Imagine selling alphabet because they are doing better than ever? Like of all the companies still standing after the AI bubble pop dust settles, you know Alphabet will be one of them. They just took AI and slapped it onto their search which had the easiest and most fsr reaching integration there was.
Just because it’s easy doesn’t make it a good idea. One of the reasons google is so profitable is that each search is super cheap computationally for them to do. So, the few ads that are on the page more than make up for the search cost.
Now add an AI overview. They still need to run the search and also generate ~100-300 tokens. Tokens are computationally expensive to generate compared to a search. So, now they have more than doubled the cost of a search. Add on top of that, fewer people click on ads in the search now because the AI overview gives them the info they need without scrolling down to the ads.
Now, I am sure google is smart enough to figure this out eventually, but when you are dealing with something at the scale of google searches, really small changes have huge impacts.
The same could be said for basically all of the big S&P500 companies right now.
There’s definitely always been strange investor sentiment around Alphabet. It’s ran basically the same way as other big companies. But investors have always tended to be skeptical and undervalue it, followed by sudden increases as it corrects.
>The same could be said for basically all of the big S&P500 companies right now.
And I think it is a problem for all of them. IMO they are approaching LLMs fundamentally wrong for the most part. Most companies seem to want to approach it similar to previous tech buildouts where the more users you get, the better your margins are. Each user was computationally inexpensive to add for things like email, streaming, search, file sharing etc. A single server could potentially serve 100 users or more depending on the type of service. So scaling fast dramatically improved your margins.
LLMs don't scale like that because they are so computationally expensive. LLMs it's more like 5-20 users per instance simultaneously, depending on the model. On top of that the server's are way more expensive and use way more electricity than other services. And that is just serving the model, nevermind the cost and capacity needed to train a new model. My company serves small open weight LLMs for our employees to use, our largest model is glm-5.3 flash and those are the kind of scale numbers I see before a model becomes too slow for a user. I have no insight into training or how much these frontier models cost to serve, but I would imagine its worse for scale.
They are approaching it like we will scale fast and loose money now and that will improve our margins to the point where it is profitable, but I fundamentally think it wont and they will have to change their approach. They will eventually all have to switch to a pay per token system for everyone and everything like OpenAI and Anthropic had to for enterprise users when tokenmaxxing took too much of a toll on their costs. They will actually be able to operate at a profit at that point, but will have dramatically fewer users willing to actually pay those costs.
> They are approaching it like we will scale fast and loose money now and that will improve our margins to the point where it is profitable, but I fundamentally think it wont and they will have to change their approach.
I think the idea is to get people hooked, and then gradually ramp up the cost, regardless of "scalability". Look at something like Uber, or Doordash, or one of the other gig apps. They also scaled really poorly from a financial perspective (a driver has a hard limit of how many orders or services at a time) but have gradually raised their prices higher and higher, while paying workers less (for example, Uber Drivers no longer get an increased payment for surge pricing; all that extra money goes directly to Uber), as they have eliminated competion like taxi services.
They dont need to neccesarily change their approach, they just "need" to slowly turn up the water's tempature so we dont realize we are boiling.
> They are approaching it like we will scale fast and loose money now and that will improve our margins to the point where it is profitable, but I fundamentally think it wont and they will have to change their approach.
I feel like this is working though.
Anecdotally more and more people I know have a subscription to some sort of AI. Google specifically has an advantage because they can bundle it with other services like Youtube Music or Storage.
Cheap models are getting amazing - Luna is better than the frontier models of 2025 for like 1/10th of the cost. Look at 5.2 vs 5.6-luna
If models keep getting cheaper at a similar rate, a model as good as what google uses for search might be a fraction of the price next year.
And it is exactly why Google is not counted among the frontier model battle. They do release good model once in a while, but it is not in their interest to keep burning cash to stay ahead of OpenAI and Anthropic when they have already been bleeding money giving people free AI access left and right (and center with search). They are more concerned with lowering token cost, which reflects partly in their API cost per 1m token (with the exception of Argon).
Its more expensive than search yes but I dont think the computations are that expensive once youre outside the top models. theyre just pouring money into frontier models because you whoever gets it will be outcompeted.
I've already had a few instances of the ad being embedded in the AI overview. People trusting the AI overview is still a delicate thing, but if google gets over that hump, they will pump it full of ads until it is about to burst.
Yeah I was only fine with giving Stadia a spin because my wife bought me a controller as a gift and it got me the service for free for a year or something. I was more curious to see how the technology worked. (We have fiber so it actually worked really well).
Also I always had a PS5, so I wasn’t really that bothered when it died. Still hilarious that Google lives up to their reputation though.
I saw an ad for Google Search the other day and almost had an aneurysm trying to figure out what I was looking at. Why are you trying to sell me an inferior version of the thing I already loved?
That approach gives itself to AI and it has also more publishers more wary of opening their content to search engines. More and more high quality content is now behind paywalls.
Not a lawyer. So there's a copyright issue and an access issue here. The copyright issue comes down to whether the summaries are a compilation of the source material into a new document, and thus a derivative work, or a fair use transformation of the source material for the purpose of reporting, scholarship, or research.
Personally, I suspect that principled courts will eventually find these summaries to be derivative because they
have a strong commercial purpose: expanding or maintaining google's search engine hegemony, which is the primary source of its revenue
have a cataclysmic effect on the market for or value of the original material because the entire purpose of these summaries is to let their users avoid having to look for elsewhere by patronising the owners of the original. The people who write the originals do so for the purpose of driving internet traffic to their websites and products/services. By knowingly and effectively curtailing traffic, they are fundamentally undermining the incentives for and revenue from these original.
But there's also a much more straight forward access issue. Just because a web server/ website is public, doesn't mean it can't impose rules on its access and use. If these original sources require you to accept their terms and conditions before accessing them, and their terms and conditions clearly prohibit google/google's automated crawlers/google's ai tools from accessing those sources, then it just becomes a straight up business tort/ contract issue. And contract issues have lots of remedy options.
The main reason why more content, even free content, is being gated by login requirements/ ToS splashscreens is because it's much easier to establish the elements needed to bring a claim against companies like google for reusing their materials even if the produced documents themselves are found not to infringe copyright law.
I've been assuming that the sheer scale of Google means they probably benefit from just returning a lot of these "AI Overviews" from cache rather than running them from scratch but who knows.
I am sure that they do this, but they also cache the searches as well. So it really comes down to how much a new search vs a new ai overview costs. Based on cost estimates for how much a google search costs from a few years ago vs the cost of 300 tokens from gemini 3.8 flash, it might actually be closer to cost parity than I thought. Both seem to be around $0.003 when you round them.
Even at cost parity though having to run both still doubles the cost of a search and it is a defensive move. Adding the overview doesn't get more people to do more google searches. It helps prevent people from moving to chatgpt. So, at best they doubled their search cost for neutral growth.
I’ve been assuming they are intentionally losing money on this to help starve out the competition. If they can keep people from using OpenAI and Anthropic by being more convenient and cheaper, then eventually those competitors fail and then Google can take AI back out of search. Classic dumping strategy.
Anyone can have record revenue if they sell at a loss and have money to burn. Record revenue and negative cash flow is a good story for a start up, not a pillar of the global economy.
But they are swimming in cash and clearly will figure this out unless everyone is too stubborn to let AI just fade into being an option and not a new deity whose outcome doesn’t make any sense long term
Yeah, they will eventually come out on top (even though their AI currently lags behind both OpenAI and Anthropic), but their current valuation is not justified with the cashflow problems they are facing. Microsoft had a similar shit earnings, but they are compensating it with the current layoffs once more.
Im gna be real, idk anything about AI, really. Very little about how cose or digital architecture works.
But i recently was gna break down and finally upgrade my google to the 400gb... and all options had a bunch of AI add ons.
I dont use any AI except the search summery and when i need to ask a weird comparison question. Sure ill use mire, but chatbots? Ai art? All that is a turnoff. I stopped using our packaging company because they made us talk to an ai first and it was always wrong.
Well... idk, i dont hate AI, but wtf i dont want to pay for these services, and frankly, i dont wanna support it more than i have to.
Its so stupid, my whole boz is on google and i am afraid to upgrade bc if i do, i have no idea wtf the AI means.
Do they need all my data for training? My pictures?
And it doesnt help, now that photos has so many layers on it... i cant even search my own, or my partners nudes! It censors my own fucking photos.
I have like 10k photos, i cannot search all of those. And they got rid of a lot of workarounds like Skin or Breast.
Annoyed. I just want my pictures and be able to search them for things, like it did before.
3 quarters of search growth decline and guided down. which is their core business . And they are valued on it growing ; the company will be okay but the valuation could continue to come in (or maybe it doesn't , who knows, great company either way )
If you want to evaluate Alphabet's "real" economic growth, you can compare its nominal growth rate against the prevailing inflation rate: Nominal Growth: Alphabet's Q3 2026 revenue grew by 25% year-over-year. Real Growth: With US CPI inflation trending at approximately 2.5% to 3% over the same period, Alphabet's real revenue growth remains exceptionally strong at over 21%, indicating that its top-line expansion is driven by actual business scaling and market demand rather than just rising prices.
Even without ads, for the last two decades Google directed people to third party websites which then collect user data and serve ads or just generally increase brand awareness.
Now with AI Google scrapes the websites and serves it up without anyone ever clicking a link. Media organizations and the like won’t be happy about it.
