Berkshire Just Put Another $10 Billion Into Alphabet. That's the Signal.
- Rich Washburn
- 4 hours ago
- 6 min read
Berkshire Hathaway's first Alphabet position caught my attention for a reason that had very little to do with Google stock. The second move made the original argument considerably harder to ignore.
Berkshire Hathaway just did something it hadn't done in a while.
After 14 consecutive quarters as a net seller of equities, Berkshire went back into the market in size. It bought roughly $20 billion more stock than it sold during the second quarter. And sitting right in the middle of that deployment was another $10 billion into Alphabet. That's interesting on its own.
Warren Buffett is no longer Berkshire's CEO. Greg Abel is. Berkshire is still sitting on an absurd amount of cash, so this wasn't a case of needing to put money somewhere. They had options. They chose Google.
More specifically, they chose to participate directly in Alphabet's capital raise as the company pours enormous amounts of money into AI infrastructure and global compute.
That's the part that got my attention. Because we've been here before.
Where This Started
When Berkshire's original roughly $4.3 billion Alphabet position became public in November 2025, most of the coverage understandably treated it as a Buffett-and-tech story.
I didn't really see it that way.
I wrote "Warren Buffett Just Bet on Google — And That's Bigger Than It Looks."
My argument was that Alphabet had crossed an important line.
Google wasn't merely a technology company anymore.
It was becoming infrastructure.
Search. Cloud. YouTube. Workspace. Android. DeepMind. Gemini. Data centers. Advertising. Distribution.
The individual products mattered less than the system underneath them.
I wrote:
"Google isn't the cool app on your phone — it's the plumbing beneath the modern internet."
And then:
"And Buffett knows plumbing."
That was the thesis.
Not *Google will outperform next quarter.*
Not *Gemini wins the model war.*
Something much simpler:
Alphabet was becoming the kind of durable economic infrastructure Berkshire has historically understood extremely well.
That same day, I pushed the argument one step further in "There Is No AI Bubble — Just a Delusion Bubble."
The point wasn't that everything carrying an AI valuation was sane.
Clearly it wasn't then, and it isn't now.
The distinction was between the speculative carnival happening on top of AI and the infrastructure being constructed underneath it.
I wrote that Alphabet wasn't simply a tech company anymore.
It was the grid.
And I ended with:
"That's not a bubble. That's the new baseline."
Three days later came "The Confirmation Effect."
That article was less about Alphabet and more about where I thought people were looking.
Everyone was watching model releases, demos, chatbots and benchmarks.
I was more interested in capital allocation, infrastructure commitments, GPUs, compute and what the people writing very large checks appeared to believe.
And now here we are.
The Second Check Matters More Than the First
The first Berkshire purchase was interesting. The second one is different.
Berkshire has changed CEOs.
The AI conversation has changed dramatically.
Valuations have moved.
Capital requirements have exploded.
Debate about an AI bubble has gotten louder, not quieter.
And Berkshire just committed another $10 billion to Alphabet while Alphabet is raising capital to build more AI infrastructure. That doesn't prove every argument I made. It doesn't make Berkshire infallible. It certainly doesn't mean everyone should run out and buy Google. But it does tell us something about how one of the world's most patient capital allocators is looking at the landscape.
They don't appear to be betting on a chatbot. They appear to be betting on the system underneath it. That distinction matters.
..And This Is Where It Gets a Little Strange
The Berkshire story also reminded me of something I've noticed in my own work. Not predictions exactly. More like recurring instances where something looked interesting at one layer, and the more useful story turned out to be underneath it. I've started keeping those because the timestamps make them useful. Not as trophies. As data.
The $5 AI Keyboard
In September 2024, I built a ridiculous little $5 ESP32 macro keyboard with dedicated AI actions. The thing itself wasn't important.
The question underneath it was: if AI becomes persistent, why should we keep navigating software to reach it?
Two years later, dedicated physical controls for agentic AI are becoming an actual product category.
The Navy Demo
In October 2024, I gave the Navy League of Fort Lauderdale a live demonstration built around onboard AI, predictive maintenance, autonomous maritime systems, local intelligence and the idea of treating nuclear-powered vessels as sovereign compute platforms.
Six months later, Frank Bell called and told me pieces of that conversation were showing up inside Navy strategic discussions.
Since then, public Navy programs have appeared around predictive machinery health, operational autonomous vessels, distributed intelligence and AI-assisted shipbuilding.
The complete architecture I described still hasn't appeared publicly. That's important. But enough of the individual pieces have that I eventually went back and scored the original demo against what has happened since.
Wall Street and the Ground
I started paying attention to banks building dedicated teams around AI infrastructure.
My thought was pretty mundane: Banks hire ahead of deal flow.
Then the conversation moved from GPUs into power, interconnects, energized land and the hard physical constraints underneath AI.
That became "The Canary Is Dead" and later "The Gravity Is in the Ground."
Wall Street has since started openly talking about an AI infrastructure capex supercycle, and power availability has become one of the defining constraints on data-center development.
I could keep going. But that would turn this into exactly the kind of article I don't want it to become.
I Don't Think the Interesting Part Is Being "Right"
Because that's too easy to abuse. Write enough things and eventually you can rummage through the archive, find the winners and construct a mythology around yourself.
No thanks.
I've been wrong...I've chased ideas that went nowhere... I've built things that were much more interesting in my head than they were once electricity got involved.
That's part of the process. What interests me is something slightly different.
The things above don't really share an industry. Naval systems...AI interfaces...Capital markets...Data centers...Energy...Alphabet.
They look unrelated until you stop organizing them by sector and start organizing them by systems. Then they start looking suspiciously similar.
The Headline Is Usually Late
I think that's the larger pattern. We naturally focus on whatever is visible.
The product.
The model.
The stock move.
The announcement.
The Navy contract.
The new device.
But by the time the headline arrives, a lot of the meaningful movement has already happened.
Someone hired the team.
Someone secured the power.
Someone signed the contract.
Someone allocated the capital.
Someone hit the constraint.
Someone built the ugly prototype.
Someone realized an old interface was no longer necessary.
The visible event is often the last step in a much longer dependency chain. And those dependency chains leave evidence.
That may be why some things look like predictions later when they didn't feel like predictions at the time. You're not necessarily seeing the future.
Sometimes you're just seeing a system that has already started moving before everybody agrees on what to call the movement.
That's what Berkshire's latest Alphabet investment brought back into focus for me.
I looked at Berkshire's first Google purchase and thought: This doesn't look like a bet on technology.
It looks like a bet on infrastructure.
Now Berkshire has changed CEOs, ended a 14-quarter run of net equity selling and put another $10 billion directly into Alphabet while Alphabet finances the next expansion of its AI infrastructure.
I don't know what Berkshire does next. I don't know who ultimately wins the model wars. And I definitely don't know what the AI economy looks like five years from now.
But the plumbing thesis? That one is holding up remarkably well. And apparently Greg Abel knows plumbing too.
Sources and Receipts
Berkshire / Alphabet
Hardware / Interface
Navy
Infrastructure / Capital
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.
Comments