I almost didn't open Berkshire's 13F this weekend. Buffett is gone from the day-to-day, Greg Abel runs the show now, and 13F season usually just confirms what everyone already assumed.

Then I actually looked at the Alphabet number.

Berkshire's stake grew 83% in a single quarter to $37.9 billion, and buried in the filing is something even bigger, Abel personally wrote Alphabet a $10 billion check as part of its own capital raise back in June. That's not a stock purchase. That's Berkshire directly funding Google's AI buildout.

And then I found the part of this story nobody's connecting to it yet.

That's what One Big Idea digs into today: what Berkshire is seeing in Alphabet right now, and the complication that showed up in the news cycle at almost the exact same moment.

Missed Friday's Playbook on Apple's guidance miss?

Reddit’s stock jumped after it was selected to join the S&P 500, replacing AvalonBay before trading opens on August 18. Reddit still has powerful growth, with latest-quarter revenue up 61% to $805 million and net income nearly tripling to $253 million. The risk is concentration: around 95% of revenue still comes from advertising, though AI data licensing with customers like Google and OpenAI gives Reddit a small but potentially valuable second growth lane.

Plug Power’s Q2 showed real operating progress after years of painful cash burn. The company beat revenue expectations, raised full-year guidance, improved gross margins toward breakeven, and cut cash usage, helped by stronger material-handling deployments, service revenue, and Project Quantum Leap cost reductions. The stock is still speculative, but the rally makes sense if investors believe Plug is finally turning restructuring into better unit economics.

Anthropic reportedly told prospective investors that Q2 revenue jumped more than 14-fold year over year to over $11.5 billion, while Reuters reported the company is being valued against extremely aggressive future revenue forecasts. The bull case is obvious: Claude demand is exploding, enterprise adoption is strong, and investors see Anthropic as one of the few serious OpenAI rivals. The risk is just as obvious: massive compute costs and a possible $2 trillion-plus valuation mean Anthropic has almost no room for disappointment.

There are over 3,000 of them in The Shift’s vault, already vetted, ready to explore.

Subscribe and get instant access to the tool vault, a 1000+ prompt library, and free AI courses built for people with actual work to do. 

Plus, the daily newsletter that keeps you sharp on everything moving in AI in under 5 minutes a day.

They’re also giving away a free 1-year Claude Pro subscription to 3 subscribers. 

Subscribe for free to enter. All free. All in one place.

Subscribe to The Shift

Berkshire Hathaway's Q2 13F, filed this past Friday, showed its Alphabet stake jumped 83% in a single quarter to $37.9 billion, now the third-largest U.S. equity position in the entire portfolio. Buried deeper in the filing is something most headlines skipped past, Greg Abel personally committed $10 billion to Alphabet's own capital raise back in June, split evenly between Class A and Class C shares.

The obvious read is simple. Berkshire's new CEO just made the boldest tech bet in the firm's history, and if Abel is willing to write a check that size directly into Alphabet's own AI buildout, that's about as strong a vote of confidence as public markets get.

I started digging into why, and the numbers back it up more than I expected. Alphabet's cloud business is compounding faster than almost anything else in big tech right now.

But then I found the part of this story that complicates the clean version, something that happened in Alphabet's own AI division at almost the exact same time Berkshire was writing that check.

Here's the first thing that stood out once I pulled the earnings apart. Alphabet's Q2 revenue hit $119.8 billion, up 24% year over year, beating estimates of $116.93 billion.

Google Cloud is the real story inside that number. Cloud revenue jumped 82% to $24.8 billion, a dramatic acceleration from the 48% growth the segment posted just two quarters earlier, and cloud operating income more than tripled to $8.8 billion from $2.8 billion a year ago.

Then there's the AI adoption data, which is the part that actually surprised me. Nearly 90% of the Fortune 100 now use Gemini Enterprise, and the Gemini App has grown to 950 million monthly active users, processing 22 billion API tokens per minute, up from 16 billion just last quarter.

Abel isn't betting on a story anymore. He's betting on a business that's already converting AI hype into an $8.8 billion quarterly profit engine most of Wall Street still talks about like it's speculative.

Here's where the story stopped being simple. Right around the same window Berkshire was disclosing this position, Alphabet's AI division lost Jeff Dean, its chief scientist and one of the actual architects behind the technical infrastructure that made this quarter's numbers possible.