Furthermore, you continously have to spend a lot on inference, otherwise you fall behind and customers straight up go to the competition. So it's not like there's an initial investment that may pay off - no, even in the best of cases it will stay like this...forever.
Gemini 3.5 flash lite is reportedly the model used in search. The paid Flash Lite API is cheaper than the paid search API. Tokens are not more computationally expensive to generate than a search. (I mean, they can be but they aren't for Google.)
Ok even let's say it even only costs 25% as much as the search does. They still have to do both. So, the search now costs 125% as much as it did before the ai overviews. This additionally doesnt make people search more, no one is doing more searches beacause it now has an ai overview. This is a defensive move to prevent people from searching on chatgpt or other services. So at best they have increased their per search costs for no increase in the number of searches.
I guarantee you their margins on search are excellent, it prints money. I doubt flash lite is more expensive; it mostly replaces the old question-answer thing which in hindsight is a shittier version of an LLM that probably cost about the same as flash lite to deliver.
But flash-lite + search probably costs about the same as search has ever cost. Google is good at this. They wouldn't do something that makes searches cost 125% more unless it made then 130% more in ads.
I have ad blocker on my laptop but on my projected that plays youtube, I noticed literally in the past few weeks that youtube is now running way more ads now than ever before.
I wonder if its to make up for the ad runs coming up short in the google search, as you just described.
I mean, I’m sure they’ve accounted for the top ~1,000,000,000,000+ searches and while that “LLM window” suggests burning a token, it’s probably just being populated with text canned language for whatever specific request
If your hypothesis is true then every single other AI company is also in trouble, but clearly the stock market does not think so. Also, tokens are quite cheap now. A lot of engineering and smarts have been spent into making it so, these are not stupid people.
Tokens are so cheap that actual consumers and users dont create enough profit for AI companies. Thats where the money is in the long run. But what happens if people dont pay? What if they go for Chinese open source AIs instead? Because as we seen, people dont actually need the best model. They need a model which creates slop/functions/whatever-they-need which is barely good enough. When a consumer chooses between "this is good enough" and "this is the best and most expensive", how many actually choose latter? I mean, just look at stuff people buy at Temu. Its inferior quality, not always even working, and rarely safe to use. And they buy that.
Same goes with AI. People might not want to pay the real price of AI at all. Then what happens for US/non-China AI firms?
I think the regular consumers arent really targets for AI. Its companies using AI to target consumers. Like imagine some defense company able to hyper target decision makers for the US military procurement. They probably spend million on lobbyist. But if they could just advertise to those 10 people they pay 500K to do it.
Indeed. And people buy possibly non-functional crap from Temu just for the chance that the cheapest option possible is somehow good enough and works out for at least for a while.
I dont think people are going to pay what AI companies want. Thats just my guess. But we the common people have so many other things to pay for...Those who can pay, i wonder if those people numbers and wallets are enough to keep the train rolling.
> If your hypothesis is true then every single other AI company is also in trouble,
Are they not? The only reason why Anthropic was able to turn a profit for 1 single quarter is because they cut a deal with Musk as they rent compute from his servers.
They are all bleeding money like crazy, with Nvidia being the only clear winner in all this. And thats because they are selling the hardware, not the service.
Not compared to the cost of a google search. That AI overview cost significantly more than the actual search does. For arguments sake though let's say they find a solution to get it to cost parity with the search. They have still doubled the cost of a search, which has to be made up somewhere, and they did it as a defensive move. This doesn't grow search, more people aren't googling things because they added an AI overview. They did it to defend against chatgpt. So in the absolute best case scenario it's neutral for them.
>these are not stupid people
I fundamentally disagree, they have smart people working for them, but the tech leaders that actually make the decisions are pack animals. Just following whatever the current hype is regardless of if it makes sense for their business or not.
They must be caching the AI overview for like the top 100k searches. When people type in exact keywords, like a movie title or athlete's name, it's got to be a cached response. That would save them massive compute.
I am sure that they cache the searches as well as the overviews. so it's really kind of irrelevant. whatever search is cached is also going to have the overview cached also. So we are really talking about the cost of new un-cached searches vs the overview cost. Even if the ai overview is cheaper to run than the search itself (lets say 25% of the cost of a search for arguments sake) they still have to do both now. That means the search is now 125% the cost before the ai overview existed. The thing is that the ai overview is not making more people search things on google. This is a defensive move to prevent people from moving to chatgpt and other services. So at best they have increased their per search costs while being neutral on the number of searches performed.
And also the AI overview results in me never seeing (or at least noticing) ads. The ads on Google are easy to ignore. The ads on the pages I used to click through to, actually caught my eye sometimes.
They're 100% caching their ai response though. It becomes a database look-up. There are even super cheap ai algos that can take similar phrases and differently ordered of search terms, categorize them cheaply, and serve the same ai response. You can also see this when some of them fall out of date.
I already feel the pull to not Google something and just use Claude, and I am the sort that is aware of why that is bad. I imagine the layperson is or soon will be no longer seeking information via a web search but just asking chatgpt. When that truly takes hold, goodbye search ad revenue, and that's their cash cow.
It's certainly eating into the ad revenue of digital content creators and their publishing houses. What AI searches are killing is the SEO driven market. At the end of the day, claude isn't going to make you a new pair of shoes or come over to fix your lawn. (At least not yet). AI search engines will still be able to extract rents from product and service providers as long as those providers need the internet to drive business.
If anything, AI will make these providers more dependent on their engines because they can no longer get customers by publishing stories like "5 secret tricks for lacing your shoes like an Olympic Sprinter!" and "How this grass will save you $$$ on your water bill!." So while AI search is definitely going to kill jobs, transform industries, and make the internet a sadder, emptier place, in the long-run, it's probably not going to affect the bottom line of Search companies themselves. You're going to need to break out the antitrust laws for that to happen.
For me I mostly replaced Google search with Gemini because I figure if it was going to show up in a Google search it would probably be in their AI product anyways, and if you just set up personal rules you can make it respond with directly quoted facts and links only if you don't want it thinking about things too much. And it actually does a better job of cutting through all the fuckin SEO'd to hell sites out there that want to charge me to view them or have 8 billion ads. And the latest versions have all been intelligent enough that you can ask for explanations of why it responded the way it did or you can even read the detailed chat log to see its "thoughts".
I also have Claude from my work and it's great for what I need from it which is prototype interfaces and useful software tools/applets I can use for whatever I'm doing, but Gemini Argon just came out and it's fucking wild to use for coding.
I am genuinely impressed as hell by it; I was able to create a tool that does hourly scans of jira and slack and email to see if anything needs my attention or response and gives me multi-choice options to respond/deal with it all, then identifies gaps between our production app and the prototype to create tickets for bringing both to parity, and more. And all I was doing was testing it because I just got access, I didn't expect it to be able to actually achieve what I was asking for on the first try.
I had seen reports of drops in Google Search but the official Google earnings call said they are experiencing a slowdown in growth, which implies search is still growing.
Just posting the latest link - https://www.searchenginejournal.com/google-search-growth-slows-again-ai-mode/594862/
MSFT went nearly 15 years without gaining any ground after 2000. Doesn’t matter if they survive if your time horizon isn’t long enough. The writing is starting to show. The market hates the amount of CAPEX being funded by borrowing and equity issuance. The market will look to companies with better FCF. That used to be Google, but not anymore.
You’re right that Alphabet is still a good company and will likely survive an AI transition or bubble pop. This is just the market revealing it’s not a good investment at its current share price.
Search and overall internet quality got so bad that I find myself using the (free) claude or chatGPT app if I need anything slightly complex. For simple things I might just go to wikipedia directly.
I certainly am not using google as much as I used to.
Queries up, rate of revenue growth down, which is what you would expect if an agent was dispatched, fired 20 queries in parallel, and returned a five lines summary.
I don't pretend to have done deep market research, that's the usage pattern I see more and more around me, where "Ask chat about it" is slowly replacing "just google it".
- whether ppl switch from traditional to chat or AI search
- whether Google remains norm in AI search
If people are in fact switching to AI, then Queries up indicates Google is remaining norm even in AI search.
With revenue still growing, but 12% vs 15% means monetizatikn for AI mode is still being figured out. That seems super easy to solve and more tactical rather than strategic (aka just add ads to ai mode screen)
> If people are in fact switching to AI, then Queries up indicates Google is remaining norm even in AI search.
Not necessarily, if the search is done from chatGPT or Claude. I'm not sure if that's the case, mind you, I'm not sure if they're using google's index.
It could also be a fundamentally harder venue to monetize. It seems to be like it's trying to achieve two opposite goals: Give you a useful summary, and make you click on the link.
The other risk I can see is chatGPT/Claude turning into the "default" apps for the normal user. If they can search as well, why switch to google at all?
Not necessarily. The ChatGPT from four years ago looks downright primitive compared to what we have today. Habits can be slow to change too.
I'm running some home renovations entirely from the free version of Claude. It handled searching for contractors, materials, crafting and sending emails, creating documents on my machine ...
It even checked our HOA bylaws by controlling my browser tab and navigating the archaic portal they're using.
That sort of affordance is not necessarily obvious, but I see more people getting up to speed. If you're not slightly tech oriented, you wouldn't guess it could do all that.
It's something the AI search mode can't compete with from a browser. I'm a software engineer, I use AI all the time, and I'm not even sure Google has an equivalent desktop/mobile app, that's how little mindshare they have in the space.
I mean honestly it doesn’t seem like you have a well formed thesis here
If users didn’t switch before, why would they switch now given Gemini is the best model on nearly every category? https://artificialanalysis.ai/#intelligence-category-tabs
You gave me a link where Gemini carbon is only first on their intelligence index and nowhere else?
