In addition, Alphabet has reached an agreement to sell $10 billion of stock to Berkshire Hathaway Inc. in a private placement, comprised of $5 billion in Class A Common Stock at a price of $351.81 per share and $5 billion in Class C Capital Stock at a price of $348.20 per share.
This investment by Berkshire Hathaway adds to the position it has built since Q3 2025.
Recently I have been asking YouTube's new AI about some videos ("when is Steam metrics mentioned in the video?" for example), which means they also index videos. This is an unthinkable amount of data.
I'm actually impressed at how bad Alphabet is with LLMs since they invented the thing as we know AND have all the data to train on, yet OpenAI and Anthropic are eating their pie.
Google's main revenue is ads based on search. LLMs are a competitor to search. Creating better LLMs will cut into search volumes.
In any large organisation this is extraordinarily difficult to manage - they have to incentivise the new tech that is actively harming the current revenues, while maintaining as much of the old revenues as possible, without creating internal conflict between these two parts of the organisation that will kill it.
Though in fairness to Google they do seem to realise this and are trying to adapt - they're letting the LLM folks mess with search. It'll be interesting to see how this goes.
I'm actually impressed by how much the Hackernews crowd is sleeping on Google & Gemini. Yes, it's lagging behind in coding, but it's consistently much better and more reliable at literally everything else.
Also there was a period of time when Gemini was the best model out there...
It's actually incredibly useful if you just want to summarize a video, or my use case, want a text tutorial of something that's a video.
I'm still on Anthropic models to code but I'm on Gemini 3.5 Flash for everything else. How can you say Google is bad at LLM when their little flash model is literally SOTA on many benchmarks?
> ... yet OpenAI and Anthropic are eating their pie.
They're eating nobody's pie: it's a new pie. Google is a $4.5 trillion company, the 2nd biggest in the world as I type this.
Seen that fact and seen how good Gemini 3.5 Flash is, I'm not really sure Google is "bad at LLMs".
Google knows LLMs are the new UI, not the new IDE.
Doing a little bit of RAG on the transcript hardly sounds impressive.
Not my impression. Lately I think Gemini is superior to ChatGPT and Claude in coding (I'm mostly using it with scientific stuff in Python).
And they have a massive amounts of TPUs. And yet... their models are way behind.
Auto Dubbing on the other hand is incredible, translating Russian/Ukranian speech with different voices and accents for each speaker, during a fire fight is wild.
I know GAAP accounting won't recognize any capital gain on these treasury operations, but from an economic standpoint this financial judo creates a lot of value for existing shareholders.
I don't know who's going to win the llm battle, but googles finance team has been doing their job fantastically.
Tech firms should always have a buffer and never get too close to the optimal debt ratio.
I think they have learned a lot re. what happens if you are asleep at the wheel now.
This is an interesting change. Essentially just gives more timing control?
Not nitpicking your answer, I just don't understand.
Capital raising is best done when markets are favorable, and Alphabet has the ability to choose how and when to raise.
Recall the froth of follow-on offerings hot circa 2000
So being down 1.7% is literally exactly what you'd expect.
It's insightful to put such documents into Claude and see how they use many different financial mechanisms to raise the money. $15B sold directly to the big banks, $40B sold to the market (but also facilitated by these banks), a direct investment (PIPE) from Berkshire. Pretty cool how financial markets do these things.
Stock based comp is another $350B a year in US markets alone. So if you think about public markets as an avenue for companies to raise capital, post-IPO firms are doing it to the tune of more than half a trillion a year.
I would have thought we have over built Datacenter before even AI came. There are enough Datacenter Rack space that replacing 5 - 8 years old server to newer 256 Core CPU would have increased their CPU per Rack by factor of 4 - 5. Saving significant space for future growth.
Instead we are so behind in Rackspace we are now building out Datacenter faster than ever.
More than a quadrillion high quality tokens per year. Pretty soon they will have an automated team of scientists doing basic and advanced research in every field. All those tokens will be fed back and make the model much more inference efficient.
All these big tech firms are spending wildly to make sure they are the one on top at the end of it all. But whoever that ends up being there’s going to be one hell of a lot of fallout underneath them.
At least part of this is slated for employee stock comp. Could be to keep their talent from running.
Why do you think there will only be one winner?
