distance. He remained Berkshire’s central allocator until he was 95 because allocating is the thing he loves, not a burden to delegate. Page and Brin wanted the opposite: to engineer away the dependence in the research core that mattered most to them, and step back at 52 and 53 with the engine still running. Same structural move, opposite motivations. One stayed in the seat by choice, the other built so the seat could be left.
Cultural Discipline
Berkshire and Alphabet share one cultural habit above all: long-term thinking as institutional reflex. Both also admit failure in public: Buffett on missing Amazon and Google; Google on Plus, Stadia, and Reader. But the more distinctive discipline, the one that separates Google even from Berkshire, is the speed of the kill.
Ask most fundamental investors about sell discipline and you’ll hear something about “conviction” or a “broken thesis”—language that is almost always more wishy-washy than concrete. Google’s kill velocity is the opposite. The product graveyard isn’t evidence of a company that can’t finish; it’s evidence of one that refuses to fund what it no longer believes in. Microsoft kept Bing alive for over 15 years; Google wouldn’t have. “Page’s Law,” Larry’s observation that software tends to get twice as slow every 18 months (which Sergey named and vowed to break), was taken seriously, and treated as a defect to engineer against. For Google, speed is the discipline—on both the way in and on the way out.
That same discipline is what makes the long-horizon bets possible and not reckless. Cutting the losers fast is precisely what earns the right to hold the winners for decades: a company that isn’t bleeding capital into dead projects can afford both to start often and to wait long. Starting is cheap when stopping is honest. Berkshire’s version of the same instinct is slower and quieter (Buffett simply declines to swing) but the underlying habit is identical: protect the compounding by refusing to let dead capital ride.
Three concerns are worth stating plainly, because an honest version of this thesis has to hold them:
Regulatory exposure: Berkshire has faced little antitrust pressure because its businesses aren’t related—GEICO doesn’t directly reinforce BNSF. Google’s do. The same connectivity that accelerates the flywheel on the way up could accelerate it on the way down. Google draws antitrust concerns precisely because their model works. So far, any attempts to break them up have been denied.
Succession: Most observers read Sundar as too passive. But Larry and Sergey are 52 and 53, far more involved than people think, and as owners—if they stay healthy—have another 30+ years of compounding ahead. Even then, they picked Sundar as heir and work closely with him, and the bench beneath is deep: Demis Hassabis at DeepMind, Thomas Kurian at Cloud, Assaf Rappaport from the Wiz acquisition.
AI as disruption: People think AI could be Google’s downfall. A few years ago the fear was that chatbots would disrupt search; now it’s migrated to cost: Google will bankrupt itself chasing AI. The capex is staggering ($180-$190bn in 2026), and no one yet knows whether it’ll earn its return. But if anyone can afford to be wrong about the size of the bet, it’s Google: the spend comes out of operating cash flow, and the same compute monetizes across Search, Ads, Cloud, and YouTube. Because the deeper truth is the one the fear keeps missing: Google has been building towards AI for 25 years. Search is basically AI, and Google enters best positioned on compute, data, and talent. Search was the precursor to AI—and it may even be its successor.
Berkshire hasn’t faced regulatory exposure—and may not. But Buffett has just stepped down, and AI is permeating even Berkshire’s portfolio via multiple sectors (insurance, energy, etc.)
Alphabet is Berkshire Hathaway 2.0
When Page announced Alphabet, he slipped in a financial pun: “We also like that it means alpha‑bet (Alpha is investment return above benchmark), which we strive for!”
Alphabet is Berkshire 2.0. It was restructured, at least twice, and deliberately, in Berkshire’s image. The founders built on the principles they learned from Buffett and Munger, improved them, and made them their own. I’m not saying it’s necessarily better—but I am saying it’s an evolution.
Page and Brin are obsessed with AI the way Buffett is obsessed with investing. They think in decades, and they’ve built a system that compounds as it scales. And the deepest divergence wasn’t a choice they made against Berkshire’s model—it was there before the model was. Search was the crown jewel from the start, and to them, search was AI. They didn’t pick a better engine than float; they already had a different one. Then they built the Berkshire architecture around it.
The whole structure routes back to obsession and long-term thinking: 1) The dual-class shares protect the long-term, 2) The money machine funds it, 3) The creation model builds on it, and 4) The reputational moat staffs it. Long-term thinking isn’t one feature among many—it’s the keystone. Pull it out and the rest is just plumbing.
In May 2000, Larry and Sergey met Buffett for the first time. Three years later they restructured around an IPO and wrote a letter in homage to his company. A decade after that, they reorganized into a holding company in his company’s image. A decade after that, Berkshire bought the company Buffett had inspired. The influence ran in one direction for 25 years—then reversed. Along the way, Google learned Berkshire’s architecture, wrote its own philosophy on top of it, and ran it on a different operating system.
Berkshire compounded for 60 years in the physical economy. Alphabet was born in the digital one—an economy where the math doesn’t stop. And now, through Waymo and robotics, it’s carrying that digital compounding back into the physical world that Berkshire never left. Google didn’t just inherit Berkshire’s model for another substrate—it outgrew the substrate itself.
But the constant in both companies (without which the architecture is just plumbing and the operating system is just code) is the obsessive, decades-long, single-pointed commitment of the people running them. Berkshire compounded because Buffett never got bored; Google is compounding because Larry and Sergey haven’t either.
There will never be another Warren Buffett, and there will never be another Berkshire Hathaway. But Larry, Sergey, and Alphabet are carrying their spirit forward.
If you’ve made it this far, thank you for reading. If the subject of this memo is something you’ve been thinking about, I’d love to hear from you (email; twitter).
Below the paywall: ~5,000 words of initial thoughts on the raise itself. The internet has spent the week dunking on Greg Abel for “YOLOing the farm” right after Buffett stepped down (first by increasing their equity stake in Alphabet by so much, now by anchoring this raise).
The addendum gets into why the raise isn’t what the headline says it is, speculates on why Alphabet is issuing equity rather than debt, the underlying benefits that Berkshire and Alphabet are each likely to get out of the deal, and of course, the one question the whole thing still can’t answer.