Check the "Intelligence Index vs. Cost per Task", that's the one that matters. ChatGPT 6.1 sol has a score of 55 to gemini's 59, but is about six times cheaper per task. That is, if you put much faith in benchmarks at all.
Best model is absolutely meaningless when a lot of them are just good enough. What matters is the best product, and google doesn't really have a great track record in that area.
Google search went to absolute crap after they started putting AI into it. You can literally search for EXACTLY what you're after, and you can't then find it in the top 10 results.
Hell, most things on prior-to-AI Google search would take you to Wikipedia for generic terms (e.g. if you searched for "fuse" or "what is a fuse" or similar). Now you get deep-linked random pages on random websites with no authority, an AI answer that hesitates forever and then answers wrongly, and relevant/useful links are pages and pages down in the results.
For instance, one that hits me regularly now... I have an energy provider called Octopus Energy.
There are NOT MANY THINGS called Octopus Energy. If you just searched for just "Octopus", sure, put me into a botanical definition or pictures of octopuses, or whatever. But it doesn't. It gives you random companies called that. If you search for "Octopus Energy", it links you to some deep-linked page on there. I ended up on an EV rebate scheme once as the top link for that.
Ironically, their domain is octopus.energy (at least in the UK). If you remove AI entirely ("-ai" in your search) it actually does far better. But you still get nonsense.
If you're smart enough to know that Octopus Energy should be top for a search for "Octopus Energy", you could instead pluck out the My Account or homepage as the first link, not some random FAQ deeply hidden in that website. That helps nobody.
They've improved it slightly recently, but for about the last year or so, Google search has been worthless even when you KNOW what you are trying to get to.
(e.g. I needed to search up a bundled PHP package in a hurry - guess why! - which I normally could ALWAYS find quite easily.
This time around, I had to literally go and find the guy's name, and use that specifically, and then I still didn't end up on the actual PHP package page that I needed, but I was able to click around and get to it.
Searches for PHP just went to (again) random deep links on PHP.net, random news articles about PHP, or vulnerability alerts that told you to update packages.)
Dunno man, when I run a search for Octopus Energy the first result I get is the energy company
My anecdotal experience is that AI search is fixing many of the issues that have plagued normal search in recent years. The question Google now faces is how to monetize them.
It comes and goes - they fix it and then break it for ages, and then fix it. It matters what computer you're on, what location you're in, whether you have AI search "turned on" or not.
But for the last year, Google has been pretty useless for basic "I know exactly what I'm asking for and it's so very specific there should be only one top result" searches with any consistency.
That's what being a software engineering director can do for you. You say it like it is a bad thing son. What's the problem? How dare I engage with the platform!
Alphabet lost 10% in premarket trading after it missed expectations on adjusted earnings per share and revenue in the third quarter. The Google parent posted negative free cash flow of $9.07 billion. The company kept its full-year capital expenditures forecast at $195 billion to $205 billion.
It's doing better than Anthropic which lost $80,000,000,000
The Q3 10-Q has the split. Other income included an unrealized net loss of $12.4B on marketable and other equity securities, which the filing says came "largely from SpaceX", partly offset by a $3.0B unrealized net gain on the non-marketable stakes. So roughly $9.4B pre-tax of the quarter was mark-to-market on holdings, and it nets the opposite way from the private-company stake people assume is the problem.
For the miss itself, operating income is the cleaner comparison, since none of those marks touch it. The real cash story is separate: operating cash flow of $42.3B against $51.4B of capex is where the negative $9.1B free cash flow comes from.
Please tell me you guys aren't so ignorant that you believe anthropic actually lost that amount of money.
You do realize that "$80 billion dollar loss" was just an accounting fair value adjustment for convertible debt, right? It means the stock went up lol...
That only speaks to the absurdity of current markets
130bn that they raised through funding rounds. Now compare that to their contracted spending obligations of the medium term
Largely driven by cheap compute deals which aren't going to continue. Operating profit is irrelevant if your whole business model relies on unprofitable model training costs. Even if they actually make a profit on inference alone (doubt), they literally can't stop throwing money at training without being caught by open weight competitors
Compute costs are going to tank over time and if you think that Chinese models will ever be used for anything relevant in the west then I have a bridge to sell you. Despite the narrative and some bench optimized "competitors", nobody is close to anthropic in practice. Big tech has and will continue to subsidize the industry until it stands in its own.
It's overvalued as hell and I wouldn't touch it. Nobody is denying that. At the same time, parroting accounting adjustments as thought they're operating costs just shows how ignorant people in this subreddit are, not that we needed another proof point after seeing some of deranged takes on the Cali billionaire tax that pop up here
“If Alphabet shares fall more than 9.77% on Thursday, that would be its worst post-earnings reaction ever.”
I think it’s incredible that Alphabet has never had a worse post-earnings reaction. That’s truly remarkable. Wall Street is so fickle. Companies plummet more than 10% when their revenues / earnings miss by just a few percent. It goes to show how well run the company is.
In this case it’s caused by greater than expected negative cash flows, I assume due to AI infrastructure build-out. It’s ultimately something they control. Maybe they can’t stop it immediately because commitments are made and contracts are signed, but they can choose to ramp it down if necessary. It’s a big bet but it’s not the farm.
Ya seems much more likely it’s about the negative FCF.
Everyone focusing on the minuscule search miss is mostly ppl who were dead wrong the last 4 years and now believe this proved them right. More human psychology than anything
omg bro we've been in late stage capitalism since the term was coined to describe the global economy after WWI. The global economy of 1927 could not be more different than today and yet it's all "late stage." Nobody of any brain power takes this term seriously.
But to your point, we are discussing what is in Google's control to do. It's not crazy to think they won't do layoffs just like the other tech firms did.
"Capitalism" in general just means "not socialism, whatever I believe socialism to be."
It's always late because the predicted socialist paradise is always just around the corner. The alternative is admitting they've been making wildly inaccurate predictions for centuries.
It's just street preachers holding up "the end is near" signs, but re-branded to seem less religious. Like the Scientology of apocalyptic preaching.
To be pedantic, generally the concept of "Late Capitalism" that typically is described dates back to the 70s with Ernest Mandel and is an entirely different concept with no connection to Sombart's theory. It just has the same name.
It tends to begin post WW2, and touches on the economic expansion, and the rise of neoliberalism and globalization. Which ARE still relevant economic factors.
> In contemporary academic or journalistic usage, "late stage capitalism" often refers to a new mix of:
- Strong growth of the digital, electronics and military industries as well as their influence in society.
- Economic concentration of corporations and banks, which control gigantic assets and market shares internationally.
- Transition from Fordist mass production in huge assembly-line factories to Post-Fordist automated production and networks of smaller, more flexible manufacturing units supplying specialized markets.
- Increasing economic inequality of income, wealth and consumption.
- Consumerism on credit and the increasing indebtedness of the population.
Terms change over time, and just because 2 words meant something 100 years ago doesnt mean its referring to the exact same thing today.
The global economy post WWII has been the most successful force in lifting people out of poverty that the world has ever witnessed. The global economy is also so much stronger and different than post ww2.
If you describe the global economy in 1954 with the same term as you do in 2026 then your term is ridiculous. If you describe the global economy in 1946 vs 2026 using the same word, then you are brainless.
You seem to really struggle with context and comprehension. If I showed you a painting I finished today, and told you I started it last week, do you think it looked the same day one as it did day 7?
No one is saying that right after we won WW2, we were plunged into late stage capitalism; that is just the moment in time used to begin charting the effects of our social, economic, and governmental systems that have been in use.
Things dont just happen, its the description of the gradual changes that have been enacted decade after decade and where that is taking us. No one is saying that everything was hunky dory, and then 1954 rolled around, created neoliberalism, and now we are living in a dystopia. That is just when a lot of the balls got rolling, such as amplifying and increasing the military industrial complex (something that was warned about before WW2 as well).
> If you describe the global economy in 1954 with the same term as you do in 2026 then your term is ridiculous.
Very oddly pro-alphabet comment section… slowing growth in search revenue with the backdrop of many people switching to other chatbots for search. Not a great story to sell and makes sense to have such a market drop. I think people are starting to wake up to the fact that these massive capital outlays are not paying off and there is no timeline for when they will
Yeah, either someone needed to cash out, Berkshire shifted out with new management replacing buffet, or the market is making the same mistake yet again betting against a monopoly.
People believing “Google” is beholden to its existing search product don’t fully understand how they’re a massive part of the internet’s core infrastructure.
Everyone has an opinion on Google because they’re so intertwined with every business and consumer application…
Google is at an entirely different stage of business than the LLM and chip fab companies- and there isn’t anything simple that can disrupt their ecosystem. They’d really need to fumble.
Their antitrust cases often fail because they’ve been so durable and non-obvious in their dominance.
The question becomes- how quickly can they grow YoY as compared to newer companies? Do they shift to offering a larger dividend?
Anecdotally, Google got me to pay them to try Gemini Pro for 3 months. If you consider that I've never paid a dime to use Google search, but happily paid a, albeit discounted, price for Gemini, that's gotta mean something for consumer behaviour shifting and consumer's willingness to pay for this tech. That's the story Google needs to sell.
Bro. They host all those bots. lol. That’s the point of CSP . They are more than the LLM which is also impressive. I dunno. Reacting to 1 Q report is just a correction. They still at 400 hundo this time next year.