"Alphabet announced that its 2026 capital expenditures are expected to be $180-$190 billion, and that it expects 2027 capital expenditures to significantly increase [...] over the 12 months ended March 31, 2026, Alphabet generated $174 billion of operating cash flow"
Like how the early railroads or oil companies shook out and cost more than expected.
At least not yet.
There's not that much cash sitting around.
Something is gonna need to get sold to transfer into those assets.
Unless central banks are just going to print money to invest in these companies, I don't know who else is going to be able to take on enough debt to prevent massive sell offs somewhere for this.
It's not like ~$400B is pocket change...
(2) Middle east oil money (Saudi Aramco's profit every year is $100B+)
(3) Public traders have been and are looking to cycle out of other investments into higher growth areas.
It's easy as fuck for Google to raise this money because they are a money printing business. They are the most profitable company in the world, so for anyone this is basically the same as buying US debt.
https://www.sec.gov/ix?doc=/Archives/edgar/data/0001652044/0...
I guess they don't want to burn it down to $40B?
High cap companies use debt for this: bank loan is located in the market where it's needed most, and the debt is serviced by interest earned from securities in other markets. The net taxes are a small percent (think 3%) relative to simply transferring funds within the company. Yes, this is the low effective tax rate the EU is quite upset about.
Other reasons for not touching their holdings usually have a similar explanation. The securities are fungible for accounting purposes but not fungible enough for actual day-to-day operations. Result: securities get "stranded" and the large company grows a hedge fund appendage.
The market thinks Alphabet is most able to efficiently turn $80B into more money by investing in AI infrastructure.
So, Alphabet is happy to oblige them, given the favorable terms.
Literally nobody.
Every company from megacorps to small fish are spending well in excess of profits on these capex expansions. No ROI timelines yet established....
Even if Alphabet has $80B sitting in the bank, they could quite reasonably arrive at a comparable decision.
How recently have you looked? I think nowadays it's quite good.
I'm pretty sure they do. They already index metadata (you can see it in the web search results) so indexing the transcript is relatively easy.
I have a boss who loves to rattle on for ages, and it gives a breakdown of what on earth he was on about
cries in Google Glass
Wild that Meta has that product now decades later, which isn't even half of what Google offered.
Yes. Their competition is deploying debt and Google has low leverage. They also have $100+ billion cash on their balance sheet.
> Tech firms should always have a buffer and never get too close to the optimal debt ratio
...why is this especially applicable to tech firms? (Or a tech firm like Google?)
Both parties get something they want this transaction. Alphabet gets the Berkshire halo effect and a guaranteed buyer of $10 billion worth of equities, Berkshire gets a large tranche of equity at a price they believe is fair.
I think they view Alphabet as their next Apple, and a relatively safe place to ride out whatever happens with AI: Alphabet is fairly well positioned for the upturn or the downturn, especially now with this expanded warchest of cash.
Even if AI crashes 90% SpaceX, OpenAI & Anthropic are worth say 200B each post IPO. In 10-20 years with similar effects to Internet they might be the next Meta. Apple, Microsoft of the world.
But Google will likely still be the leader if it can make good on it's advantages.
Yes, but we are talking about liquidity not valuations...
The risk if it doesn't work out is that everyone gets diluted 1.67%
In this statement, their 2025 capex was $91.45B. They expect their 2026 capex to be $180B-190B. And they expect their "2027 capital expenditures to significantly increase compared to 2026."
So they simply don't have the money. Up until now, I thought the bubble talks about AI were silly because all these companies were using cash flow to fund their capex. These numbers are so astronomical now that a company that had $132B in net income has to take debt or issue stock to pay for it.
If i was a google cfo and was trading at a premium to my peers before that, i'd want to raise the cash now. Look at MSFT, they're trading at 25 forward p/e and were buying back shares at 40. If they have to issue equity over the next few years the spread between teh performance of the 2 cfos could be 40-50b on that alone.
Semiconductor/ Big Oil/ Rail/ Telco have.
I can invest perfectly in an always up market.
1. There’s real profit/value expected in pursuing the full automation of the labor market to the extent that the Board will approve large debts to known allies (BH) who only invest in long term infrastructure.
So they are investing in more AI infrastructure with long term capital because they see the payoff in the long term.
2. That also means they aren’t doing market moving plays in public like selling corporate debt because they don’t want to be in the short term froth with a long term bet.
But null hypothesis p=0.3 or something right?
Because the obvious answer is that he has compelling financial data telling him that this $80B now will produce a positive return on investment in the future. But you of course seem to disagree.
