gleepskip
macrumors 6502a
I finally read it.
He presumes the business model will never work and the entire industry will collapse. That's certainly not impossible, but it's looking less likely.
Some people are taking the success of AI quite hard.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Will 'Watch Everything Burn' When AI Bubble Bursts - Ed Zitron
- Thread starter MacRumors
- Start date
- Sort by reaction score
I finally read it.
He presumes the business model will never work and the entire industry will collapse. That's certainly not impossible, but it's looking less likely.
Some people are taking the success of AI quite hard.
Such an bloviated and frankly extremist take- I know this sort of opinion makes a lot of People (haters) happy- but it's almost certainly a junk opinion. Sure Ai is doomed but the future is the Vision Pro 😀 anecdotes like the fact I cant be bothered to use my Vision Pro even once a month but I do use Ai tools, love them or hate them, almost every day. Will there be a correction? of course. A full b bubble burst? perhaps. But this technology is not going back into the bag and it will mature and become standard technology in our day to day lives- mostly in ways we haven't even considered yet. And I don't think anyone serious thinks LLMs are where this ends up.Even without being an expert in these things its pretty obvious his market take is short sighted.
g\
g\
adrianlondon
macrumors 604
One could argue that googling did that once people stopped going to the library or buying encyclopaedias.
For me, the main issue is that people will just "ask AI" and do what it says and if it doesn't work, just ask AI again or give up. This means the number of people coming to forums to ask questions, and the number of people (with the opportunity and knowledge for) posting answers drops.
As AI, so far, just scrapes existing information and collates it into convincing language, how will it "learn" when no one is adding any new info to the places it scrapes?
The worse problem is people posting the output straight out of AI. I'm not referring to people using AI to help format their thoughts. I mean copy-paste from AI. The detrimental result is AI consuming its own output and slowly degrading. The problem is known as model collapse.
thebarrkbyte
macrumors member
Oh, there's a private school chain (whenever "chain" is used to describe a school, you know where the emphasis is) that's AI-forward—Alpha School. Wired published an investigation into them back in the fall and discussed how they're failing. Having worked for startups and being a teacher, i can say startup culture doesn't belong in K-8 education (I can see an argument for some of it in high school, particularly an alternative high school or a CTE program).
One of my students a couple of years ago swore she didn't copy from AI. I sat her down and started to read the first line: "Okay, it looks like you need to write a short essay response to the prompt your teacher gave you. Let's get started!" 😂
gleepskip
macrumors 6502a
No, that's not in there. Such hyperbole.
gleepskip
macrumors 6502a
- Private credit funds
- CalPERS
- San Francisco teachers (and similar public pension plans)
- Special-purpose vehicles (SPVs)
- Oracle
- Taiwanese ODMs (Quanta, Foxconn / Hon Hai)
- Korean investors
- U.S. hyperscaler stocks
- U.S. semiconductor stocks
You can come back to this thread every year for the next decade and people like Ed will always be wrong but whenever there is a 10-20% down move in the market he and his echo chamber will pat themselves on the back and act like prophets even though 10-20% down movements happen almost every year. They are called profit taking, manipulation, corrections and stock rotations.
www.cnbc.com
“Samsung said it was seeing exponential growth in demand for general-purpose computing alongside AI, and was closely monitoring the relative pace of demand for HBM and server DRAM while maintaining an optimal product mix to support long-term AI demand.
The company expects industry supply constraints to persist next year and said it will manage its HBM and DRAM businesses in a balanced manner to achieve an HBM market share in line with its conventional DRAM business.
For the second half of the year, the company expects robust demand for memory centered on servers and continuing AI infrastructure capex amid a broader adoption of agentic AI.
It also expects to see accelerating demand for server DRAM, eSSDs, and HBM. Enterprise SSDs are high-performance storage devices designed to deliver data quickly and are essential to AI infrastructure.
Samsung said it expects supply constraints to tighten further in 2027, with rapidly growing AI token generation driving exponential demand over the medium to long term. The company said more customers are seeking multi-year supply agreements to secure AI infrastructure capacity, giving Samsung a better visibility into future demand and greater flexibility in planning investments.