Its very simple, google search engine just gets used less. The first option is an AI prompt. Dont forget that the cloud/AI activity grew big time.
Edit: dont forget they're part of Anthropic/gemini/claude, etc. people dont know how to read reports, but then again it's a share that lots of people picked up without knowing something about it and they get freaked out.
Also, people avoiding negative facts on trading or stocks is always funny to me. Take the hit haha. If you can't stand the heat, don't play the game.
Edit 2: search still grew 12% lol. Even with the AI prompts. They're still steady, people just keep expecting this crazy jumps they're used to now.
Think big. Now is the time Google goes all-in to ensure a leading position down the line. This is more important than the short-term stock price, and there’s no cause for alarm. Alphabet will be the winner here. They are ahead of the pack in developing quantum computing technology—evidenced by their "Willow" processor and their cutting-edge research into error correction. The synergy between AI (DeepMind) and quantum computing will be the decisive factor, and Alphabet has the deep pockets required to see it through to the end. If the stock dips a bit further—which is entirely possible—it could present a good entry point. That said, the company is currently contending with a wave of negative sentiment toward AI from those who view its development critically, as well as the non-commercial, open-source approach taken by Zuckerberg with Llama. For Meta, advertising revenue is sufficient; their investment in AI is primarily aimed at ensuring they never have to depend on others again.
You got me. Let`s say, Meta, Google, OpenAI and Anthropic, maybe even more players will fight it out. There is a lot of money left for number 5. It´s just my thinking, that as it sort of goes bad for Google now, they always surprised us with positive changes in the past. It´s just my personal bet they will make it.
And a postscript following the disaster involving Amazon, Intel, and Reddit: this is perfectly normal and by no means the death knell for the "AI brothers"—it is simply a sharp pullback during a period of stagnation, coming after an entirely excessive 6,000-point rally in April/May.
The stock plummeted about 10% in extended trading, reflecting the weaker-than-expected earnings report and negative free cash flow of $9 billion for the quarter, three times the estimated deficit. In the second quarter, Alphabet’s free cash flow went negative for the first time in its history as a public company, at $5.9 billion.
People focusing on the negative free cash flow are missing so much. Google brings in more revenue in interest than it's sending out. Through something like Waymo they can bring ad revenue into the real world. This truly is a monster company throwing money back into the business to become an even bigger monster.
Yes. Google has the luxury of sitting on a pile of cash while still being able to spend freely to build out their AI, even if they take a hit in the short run. They can outlast the vast majority of other AI companies and will reap the rewards when the bubble bursts. I’m sure they’ll even gobble up a few other the smaller players when that time comes.
Nobody beats the Wiz. It's got to be a completely different company but Bernie Williams was my favorite baseball player growing up. The Wiz reminds me of that.
Well, Wiz is basically Bernie’s lucrative ‘99 contract…so they’re hoping for the dynasty to continue on. I think they’ll be able to take on Boston for years to come.
I am very much of the opinion that Alphabet is going to emerge as victorious from the "AI wars", but they will not be immune to the "pop" - I will wait until some massive drop.
JamesLahey08 | 20 hours ago
Imagine selling alphabet because they are doing better than ever? Like of all the companies still standing after the AI bubble pop dust settles, you know Alphabet will be one of them. They just took AI and slapped it onto their search which had the easiest and most fsr reaching integration there was.
waitmarks | 20 hours ago
Just because it’s easy doesn’t make it a good idea. One of the reasons google is so profitable is that each search is super cheap computationally for them to do. So, the few ads that are on the page more than make up for the search cost.
Now add an AI overview. They still need to run the search and also generate ~100-300 tokens. Tokens are computationally expensive to generate compared to a search. So, now they have more than doubled the cost of a search. Add on top of that, fewer people click on ads in the search now because the AI overview gives them the info they need without scrolling down to the ads.
Now, I am sure google is smart enough to figure this out eventually, but when you are dealing with something at the scale of google searches, really small changes have huge impacts.
throwaway1847384728 | 20 hours ago
The same could be said for basically all of the big S&P500 companies right now.
There’s definitely always been strange investor sentiment around Alphabet. It’s ran basically the same way as other big companies. But investors have always tended to be skeptical and undervalue it, followed by sudden increases as it corrects.
waitmarks | 19 hours ago
>The same could be said for basically all of the big S&P500 companies right now.
And I think it is a problem for all of them. IMO they are approaching LLMs fundamentally wrong for the most part. Most companies seem to want to approach it similar to previous tech buildouts where the more users you get, the better your margins are. Each user was computationally inexpensive to add for things like email, streaming, search, file sharing etc. A single server could potentially serve 100 users or more depending on the type of service. So scaling fast dramatically improved your margins.
LLMs don't scale like that because they are so computationally expensive. LLMs it's more like 5-20 users per instance simultaneously, depending on the model. On top of that the server's are way more expensive and use way more electricity than other services. And that is just serving the model, nevermind the cost and capacity needed to train a new model. My company serves small open weight LLMs for our employees to use, our largest model is glm-5.3 flash and those are the kind of scale numbers I see before a model becomes too slow for a user. I have no insight into training or how much these frontier models cost to serve, but I would imagine its worse for scale.
They are approaching it like we will scale fast and loose money now and that will improve our margins to the point where it is profitable, but I fundamentally think it wont and they will have to change their approach. They will eventually all have to switch to a pay per token system for everyone and everything like OpenAI and Anthropic had to for enterprise users when tokenmaxxing took too much of a toll on their costs. They will actually be able to operate at a profit at that point, but will have dramatically fewer users willing to actually pay those costs.
brutinator | 18 hours ago
> They are approaching it like we will scale fast and loose money now and that will improve our margins to the point where it is profitable, but I fundamentally think it wont and they will have to change their approach.
I think the idea is to get people hooked, and then gradually ramp up the cost, regardless of "scalability". Look at something like Uber, or Doordash, or one of the other gig apps. They also scaled really poorly from a financial perspective (a driver has a hard limit of how many orders or services at a time) but have gradually raised their prices higher and higher, while paying workers less (for example, Uber Drivers no longer get an increased payment for surge pricing; all that extra money goes directly to Uber), as they have eliminated competion like taxi services.
They dont need to neccesarily change their approach, they just "need" to slowly turn up the water's tempature so we dont realize we are boiling.
dust4ngel | 15 hours ago
> the idea is to get people hooked
aka wait for people to cognitively de-skill
ThingsThatMakeMeMad | 17 hours ago
> They are approaching it like we will scale fast and loose money now and that will improve our margins to the point where it is profitable, but I fundamentally think it wont and they will have to change their approach.
I feel like this is working though.
Anecdotally more and more people I know have a subscription to some sort of AI. Google specifically has an advantage because they can bundle it with other services like Youtube Music or Storage.
Cheap models are getting amazing - Luna is better than the frontier models of 2025 for like 1/10th of the cost. Look at 5.2 vs 5.6-luna
If models keep getting cheaper at a similar rate, a model as good as what google uses for search might be a fraction of the price next year.
hopefuldonkeykick | 13 hours ago
> If models keep getting cheaper at a similar rate, a model as good as what google uses for search might be a fraction of the price next year.
If models keep getting cheaper at a similar rate, you're just going to be putting a FOSS one on your phone in a year.
kronpas | 8 hours ago
And it is exactly why Google is not counted among the frontier model battle. They do release good model once in a while, but it is not in their interest to keep burning cash to stay ahead of OpenAI and Anthropic when they have already been bleeding money giving people free AI access left and right (and center with search). They are more concerned with lowering token cost, which reflects partly in their API cost per 1m token (with the exception of Argon).
HumorAccomplished611 | 17 hours ago
Its more expensive than search yes but I dont think the computations are that expensive once youre outside the top models. theyre just pouring money into frontier models because you whoever gets it will be outcompeted.
Spez_is-a-nazi | 20 hours ago
Ad blockers are also more popular than ever. It’s a cat and mouse game there but Google is starting to take the fight more seriously.
seeyam14 | 20 hours ago
Does an adblocker block links in an ai response?
ImamTrump | 18 hours ago
No. Ads have different tags
falooda1 | 20 hours ago
No evidence for this btw other than Reddit feelings
Spez_is-a-nazi | 19 hours ago
LOL https://backlinko.com/ad-blockers-users
If it wasn’t a problem why would Google be trying to stop it, but hey you said something incredibly stupid without researching it, so that’s fun :)
https://trends.google.com/explore?q=Ad%2520blocker&date=all&geo=Worldwide
But go on :)
arstin | 19 hours ago
I've already had a few instances of the ad being embedded in the AI overview. People trusting the AI overview is still a delicate thing, but if google gets over that hump, they will pump it full of ads until it is about to burst.
214ObstructedReverie | 12 hours ago
> they will pump it full of ads until it is about to burst.
Google admits AI search makes less money, but it’s fixing that with more ads
karl4319 | 19 hours ago
It will take them 6 to 9 years to figure out a project is a failure. Google+, google stradia, and google glass all had similar timescales
DJanomaly | 17 hours ago
Stadia was 3 years. I had a founders edition. At least they refunded me for the controller.
considerspiders | 12 hours ago
I think about this comic often whenever one of the big boys just decides to turn something off
https://www.penny-arcade.com/comic/2019/06/10/stadioid
DJanomaly | 11 hours ago
Yeah I was only fine with giving Stadia a spin because my wife bought me a controller as a gift and it got me the service for free for a year or something. I was more curious to see how the technology worked. (We have fiber so it actually worked really well).