They are considered a very low risk and can borrow for a long time at low rates. They recently issued a 100 year bond.
They seem to have decided to issue equity rather than borrow more. This is probably so that they can maintain the ability to borrow very cheaply in future if necessary.
Issuing new equity might be a financial engineering experiment. No other mag7 has tried it. Plus they got BH name on the plate.
So I guess Google doesn’t think their stock is particularly cheap, but Berkshire Hathaway wants to buy more anyway. (At a slight discount.)
2008 wasn't a serious downturn?
Sundar and many of his executives have certainly read or heard of The Innovator's Dilemma, and I expect they're all moderately paranoid that it will be their downfall.
Also, that's not it. Google has a great ai app called Gemini where they have at various points hosted the top ai image generation model (certainly for speed, and for a while for accuracy) and have innovated with features like deep research
They are monetizing their ai conversations more effectively than OpenAI could dream of via ads and chat in Google search.
They are heavily investing in compute and talent.
When they've added llm results to Google search it has _increased_ engagement and re-engagement.
What part of the competition are they blissfully ignoring?
(I have counter arguments to some of these points, but I would rather hear other people's)
Are they actually implementing ads in chat yet? I haven't seen an ad in Gemini yet.
Again, the results I've seen is that LLM results in search have resulted in more zero-click searches (as a proportion of all searches), which isn't increasing engagement? But again, I may be wrong, what are you basing your assertion on?
I didn't say they were blissfully ignoring anything. I gave them credit for knowing the situation they're in and doing something about it.
The problem that I was talking about (probably badly getting my point across) is that it's internal conflict and strife that causes the pain here. One part of the company is incentivised on increasing revenue on the existing business. The other part of the company is incentivised on increasing revenue for the new business. But the new business is at the expense of the old business, so it sets up internal conflict where each part of the business tries to protect its own incentives. And Google has always been afflicted with rife internal politics.
If there is a "crack" there you might be able to get out of it, or it will let a disruptive idea to grow, but my way of thinking about innovator's dilemma is that is it a "culture bias": knowing about it give you some small advantage but it needs a real change to maybe have a chance to escape/act on it and the most important part is that under pressure it will quickly and imperceptible run the entire process or decision making.
But I also disagree with your reading of the innovators dilemma. You're being far too absolute
Have you seen their Cloud business?
Moreover, Google has continued to drive search growth since ChatGPT arrived and is executing competently. Their models are good (not great), but they have enough compute and one of the best ML-focused chips such that they aren't beholden to Nvidia (instead, they're beholden to fabs: tsmc - this is a much better dependency since Nvidia is hell bent on extracting as much value as they can from their position in the stack and it would be against the nature of tsmc to behave similarly)
Will Google's ad revenue decrease? Advertising is an incredible business because it is anti fragile.^ Even if search revenues decrease from their current highs (I would bet heavily against this), they still have YouTube with shorts and a robust display ads business that is going to improve if AI supercharges the economy (more companies - # startups founded in Jan 2026 is much higher than # founded the previous January, more products, advertising and distribution become the differentiators for these products)
If you're wondering how anthropic is going to continue to grow its base, the answer is advertising. In fact, Google is situated to fundamentally support everything that anthropic needs. Who cares if they make worse margins than anthropic? They'll benefit from the entire ride up, and they'll do the same for the next startup of that scale.
coding models? their own devs use claude code.
a bit more nuanced take on the failure would also account for executives backgrounds at the critical period:
- in 1981 Vince Barabba — Kodak's Head of Market Intelligence — conducted an extensive internal study that explicitly concluded digital photography could replace film and that Kodak had approximately 10 years to prepare for the transition.
- Kodak's leadership in 1980–1993 saw the company through the lens of its founding identity — silver-halide chemitry, precision coating and manufacturing, and the extraordinarily high margins of the film-plus-processing business. This identity-driven decade was spent on failed diversification and defending film instead of building an electronics cost structure and a defensible high-margin position. They steered capital and attention toward businesses that fit that self-image (specialty chemicals, pharmaceuticals, hybrid film products) rather than toward digital cameras, which meant fighting Sony and Canon on low-margin electronics turf where Kodak felt no competence and feared cannibalizing film.