Dean spent 27 years at Google, and he's not leaving quietly. He's co-founding a new venture called Discovery Loop alongside three other senior AI figures, Sanjay Ghemawat, a longtime Google senior fellow, Oriol Vinyals, a DeepMind vice president, and Quoc Le, one of the original co-founders of Google Brain.

That's not a single departure. That's four of the people most responsible for Google's underlying AI research capability walking out the door at once, and the stock fell about 4% specifically on that news.

Google restructured quickly, moving Demis Hassabis into a chair role at DeepMind and giving Koray Kavukcuoglu operational control, separating long-term research from day-to-day Gemini development. Whether that split makes the company more efficient or just formalizes a talent gap is exactly the question Berkshire's $37 billion is now riding on.

Alphabet has been one of the more volatile mega-cap stories of the year, and this week added another chapter to that pattern. The stock hit an all-time high of $404.47 back on May 18, then spent the summer grinding lower before earnings in late July added another leg down on capex guidance concerns.

By this week it was trading around $344.70, roughly 15% below that May peak, and the Jeff Dean news knocked another 4% off in a single session on top of that. That's a stock that's had to absorb several separate, unrelated pieces of bad news in a short window, a capex scare and a leadership shakeup, all while the actual business kept accelerating underneath it.

The level worth watching is $330. That's roughly where the stock found buyers during the post-earnings dip in July, and whether it holds there again will tell you whether the market is treating Dean's exit as a real dent in the AI story or just another headline for a stock that's already digested worse this year.

Koray Kavukcuoglu has operational control of Gemini development, but the market hasn't had a real chance to judge his output yet. The next major model release or research announcement out of Google will be the first genuine test of whether the post-Dean structure actually works.

Abel just made his boldest move yet, and the next 13F, due in November, will show whether this was a one-time conviction bet or the start of a real position build. A further increase would tell you Berkshire isn't worried about the leadership churn at all.

Dean's new venture is framed as a public benefit corporation focused on scientific and engineering discovery, not necessarily a direct Gemini competitor. Watch for its first real announcements, since whether it becomes a genuine AI lab rival or something more academic changes how much this departure actually costs Google competitively.

Cloud's acceleration from 48% to 82% growth in two quarters is the single number underneath Berkshire's entire thesis. The next print needs to show that pace is durable, not a peak, for the "cash machine, not a story" read to keep holding up.

Alphabet is already guiding to $195 to $205 billion in 2026 capex, funded partly by the same capital raise Berkshire participated in. Any further increase to that number, on top of a leadership shakeup, is the combination that would actually worry me more than either issue alone.

I went into this weekend expecting a routine 13F, and I came out of it more interested in Alphabet than I've been in months, for reasons that have almost nothing to do with the stock price.

Greg Abel didn't just buy Alphabet. He personally wrote a $10 billion check directly into the company's own AI capital raise, which is a very different kind of conviction than clicking buy on a public exchange. I agreed with Buffett's original, much smaller $4.3 billion Google bet back in November, and Abel just showed he's willing to go dramatically further on the same thesis.

The Jeff Dean departure is real and I'm not waving it away, four senior AI leaders walking out the same door in the same week is not nothing. But Berkshire isn't betting on any single scientist, it's betting on an $8.8 billion quarterly cloud profit engine, 950 million Gemini users, and a company that's proven it can keep compounding even while absorbing bad headlines.

I think Wall Street is currently more spooked by the departure than the numbers justify, and Abel's check is the clearest signal yet that somebody with a much better view of the business than any of us doesn't think this changes the thesis at all.

Most coverage tells you what happened. Fintech Takes is the free newsletter that tells you why it matters. Each week, I break down the trends, deals, and regulatory shifts shaping the industry — minus the spin. Clear analysis, smart context, and a little humor so you actually enjoy reading it. Subscribe free.

Sign Up Free

The best investors rarely react to the headline that's loudest. They react to the one that actually changes the math.

Berkshire's $10 billion check and Jeff Dean's exit happened almost on top of each other this month, and only one of them moves Alphabet's actual earnings power.

Capital commitments from people with real information are a stronger signal than personnel news, even dramatic personnel news, because money is a harder thing to fake than a press release.

That doesn't mean talent departures never matter. It means the size of the reaction should usually match the size of the actual financial impact, and this week those two things came apart.

The best readers here were referred by readers like you. Pass it on.