Also I always had a PS5, so I wasn’t really that bothered when it died. Still hilarious that Google lives up to their reputation though.
Economy_Zombie_3026 | 15 hours ago
I saw an ad for Google Search the other day and almost had an aneurysm trying to figure out what I was looking at. Why are you trying to sell me an inferior version of the thing I already loved?
broadcastday | 20 hours ago
And I'm still unclear: are there any IP/copyright implications for "summarizing" a website in search results?
gimpwiz | 17 hours ago
Not a lawyer but I don't think that summarizing is protected by copyright.
ogaat | 20 hours ago
That approach gives itself to AI and it has also more publishers more wary of opening their content to search engines. More and more high quality content is now behind paywalls.
Un-resigned | 10 hours ago
Not a lawyer. So there's a copyright issue and an access issue here. The copyright issue comes down to whether the summaries are a compilation of the source material into a new document, and thus a derivative work, or a fair use transformation of the source material for the purpose of reporting, scholarship, or research.
Personally, I suspect that principled courts will eventually find these summaries to be derivative because they
have a strong commercial purpose: expanding or maintaining google's search engine hegemony, which is the primary source of its revenue
have a cataclysmic effect on the market for or value of the original material because the entire purpose of these summaries is to let their users avoid having to look for elsewhere by patronising the owners of the original. The people who write the originals do so for the purpose of driving internet traffic to their websites and products/services. By knowingly and effectively curtailing traffic, they are fundamentally undermining the incentives for and revenue from these original.
But there's also a much more straight forward access issue. Just because a web server/ website is public, doesn't mean it can't impose rules on its access and use. If these original sources require you to accept their terms and conditions before accessing them, and their terms and conditions clearly prohibit google/google's automated crawlers/google's ai tools from accessing those sources, then it just becomes a straight up business tort/ contract issue. And contract issues have lots of remedy options.
The main reason why more content, even free content, is being gated by login requirements/ ToS splashscreens is because it's much easier to establish the elements needed to bring a claim against companies like google for reusing their materials even if the produced documents themselves are found not to infringe copyright law.
Toribor | 18 hours ago
> and also generate ~100-300 tokens.
I've been assuming that the sheer scale of Google means they probably benefit from just returning a lot of these "AI Overviews" from cache rather than running them from scratch but who knows.
cogit4se | 11 hours ago
They use semantic caching so they can return cached AI overviews even for queries that are phrased slightly differently.
waitmarks | 18 hours ago
I am sure that they do this, but they also cache the searches as well. So it really comes down to how much a new search vs a new ai overview costs. Based on cost estimates for how much a google search costs from a few years ago vs the cost of 300 tokens from gemini 3.8 flash, it might actually be closer to cost parity than I thought. Both seem to be around $0.003 when you round them.
Even at cost parity though having to run both still doubles the cost of a search and it is a defensive move. Adding the overview doesn't get more people to do more google searches. It helps prevent people from moving to chatgpt. So, at best they doubled their search cost for neutral growth.
JoshuaEdwardSmith | 16 hours ago
I’ve been assuming they are intentionally losing money on this to help starve out the competition. If they can keep people from using OpenAI and Anthropic by being more convenient and cheaper, then eventually those competitors fail and then Google can take AI back out of search. Classic dumping strategy.
poopzains | 20 hours ago
They had record revenue
IKillZombies4Cash | 20 hours ago
Anyone can have record revenue if they sell at a loss and have money to burn. Record revenue and negative cash flow is a good story for a start up, not a pillar of the global economy.
But they are swimming in cash and clearly will figure this out unless everyone is too stubborn to let AI just fade into being an option and not a new deity whose outcome doesn’t make any sense long term
Kindly_Type2327 | 20 hours ago
Yeah, they will eventually come out on top (even though their AI currently lags behind both OpenAI and Anthropic), but their current valuation is not justified with the cashflow problems they are facing. Microsoft had a similar shit earnings, but they are compensating it with the current layoffs once more.
poopzains | 8 hours ago
No. Not anyone can have record revenue. Sorry. Just not true.
Djaja | 20 hours ago
Im gna be real, idk anything about AI, really. Very little about how cose or digital architecture works.
But i recently was gna break down and finally upgrade my google to the 400gb... and all options had a bunch of AI add ons.
I dont use any AI except the search summery and when i need to ask a weird comparison question. Sure ill use mire, but chatbots? Ai art? All that is a turnoff. I stopped using our packaging company because they made us talk to an ai first and it was always wrong.
Well... idk, i dont hate AI, but wtf i dont want to pay for these services, and frankly, i dont wanna support it more than i have to.
Its so stupid, my whole boz is on google and i am afraid to upgrade bc if i do, i have no idea wtf the AI means.
Do they need all my data for training? My pictures?
And it doesnt help, now that photos has so many layers on it... i cant even search my own, or my partners nudes! It censors my own fucking photos.
I have like 10k photos, i cannot search all of those. And they got rid of a lot of workarounds like Skin or Breast.
Annoyed. I just want my pictures and be able to search them for things, like it did before.
akmalhot | 20 hours ago
3 quarters of search growth decline and guided down. which is their core business . And they are valued on it growing ; the company will be okay but the valuation could continue to come in (or maybe it doesn't , who knows, great company either way )
Churchbushonk | 20 hours ago
When considering inflation?
poopzains | 8 hours ago
Here ya go. Enjoy. Google gotcha.
Factoring in Real Growth
If you want to evaluate Alphabet's "real" economic growth, you can compare its nominal growth rate against the prevailing inflation rate:
Nominal Growth: Alphabet's Q3 2026 revenue grew by 25% year-over-year.
Real Growth: With US CPI inflation trending at approximately 2.5% to 3% over the same period, Alphabet's real revenue growth remains exceptionally strong at over 21%, indicating that its top-line expansion is driven by actual business scaling and market demand rather than just rising prices.
waitmarks | 19 hours ago
Yes, but it’s mostly from massive gains in their cloud business.
poopzains | 8 hours ago
Yes. Cloud services is how private corporations are hosting. That include private LLM hosting for compliance isolation. Its a good thing.
PMmeDonutHoles | 14 hours ago
Literally means nothing. Every one of these big companies report record revenue pretty much every quarter
poopzains | 8 hours ago
I mean we knew they were spending. Estimates were off is all. People acting llke they are in trouble are silly.
ResearcherSad9357 | 14 hours ago
Cisco had record revenue in 2000, ~55.5% growth!
ToInfinity_MinusOne | 19 hours ago
Even without ads, for the last two decades Google directed people to third party websites which then collect user data and serve ads or just generally increase brand awareness.
Now with AI Google scrapes the websites and serves it up without anyone ever clicking a link. Media organizations and the like won’t be happy about it.
ys2020 | 19 hours ago
Adds (marketed items) are in the llm output now. They're doing just fine.
Odra_dek | 18 hours ago
Furthermore, you continously have to spend a lot on inference, otherwise you fall behind and customers straight up go to the competition. So it's not like there's an initial investment that may pay off - no, even in the best of cases it will stay like this...forever.
FlyingBishop | 17 hours ago
Gemini 3.5 flash lite is reportedly the model used in search. The paid Flash Lite API is cheaper than the paid search API. Tokens are not more computationally expensive to generate than a search. (I mean, they can be but they aren't for Google.)
waitmarks | 17 hours ago
Ok even let's say it even only costs 25% as much as the search does. They still have to do both. So, the search now costs 125% as much as it did before the ai overviews. This additionally doesnt make people search more, no one is doing more searches beacause it now has an ai overview. This is a defensive move to prevent people from searching on chatgpt or other services. So at best they have increased their per search costs for no increase in the number of searches.
FlyingBishop | 17 hours ago
I guarantee you their margins on search are excellent, it prints money. I doubt flash lite is more expensive; it mostly replaces the old question-answer thing which in hindsight is a shittier version of an LLM that probably cost about the same as flash lite to deliver.
But flash-lite + search probably costs about the same as search has ever cost. Google is good at this. They wouldn't do something that makes searches cost 125% more unless it made then 130% more in ads.
TheTimberFactory | 17 hours ago
I have ad blocker on my laptop but on my projected that plays youtube, I noticed literally in the past few weeks that youtube is now running way more ads now than ever before.
I wonder if its to make up for the ad runs coming up short in the google search, as you just described.
EclipsePixy | 16 hours ago
You’re right, it seems like every two minutes there’s 1 or 2 ads. It’s annoying and I don’t think it will be good in the long run
flbnah | 13 hours ago
I mean, I’m sure they’ve accounted for the top ~1,000,000,000,000+ searches and while that “LLM window” suggests burning a token, it’s probably just being populated with text canned language for whatever specific request
kajaktumkajaktum | 19 hours ago
If your hypothesis is true then every single other AI company is also in trouble, but clearly the stock market does not think so. Also, tokens are quite cheap now. A lot of engineering and smarts have been spent into making it so, these are not stupid people.
retro_ditto | 18 hours ago
Tokens are so cheap that actual consumers and users dont create enough profit for AI companies. Thats where the money is in the long run. But what happens if people dont pay? What if they go for Chinese open source AIs instead? Because as we seen, people dont actually need the best model. They need a model which creates slop/functions/whatever-they-need which is barely good enough. When a consumer chooses between "this is good enough" and "this is the best and most expensive", how many actually choose latter? I mean, just look at stuff people buy at Temu. Its inferior quality, not always even working, and rarely safe to use. And they buy that.