- It was an inside executive culture, crystallized in the 1990 choice of film-lifer Kay Whitmore over the digital-minded Phil Samper. When Chandler retired, the finalists were Whitmore and vice-chairman Phil Samper, who had a deep appreciation for digital technology. The board chose Whitmore, and was explicit about why: as the New York Times reported, Whitmore said he would keep Kodak closer to its core businesses in film and photographic chemicals. Samper resigned and went on to become president of Sun Microsystems and then CEO of Cray Research — i.e., to lead exactly the kind of digital/computing companies Kodak was avoiding becoming.
- so when Kodak did get serious to compete in digital (in 1993 board made Fisher the CEO, he came from running Motorola and held an engineering degree plus a doctorate in applied mathematics) it did so as one commodity hardware maker among many and that was too late since film began to drop as digital started to pick up, exactly as Vince Barabba predicted in 1981
Google is well positioned to earn from this service, especially if they can prove that their search service is superior to competitors. While they lose some of their moat, they are well positioned to dominate the market, just like they did in the consumer space.
People asking any AI chat interface for ideas for their honeymoon will trigger some kind of search. SEO is still relevant and Google might still be able to sell top spots in their search so LLMs will pick it up.
"You tried to find a recipe for cupcakes, well all I can offer you is an advert on kitchen appliances"
Some already do, and some of the ones that don't will in the future.
See for example https://help.openai.com/en/articles/20001047-ads-in-chatgpt
Of course that's not to say that the advertising situation will be identical to that of pre-LLM search engines, and the differences may lead to radically different economic models and user experiences. But I was just correcting your statement.
In practice there's a lot of issues with asymmetric information. The company knows its own operations and financial position better than random traders on Wall Street. It is rational for it to buy back stock when the market value is lower than the true intrinsic value of the company, and to sell stock when the market value is higher than the true intrinsic value of the company. Therefore, traders often treat buybacks as a signal that the company is "cheap" (at least in the company's own view) and pump up the price accordingly, and treat stock issuances as a sign that company management believes that the stock is "expensive" and push it down accordingly. Company management has more inside information than market participants do, but is usually prohibited from trading on it. Stock issuances and stock buybacks are one of the few cases where insider-initiated trading is legal, because the benefits accrue to the company as a whole rather than a few individuals.
And in this specific case, selling shares to Berkshire at a 5% discount has a pretty clear signalling effect.
The company has less cash in the balance sheet, so its market cap decreases. But there are fewer shares, so the share price is the same.
(This allows hypothetical future growth to disproportionately benefit existing shareholders, but does not intrinsically increase stock price.)
In practice, like another poster pointed out, it signals the company’s belief that its own shares are undervalued, so the market usually increases its estimation of value.
price is more broad and brings in supply vs demand effects.
However, if someone gives you a dividend you typically have to pay tax, and lots of people really hate paying tax.
So buybacks are the preferred price neutral way of dealing with excess cash.
It's not based on the fundamental value of the stock so maybe you wouldn't consider it "first order," but I think you can still call it "mechanical."
But before-paying-dividend versus after-paying-dividend decreases the value of a share.
ChatGPT's growth is incredible, but they essentially have to get all of their growth from inside codex or ChatGPT apps. Google can auto query Gemini with every search. There's an interesting piece of data which is that this tells them (on a conditioned basis^^) when is the chat result more effective than the search and vice versa?
Google can force growth of Gemini by leveraging their existing properties. This is a huge asset, and if you're wondering why Meta has artificially high usage of their LLMs it's because distribution is hard and Meta and Google have a lot of surface area to distribute on
^ no, I don't mean this as a compliment, although it does lend credence to the idea that Google is willing to update its current products with AI.
^^ ie conditional on the user's willingness to view the chat result
If you ask questions, it will enable "AI overview" , but if we search about particular object/platform like "Google stock" or "bbc news", it will give the old classic search experience and we woulnd't need to swallow "AI overview" pill in that case.
Turns out licensing is separate for "code" and "pro"...
Well done lads
we started model testing the cost/performance of our skills and agents and flash 3.5 wins in most things.
As people develop harnesses for their codebase i think the intelligence required comes down a lot.
Google makes it very hard to use their shit and it was full of bugs.
Anthropic's current run is based entirely around Claude Code in this space and the last time I used the gemeini-cli it wouldnt give me access to the latest models and I was paying them for the privilege
https://github.com/google-gemini/gemini-cli/discussions/2727...
I get the complaints in that thread but I still think it is hilarious. That repo is a gong show to random shit and perhaps one of the best worst examples of "opensource" LLM development.