Same goes with AI. People might not want to pay the real price of AI at all. Then what happens for US/non-China AI firms?
HumorAccomplished611 | 17 hours ago
I think the regular consumers arent really targets for AI. Its companies using AI to target consumers. Like imagine some defense company able to hyper target decision makers for the US military procurement. They probably spend million on lobbyist. But if they could just advertise to those 10 people they pay 500K to do it.
twenty_three_three | 9 hours ago
Its tough because your example, targeted hyper specific marketing and communication, is not something the LLM AI can do all that well.
usernameelmo | 18 hours ago
> When a consumer chooses between "this is good enough" and "this is the best and most expensive", how many actually choose latter?
going by smartphone cameras vs. actual cameras, most choose this is good enough
retro_ditto | 15 hours ago
Indeed. And people buy possibly non-functional crap from Temu just for the chance that the cheapest option possible is somehow good enough and works out for at least for a while.
I dont think people are going to pay what AI companies want. Thats just my guess. But we the common people have so many other things to pay for...Those who can pay, i wonder if those people numbers and wallets are enough to keep the train rolling.
brutinator | 18 hours ago
> If your hypothesis is true then every single other AI company is also in trouble,
Are they not? The only reason why Anthropic was able to turn a profit for 1 single quarter is because they cut a deal with Musk as they rent compute from his servers.
They are all bleeding money like crazy, with Nvidia being the only clear winner in all this. And thats because they are selling the hardware, not the service.
> Also, tokens are quite cheap now.
Thats how a loss leader works.
waitmarks | 18 hours ago
>tokens are quite cheap now
Not compared to the cost of a google search. That AI overview cost significantly more than the actual search does. For arguments sake though let's say they find a solution to get it to cost parity with the search. They have still doubled the cost of a search, which has to be made up somewhere, and they did it as a defensive move. This doesn't grow search, more people aren't googling things because they added an AI overview. They did it to defend against chatgpt. So in the absolute best case scenario it's neutral for them.
>these are not stupid people
I fundamentally disagree, they have smart people working for them, but the tech leaders that actually make the decisions are pack animals. Just following whatever the current hype is regardless of if it makes sense for their business or not.
Spektr44 | 17 hours ago
They must be caching the AI overview for like the top 100k searches. When people type in exact keywords, like a movie title or athlete's name, it's got to be a cached response. That would save them massive compute.
waitmarks | 17 hours ago
I am sure that they cache the searches as well as the overviews. so it's really kind of irrelevant. whatever search is cached is also going to have the overview cached also. So we are really talking about the cost of new un-cached searches vs the overview cost. Even if the ai overview is cheaper to run than the search itself (lets say 25% of the cost of a search for arguments sake) they still have to do both now. That means the search is now 125% the cost before the ai overview existed. The thing is that the ai overview is not making more people search things on google. This is a defensive move to prevent people from moving to chatgpt and other services. So at best they have increased their per search costs while being neutral on the number of searches performed.
crackanape | 7 hours ago
And also the AI overview results in me never seeing (or at least noticing) ads. The ads on Google are easy to ignore. The ads on the pages I used to click through to, actually caught my eye sometimes.
Richandler | 8 hours ago
> super cheap computationally for them to do.
They're 100% caching their ai response though. It becomes a database look-up. There are even super cheap ai algos that can take similar phrases and differently ordered of search terms, categorize them cheaply, and serve the same ai response. You can also see this when some of them fall out of date.
thx1138inator | 19 hours ago
I don't search anymore. I ask AI for the answer. No ADS! Love it!
crackanape | 7 hours ago
As your brain atrophies and you have an increased risk of getting bogus responses.
JamesLahey08 | 16 hours ago
No.
ogaat | 20 hours ago
They make most of their money from ads and search.
Search is being subsumed by AI and competition. Ads too are less effective because of AI content as well as AI searches and agents.
They own the whole AI stack but its ability to replace existing revenue is unproven.
pagerussell | 17 hours ago
I suspect AI is eating their ad revenue.
I already feel the pull to not Google something and just use Claude, and I am the sort that is aware of why that is bad. I imagine the layperson is or soon will be no longer seeking information via a web search but just asking chatgpt. When that truly takes hold, goodbye search ad revenue, and that's their cash cow.
Un-resigned | 9 hours ago
It's certainly eating into the ad revenue of digital content creators and their publishing houses. What AI searches are killing is the SEO driven market. At the end of the day, claude isn't going to make you a new pair of shoes or come over to fix your lawn. (At least not yet). AI search engines will still be able to extract rents from product and service providers as long as those providers need the internet to drive business.
If anything, AI will make these providers more dependent on their engines because they can no longer get customers by publishing stories like "5 secret tricks for lacing your shoes like an Olympic Sprinter!" and "How this grass will save you $$$ on your water bill!." So while AI search is definitely going to kill jobs, transform industries, and make the internet a sadder, emptier place, in the long-run, it's probably not going to affect the bottom line of Search companies themselves. You're going to need to break out the antitrust laws for that to happen.
Tim_Apple_938 | 16 hours ago
What was their ad revenue yesterday vs last quarter and/or last years Q3?
DistinctlyIrish | 9 hours ago
For me I mostly replaced Google search with Gemini because I figure if it was going to show up in a Google search it would probably be in their AI product anyways, and if you just set up personal rules you can make it respond with directly quoted facts and links only if you don't want it thinking about things too much. And it actually does a better job of cutting through all the fuckin SEO'd to hell sites out there that want to charge me to view them or have 8 billion ads. And the latest versions have all been intelligent enough that you can ask for explanations of why it responded the way it did or you can even read the detailed chat log to see its "thoughts".
I also have Claude from my work and it's great for what I need from it which is prototype interfaces and useful software tools/applets I can use for whatever I'm doing, but Gemini Argon just came out and it's fucking wild to use for coding.
I am genuinely impressed as hell by it; I was able to create a tool that does hourly scans of jira and slack and email to see if anything needs my attention or response and gives me multi-choice options to respond/deal with it all, then identifies gaps between our production app and the prototype to create tickets for bringing both to parity, and more. And all I was doing was testing it because I just got access, I didn't expect it to be able to actually achieve what I was asking for on the first try.
Tim_Apple_938 | 16 hours ago
> search is being subsumed
Is it?
ogaat | 16 hours ago
Yes and maybe.
I had seen reports of drops in Google Search but the official Google earnings call said they are experiencing a slowdown in growth, which implies search is still growing.
Just posting the latest link - https://www.searchenginejournal.com/google-search-growth-slows-again-ai-mode/594862/
Tim_Apple_938 | 16 hours ago
That exclusively says growth - where’s report of drop?
chockeysticks | 16 hours ago
Speaking as a consumer I ask ChatGPT more for typical search than Google now, so I can see that.
Tim_Apple_938 | 16 hours ago
What was google search revenue when ChatGPT came out and what is it now?
DocumentOk7579 | 20 hours ago
It's about relative what is already priced in.
FearlessPark4588 | 6 hours ago
It's about other people doing even better than Alphabet, despite how well Alphabet itself is performing.
belovedkid | 20 hours ago
MSFT went nearly 15 years without gaining any ground after 2000. Doesn’t matter if they survive if your time horizon isn’t long enough. The writing is starting to show. The market hates the amount of CAPEX being funded by borrowing and equity issuance. The market will look to companies with better FCF. That used to be Google, but not anymore.
bigred4723 | 20 hours ago
You’re right that Alphabet is still a good company and will likely survive an AI transition or bubble pop. This is just the market revealing it’s not a good investment at its current share price.
DJanomaly | 17 hours ago
Yeah it’s just reverted to its price…from March. It was just overpriced and now it’s inline with reasonable expectations.
lbreakjai | 20 hours ago
Search and overall internet quality got so bad that I find myself using the (free) claude or chatGPT app if I need anything slightly complex. For simple things I might just go to wikipedia directly.
I certainly am not using google as much as I used to.
Tim_Apple_938 | 16 hours ago
But search queries are up and growing more rapidly than ever?
lbreakjai | 16 hours ago
Queries up, rate of revenue growth down, which is what you would expect if an agent was dispatched, fired 20 queries in parallel, and returned a five lines summary.
I don't pretend to have done deep market research, that's the usage pattern I see more and more around me, where "Ask chat about it" is slowly replacing "just google it".
Tim_Apple_938 | 16 hours ago
I think you’re conflating two orthogonal concepts
- whether ppl switch from traditional to chat or AI search
- whether Google remains norm in AI search
If people are in fact switching to AI, then Queries up indicates Google is remaining norm even in AI search.
With revenue still growing, but 12% vs 15% means monetizatikn for AI mode is still being figured out. That seems super easy to solve and more tactical rather than strategic (aka just add ads to ai mode screen)
lbreakjai | 15 hours ago
> If people are in fact switching to AI, then Queries up indicates Google is remaining norm even in AI search.
Not necessarily, if the search is done from chatGPT or Claude. I'm not sure if that's the case, mind you, I'm not sure if they're using google's index.
It could also be a fundamentally harder venue to monetize. It seems to be like it's trying to achieve two opposite goals: Give you a useful summary, and make you click on the link.
The other risk I can see is chatGPT/Claude turning into the "default" apps for the normal user. If they can search as well, why switch to google at all?
Tim_Apple_938 | 15 hours ago
Surely if that was the case it would have happened already given ChatGPT went giga viral 4 years ago
Re difficulty it actually seems way easier to monetize - much more data and thus more intent
lbreakjai | 15 hours ago
Not necessarily. The ChatGPT from four years ago looks downright primitive compared to what we have today. Habits can be slow to change too.
I'm running some home renovations entirely from the free version of Claude. It handled searching for contractors, materials, crafting and sending emails, creating documents on my machine ...
It even checked our HOA bylaws by controlling my browser tab and navigating the archaic portal they're using.
That sort of affordance is not necessarily obvious, but I see more people getting up to speed. If you're not slightly tech oriented, you wouldn't guess it could do all that.
It's something the AI search mode can't compete with from a browser. I'm a software engineer, I use AI all the time, and I'm not even sure Google has an equivalent desktop/mobile app, that's how little mindshare they have in the space.
Tim_Apple_938 | 15 hours ago
I mean honestly it doesn’t seem like you have a well formed thesis here
If users didn’t switch before, why would they switch now given Gemini is the best model on nearly every category? https://artificialanalysis.ai/#intelligence-category-tabs
lbreakjai | 14 hours ago
You gave me a link where Gemini carbon is only first on their intelligence index and nowhere else?
Check the "Intelligence Index vs. Cost per Task", that's the one that matters. ChatGPT 6.1 sol has a score of 55 to gemini's 59, but is about six times cheaper per task. That is, if you put much faith in benchmarks at all.
Best model is absolutely meaningless when a lot of them are just good enough. What matters is the best product, and google doesn't really have a great track record in that area.
ledow | 17 hours ago
Google search went to absolute crap after they started putting AI into it. You can literally search for EXACTLY what you're after, and you can't then find it in the top 10 results.
Hell, most things on prior-to-AI Google search would take you to Wikipedia for generic terms (e.g. if you searched for "fuse" or "what is a fuse" or similar). Now you get deep-linked random pages on random websites with no authority, an AI answer that hesitates forever and then answers wrongly, and relevant/useful links are pages and pages down in the results.
For instance, one that hits me regularly now... I have an energy provider called Octopus Energy.
There are NOT MANY THINGS called Octopus Energy. If you just searched for just "Octopus", sure, put me into a botanical definition or pictures of octopuses, or whatever. But it doesn't. It gives you random companies called that. If you search for "Octopus Energy", it links you to some deep-linked page on there. I ended up on an EV rebate scheme once as the top link for that.
Ironically, their domain is octopus.energy (at least in the UK). If you remove AI entirely ("-ai" in your search) it actually does far better. But you still get nonsense.
If you're smart enough to know that Octopus Energy should be top for a search for "Octopus Energy", you could instead pluck out the My Account or homepage as the first link, not some random FAQ deeply hidden in that website. That helps nobody.
They've improved it slightly recently, but for about the last year or so, Google search has been worthless even when you KNOW what you are trying to get to.
(e.g. I needed to search up a bundled PHP package in a hurry - guess why! - which I normally could ALWAYS find quite easily.
This time around, I had to literally go and find the guy's name, and use that specifically, and then I still didn't end up on the actual PHP package page that I needed, but I was able to click around and get to it.
Searches for PHP just went to (again) random deep links on PHP.net, random news articles about PHP, or vulnerability alerts that told you to update packages.)
Imnejjek | 15 hours ago
Dunno man, when I run a search for Octopus Energy the first result I get is the energy company
My anecdotal experience is that AI search is fixing many of the issues that have plagued normal search in recent years. The question Google now faces is how to monetize them.
ledow | 15 hours ago
It comes and goes - they fix it and then break it for ages, and then fix it. It matters what computer you're on, what location you're in, whether you have AI search "turned on" or not.
But for the last year, Google has been pretty useless for basic "I know exactly what I'm asking for and it's so very specific there should be only one top result" searches with any consistency.
deuteronpsi | 19 hours ago
People are selling plenty of companies that are currently doing better than ever. Rivian is a great example.
node19 | 18 hours ago
The stock can drop short-term, even if the long-term outcome is positive.
Argosnautics | 18 hours ago
Isn't selling high what you're supposed to do with stock?
reelphopkins | 15 hours ago
Think it is important to note this comment comes from an account with 29k contributions in a year equalling about 79 contributions a day.
JamesLahey08 | 15 hours ago
That's what being a software engineering director can do for you. You say it like it is a bad thing son. What's the problem? How dare I engage with the platform!
TimSylvester_ | 14 hours ago
> Like of all the companies still standing after the AI bubble pop dust settles, you know Alphabet will be one of them.
Except that AI completely undoes their entire revenue method, even if they're the one providing the AI.
JamesLahey08 | 14 hours ago
No.
TimSylvester_ | 14 hours ago
Wrong.
JamesLahey08 | 12 hours ago
Wrong.
[OP] One-Emu-1103 | 21 hours ago
Alphabet lost 10% in premarket trading after it missed expectations on adjusted earnings per share and revenue in the third quarter. The Google parent posted negative free cash flow of $9.07 billion. The company kept its full-year capital expenditures forecast at $195 billion to $205 billion.
It's doing better than Anthropic which lost $80,000,000,000
AskngAQstn | 20 hours ago
Google has a big stake in Anthropic and Anthropic is also a major customer and business partner, providing models that run on Google’s custom TPUs.
Jellicle_KitCat | 19 hours ago
That's a lot of zeros.
Equivalent_Song3771 | 19 hours ago
Anthropic was a 8$ billion loss when taking into account the senior debt notes converting into stock
Tim_Apple_938 | 16 hours ago
What was Google’s miss if you ignore the paper loss of SpaceX stock like you’re doing w anthropic?
EquiblesAI | 13 hours ago
The Q3 10-Q has the split. Other income included an unrealized net loss of $12.4B on marketable and other equity securities, which the filing says came "largely from SpaceX", partly offset by a $3.0B unrealized net gain on the non-marketable stakes. So roughly $9.4B pre-tax of the quarter was mark-to-market on holdings, and it nets the opposite way from the private-company stake people assume is the problem.
For the miss itself, operating income is the cleaner comparison, since none of those marks touch it. The real cash story is separate: operating cash flow of $42.3B against $51.4B of capex is where the negative $9.1B free cash flow comes from.
No_Cell6708 | 21 hours ago
Please tell me you guys aren't so ignorant that you believe anthropic actually lost that amount of money.
You do realize that "$80 billion dollar loss" was just an accounting fair value adjustment for convertible debt, right? It means the stock went up lol...
diamondgrin | 20 hours ago
Yeah true - they've only burned 9bn in cash to date....
No_Cell6708 | 20 hours ago
$9 billion is nothing in this world lol
They have like $140 billion cash on their balance sheet.
They turned an operating profit in the most recent q
diamondgrin | 20 hours ago
No_Cell6708 | 20 hours ago
Compute costs are going to tank over time and if you think that Chinese models will ever be used for anything relevant in the west then I have a bridge to sell you. Despite the narrative and some bench optimized "competitors", nobody is close to anthropic in practice. Big tech has and will continue to subsidize the industry until it stands in its own.
bjornuntuit | 20 hours ago
We get it, you like the stonk. Go spend all your money on it and stop bothering people with fanboism.
No_Cell6708 | 20 hours ago
It's overvalued as hell and I wouldn't touch it. Nobody is denying that. At the same time, parroting accounting adjustments as thought they're operating costs just shows how ignorant people in this subreddit are, not that we needed another proof point after seeing some of deranged takes on the Cali billionaire tax that pop up here
kronpas | 19 hours ago
You still havent answered me where do you think that pre-adjustment accouting 'cost' came from
brutinator | 17 hours ago
> 3) They turned an operating profit in the most recent q
Because they got a one time discount on renting compute from Elon Musk; that discount wont apply for future quarters.
kronpas | 20 hours ago
And where do you think that debt came from, even before valuation adjusment?
Happy_Incident_9982 | 18 hours ago
Notably you people only come out of the woodwork for headlines that negatively portray shartmodei, instead of the blatantly bullshit positive press
Smooth-Way-8724 | 20 hours ago
“If Alphabet shares fall more than 9.77% on Thursday, that would be its worst post-earnings reaction ever.”
I think it’s incredible that Alphabet has never had a worse post-earnings reaction. That’s truly remarkable. Wall Street is so fickle. Companies plummet more than 10% when their revenues / earnings miss by just a few percent. It goes to show how well run the company is.
In this case it’s caused by greater than expected negative cash flows, I assume due to AI infrastructure build-out. It’s ultimately something they control. Maybe they can’t stop it immediately because commitments are made and contracts are signed, but they can choose to ramp it down if necessary. It’s a big bet but it’s not the farm.
Tim_Apple_938 | 16 hours ago
Ya seems much more likely it’s about the negative FCF.
Everyone focusing on the minuscule search miss is mostly ppl who were dead wrong the last 4 years and now believe this proved them right. More human psychology than anything
Keeltoodeep | 20 hours ago
They probably need to do more layoffs like MSFT and other tech firms did. They are pretty bloated with SWEs
BroughtBagLunchSmart | 19 hours ago
just fire everyone right? If you don't like that you don't like end stage capitalism!
Keeltoodeep | 19 hours ago
omg bro we've been in late stage capitalism since the term was coined to describe the global economy after WWI. The global economy of 1927 could not be more different than today and yet it's all "late stage." Nobody of any brain power takes this term seriously.
But to your point, we are discussing what is in Google's control to do. It's not crazy to think they won't do layoffs just like the other tech firms did.
ants_are_everywhere | 18 hours ago
"Capitalism" in general just means "not socialism, whatever I believe socialism to be."