Sometimes you have to tab across and give it a PW, but it seemingly is incapable of parsing that, and just asking.
Kiro, what we use at work, on the other hand will just prompt you. (And doesn't like taking credentials directly)
Kiro is of course really good to back into AWS stuff, it knows more about AWS than Amazon themselves!
Gemini is really good at understanding my inane ramble and mis-spelling
In any case, it's well known that devs in Google have liked anthropic/openai models for coding more than gemini, so unless they're hiding their best models from the people within, I think it's just the case that they're behind.
Even with anthropics record breaking revenue growth I don't see how the pure AI companies can sustain, but the catch-22 is that any obvious pivot proves that. This puts the more traditional tech companies in position to ride the back of the wave until the growth curve tops.
However the agent-based conversational future simply does not support that level of valuable [1] ad volume, which collapses Google's carefully optimized tech+business stack.
Like I said, they will still thrive, but more because of GCP (which might see the biggest growth due to AI and the other tech + infrastructural advantages you mentioned) and the other businesses (YouTube, Waymo, etc.) However, their current cash cow is being disrupted, primarily by themselves, and I don't yet see how they can monetize agents nearly as lucratively as they've monetized search.
[1.] Sure, they could keep stuffing ads into each turn of the conversation, but 1) as I theorized in the linked post, those would be meaningless and low-value, and 2) that could just push users to competitors like ChatGPT, Anthropic or Perplexity that can offer a cleaner UX because they're starting from a clean slate and don't (yet) have the same revenue expectations to meet.
While your linked post was explicitly about ads, this line wasn't and is just baseless
If you think that (the quoted line) you should short Google (haha you will lose all your money even if you're right, market...irrational...solvent)
Search is going to be fine. It's about half their revenue. Display is another large chunk and feels largely immune to the affects of AI, and Google is not a static entity. And their Cloud business is seeing unbelievable growth.
I mean, the ad business is still by far the largest chunk of their whole business, which is why I think their overall revenues, profits and dominance will decrease ¯\_(ツ)_/¯
Display is another good example of why. Like half (?) of it is YouTube, which will be fine, but the rest comes from non-first party properties, which are already seeing drastic drops in traffic. Just as with the shrinking of SERP ads, there's no way to stuff enough ads in chatbot conversations or to compensate for the display ad revenue dropping either.
I'm long Google (monopolies are usually good bets, bullish on AI + GCP, plus it is disrupting itself before others can so has good long-term prospects) but I can't see how that will compensate for its main cash cow today being cannibalized, so I fully expect future returns to be less than historic ones.
I highly suspect they opaquely lowered usage limits on me.
https://www.theverge.com/news/627849/auto-draft
But I believe since then Anthropic have raised more money, almost certainly diluting Google's stake (I could be wrong and misremembering that Google didn't partake in the additional fundraising). I have in the back of my head that Google is down to something like 10% now, but don't have time to go and find details to fact check that, sorry!
I don't think that's true, mostly in that a lot of usecases are solved via coding models + a harness.
> Google's models are driving cars right now.
Yes + other models like alphafold. But those are (relatively) specialized models. Besides, the comment I was responding to was saying Google is sandbagging the market to keep it calm or something. I don't disagree that Google is doing well overall and has some clear advantages
I mean, the product might not be called Google search, but the thing that Google search transforms into will likely make more money than it does today and my bear case is that it would still make >80% of today's watermark.
Also Google's going to grow auxiliary businesses around AI that will change the balance of their assets dramatically (waymo is ~nothing right now, cloud is growing well, and there's no reason Google shouldn't be able to produce a top coding model for speed if not frontier level)
Depends on the product - whether protein bars, salty chips, cellular service, or IPhone or something else. If your product has a flavor, it’s never going to get commoditized. Coke still tastes better than Pepsi.
It will be interesting to see if the LLM companies can establish their own "brand" and how they will do that. LLM voice is a thing but not sure if it's a good thing people will use to hang their self identity on. Distillation of models and constant training also make this complicated. Claude code is winning on harness and ux right now but it seems precarious and also easy to commoditize. I think elon tried to add branding to his chatbot pretty intelligently by being iconically crude/evil/"anti woke" since it's both highly visible and less likely to be copied.
We live in fascinating times!
Google has gone all in on AI. To the point of challenging their own core product. Apple is waiting and seeing. Google is building and distributing, albeit with terrible marketing.