It's always late because the predicted socialist paradise is always just around the corner. The alternative is admitting they've been making wildly inaccurate predictions for centuries.
It's just street preachers holding up "the end is near" signs, but re-branded to seem less religious. Like the Scientology of apocalyptic preaching.
brutinator | 18 hours ago
To be pedantic, generally the concept of "Late Capitalism" that typically is described dates back to the 70s with Ernest Mandel and is an entirely different concept with no connection to Sombart's theory. It just has the same name.
It tends to begin post WW2, and touches on the economic expansion, and the rise of neoliberalism and globalization. Which ARE still relevant economic factors.
> In contemporary academic or journalistic usage, "late stage capitalism" often refers to a new mix of:
- Strong growth of the digital, electronics and military industries as well as their influence in society.
- Economic concentration of corporations and banks, which control gigantic assets and market shares internationally.
- Transition from Fordist mass production in huge assembly-line factories to Post-Fordist automated production and networks of smaller, more flexible manufacturing units supplying specialized markets.
- Increasing economic inequality of income, wealth and consumption.
- Consumerism on credit and the increasing indebtedness of the population.
Terms change over time, and just because 2 words meant something 100 years ago doesnt mean its referring to the exact same thing today.
Keeltoodeep | 18 hours ago
The global economy post WWII has been the most successful force in lifting people out of poverty that the world has ever witnessed. The global economy is also so much stronger and different than post ww2.
If you describe the global economy in 1954 with the same term as you do in 2026 then your term is ridiculous. If you describe the global economy in 1946 vs 2026 using the same word, then you are brainless.
brutinator | 17 hours ago
You seem to really struggle with context and comprehension. If I showed you a painting I finished today, and told you I started it last week, do you think it looked the same day one as it did day 7?
No one is saying that right after we won WW2, we were plunged into late stage capitalism; that is just the moment in time used to begin charting the effects of our social, economic, and governmental systems that have been in use.
Things dont just happen, its the description of the gradual changes that have been enacted decade after decade and where that is taking us. No one is saying that everything was hunky dory, and then 1954 rolled around, created neoliberalism, and now we are living in a dystopia. That is just when a lot of the balls got rolling, such as amplifying and increasing the military industrial complex (something that was warned about before WW2 as well).
> If you describe the global economy in 1954 with the same term as you do in 2026 then your term is ridiculous.
Wait til you hear about the term "Modern"!
Keeltoodeep | 17 hours ago
The global economy is just getting started. Nothing "late" about it. The party has just opened its doors.
brutinator | 17 hours ago
And the conditions that started them didnt just pop into existance today.
Keeltoodeep | 16 hours ago
I don't think they have. The wealth of nations was written in the 70s and explains it far better without pretending like it's a "late" cycle to cope.
liberTyrion | 18 hours ago
Very oddly pro-alphabet comment section… slowing growth in search revenue with the backdrop of many people switching to other chatbots for search. Not a great story to sell and makes sense to have such a market drop. I think people are starting to wake up to the fact that these massive capital outlays are not paying off and there is no timeline for when they will
freeatlast101 | 17 hours ago
Cloud growth was +100%.
aaahhhhhhfine | 17 hours ago
Still not an enormous percentage of overall income though...
Gloobbie | 17 hours ago
You have to give it time to actually grow
Tim_Apple_938 | 16 hours ago
Are people switching to other chatbots for search?
Ppl said that since 2022 and search keeps growing double digits YoY continually since then
(last nights growth rate being 12%)
Cute-Illustrator-862 | 16 hours ago
They are growing 25% YOY and trading at sub 15 p/e. This stock is oversold.
Richandler | 8 hours ago
Yeah, either someone needed to cash out, Berkshire shifted out with new management replacing buffet, or the market is making the same mistake yet again betting against a monopoly.
RedErin | 16 hours ago
it's google, they'll be fine
JC_Hysteria | 13 hours ago
People believing “Google” is beholden to its existing search product don’t fully understand how they’re a massive part of the internet’s core infrastructure.
Everyone has an opinion on Google because they’re so intertwined with every business and consumer application…
Google is at an entirely different stage of business than the LLM and chip fab companies- and there isn’t anything simple that can disrupt their ecosystem. They’d really need to fumble.
Their antitrust cases often fail because they’ve been so durable and non-obvious in their dominance.
The question becomes- how quickly can they grow YoY as compared to newer companies? Do they shift to offering a larger dividend?
ths33333 | 9 hours ago
Anecdotally, Google got me to pay them to try Gemini Pro for 3 months. If you consider that I've never paid a dime to use Google search, but happily paid a, albeit discounted, price for Gemini, that's gotta mean something for consumer behaviour shifting and consumer's willingness to pay for this tech. That's the story Google needs to sell.
poopzains | 8 hours ago
Bro. They host all those bots. lol. That’s the point of CSP . They are more than the LLM which is also impressive. I dunno. Reacting to 1 Q report is just a correction. They still at 400 hundo this time next year.
justtobeherenotsure | 8 hours ago
Its very simple, google search engine just gets used less. The first option is an AI prompt. Dont forget that the cloud/AI activity grew big time.
Edit: dont forget they're part of Anthropic/gemini/claude, etc. people dont know how to read reports, but then again it's a share that lots of people picked up without knowing something about it and they get freaked out.
Also, people avoiding negative facts on trading or stocks is always funny to me. Take the hit haha. If you can't stand the heat, don't play the game.
Edit 2: search still grew 12% lol. Even with the AI prompts. They're still steady, people just keep expecting this crazy jumps they're used to now.
DapperTicket1564 | 15 hours ago
Think big. Now is the time Google goes all-in to ensure a leading position down the line. This is more important than the short-term stock price, and there’s no cause for alarm. Alphabet will be the winner here. They are ahead of the pack in developing quantum computing technology—evidenced by their "Willow" processor and their cutting-edge research into error correction. The synergy between AI (DeepMind) and quantum computing will be the decisive factor, and Alphabet has the deep pockets required to see it through to the end. If the stock dips a bit further—which is entirely possible—it could present a good entry point. That said, the company is currently contending with a wave of negative sentiment toward AI from those who view its development critically, as well as the non-commercial, open-source approach taken by Zuckerberg with Llama. For Meta, advertising revenue is sufficient; their investment in AI is primarily aimed at ensuring they never have to depend on others again.
Raspberrybye | 14 hours ago
Deepminders are being poached in droves
DapperTicket1564 | 13 hours ago
You got me. Let`s say, Meta, Google, OpenAI and Anthropic, maybe even more players will fight it out. There is a lot of money left for number 5. It´s just my thinking, that as it sort of goes bad for Google now, they always surprised us with positive changes in the past. It´s just my personal bet they will make it.
And a postscript following the disaster involving Amazon, Intel, and Reddit: this is perfectly normal and by no means the death knell for the "AI brothers"—it is simply a sharp pullback during a period of stagnation, coming after an entirely excessive 6,000-point rally in April/May.
Shot-Possibility-399 | 14 hours ago
The google ai search box is incredibly clunky and unpolished. Very annoying to use. Regardless of the results being generated.
[OP] One-Emu-1103 | 20 hours ago
The stock plummeted about 10% in extended trading, reflecting the weaker-than-expected earnings report and negative free cash flow of $9 billion for the quarter, three times the estimated deficit. In the second quarter, Alphabet’s free cash flow went negative for the first time in its history as a public company, at $5.9 billion.
Environmental_Park_6 | 20 hours ago
People focusing on the negative free cash flow are missing so much. Google brings in more revenue in interest than it's sending out. Through something like Waymo they can bring ad revenue into the real world. This truly is a monster company throwing money back into the business to become an even bigger monster.
bigtdaddy | 19 hours ago
yeah idk why people rather buy a company that doesnt have anything better to do with their money than buybacks
neddiddley | 15 hours ago
Yes. Google has the luxury of sitting on a pile of cash while still being able to spend freely to build out their AI, even if they take a hit in the short run. They can outlast the vast majority of other AI companies and will reap the rewards when the bubble bursts. I’m sure they’ll even gobble up a few other the smaller players when that time comes.
JC_Hysteria | 13 hours ago
Outside of its already dominant ecosystem, it’s nuts they’re sitting on Waymo…
People are also sleeping on the Wiz acquisition in the age of AI needs in security/cloud monitoring, etc.
Environmental_Park_6 | 13 hours ago
Nobody beats the Wiz. It's got to be a completely different company but Bernie Williams was my favorite baseball player growing up. The Wiz reminds me of that.
JC_Hysteria | 12 hours ago
Well, Wiz is basically Bernie’s lucrative ‘99 contract…so they’re hoping for the dynasty to continue on. I think they’ll be able to take on Boston for years to come.
Diligent_Fondant6761 | 20 hours ago
Google before AI was like Usain Bolt vs Average Man
After AI is like Usain Bolt vs Asafa Powell
Google would still win but I hope you get the message!!!
XmasMancer | 15 hours ago
You do know google has Gemini?
postexitus | 14 hours ago
I am very much of the opinion that Alphabet is going to emerge as victorious from the "AI wars", but they will not be immune to the "pop" - I will wait until some massive drop.
RetrieverDoggo | 10 hours ago
Google Gemini is one of the worst AI there is. Consistently tells me wrong info. Not sure how they're going to win when it's so bad.
postexitus | 7 hours ago
Doesnt matter. When others run out of money their models can be part of the dropdown.
Victorythagr8 | 10 hours ago
And which a.i model you are using? Because we can choose which model we can use. The standard Gemini is nowhere as powerful as the paid version.