The South Korean tech giant added it has finalized agreements with top five global data center customers and is in the final stages of talks with five additional major accounts to support their AI related demand.“
Samsung Electronics second-quarter operating profit beats estimates on soaring AI chip demand
Samsung Electronics extended its record run with second-quarter operating profit topping analysts' estimates.
“Samsung said it was seeing exponential growth in demand for general-purpose computing alongside AI, and was closely monitoring the relative pace of demand for HBM and server DRAM while maintaining an optimal product mix to support long-term AI demand.
The company expects industry supply constraints to persist next year and said it will manage its HBM and DRAM businesses in a balanced manner to achieve an HBM market share in line with its conventional DRAM business.
For the second half of the year, the company expects robust demand for memory centered on servers and continuing AI infrastructure capex amid a broader adoption of agentic AI.
It also expects to see accelerating demand for server DRAM, eSSDs, and HBM. Enterprise SSDs are high-performance storage devices designed to deliver data quickly and are essential to AI infrastructure.
Samsung said it expects supply constraints to tighten further in 2027, with rapidly growing AI token generation driving exponential demand over the medium to long term. The company said more customers are seeking multi-year supply agreements to secure AI infrastructure capacity, giving Samsung a better visibility into future demand and greater flexibility in planning investments.
The South Korean tech giant added it has finalized agreements with top five global data center customers and is in the final stages of talks with five additional major accounts to support their AI related demand.“
robertosh
macrumors 65816
You know, I was thinking about this for a while and I think that Apple -by pure chance or by strategy- is not bad at all on all of this. At least if we assume that the bubble will burst. They arrived bad and late, and they decided to not continue investing in own infrastructure and use technology and resources from somebody else. This is a good move if the bubble burst, as they can just stop paying Google and that's it. But if AI finally became profitable, could be terrible as depending on others at this scale is always bad and it will be too late to build its own infrastructure.
MacAddict1978
macrumors 68000
You don't seem to grasp the economics. This has nothing to do with that OpenAI, Anthropic, et. al, are charging for tokens - its everything to do with their cost of doing business. Tokens are in use to try and reduce the bleeding which is silly because they're impaled on a telephone pole.
Let's break this out. Even if at a certain point enough data centers are built - they'll need eternal maintenance, upgrades, and the electric costs (which are insane) will be an never ending expense. Adoption of AI is only going to be 100X more by this magical year you cite.
At some point, investors will need a return - AI will need to make money - and that day is ages away and the time to recoup the investment is in a galaxy far, far, away.
The points ED makes have been long said by every expert out there - and everyone with a brain that can type basic math into a calculator.
Where this becomes a mess - tons of companies are building businesses using API's to call multiple LLM models. At some point, AI companies will need to course correct that pricing because they cannot be selling at massive losses. Then their own direct to consumer products will need to follow suit. So even in a best case scenario as above where data center builds slow, the cost will still be insane - and no one will be able to afford it.
Look at things like Gamma, or even Canva with AI integration - that's the stuff that will have the bubbles BURST and hard. Free use? That will go pop. Paid use? Corporate only and Agentic AI will need to prove its value which its not yet done but I beleive will in time.... which will create an entire other mess with all the jobs it will wipe out.
This is unprecedented because never in history has there been anything like this. It's kind of like inventing a real life Startship Enterprise but not having a power source for it that wouldn't cost 2 million times the daily electric bill of the entire planet earth to fly 5 miles.
MacAddict1978
macrumors 68000
And he didn't say they were in trouble because of "AI Technology." Though, that tech is the driving force behind some very bad economics.
Correct, I was responding to the message in the quote fwiw not a comment by Zitron.
MacAddict1978
macrumors 68000
Working in a data center and being one who manages the cost structure, expenses, and continued investment are very different things.
Cloud computing caused a similiar stir with data center build outs, but the processing power wasn't at play. They largely operated as hard drives where AI is a super computer and a hard drive - using far more power and processing need than what cloud computing/gaming do where most a lot of the load of on the end users device.
AI tech is just ahead of processing tech (power, cores, efficiency, etc.) for what everyone is trying to do with it. That's really the problem. The AI category is growing faster than the tech to support what it is doing, but the competition between AI companies is so fierce, they're throwing out features and abilities that should be dripped out over years and decades in weeks and months. Said in a past comment, but it's the equivilent of inventing a real life Star Ship Enterprise - or Star Destroyer - without the ability to power it yet.
Look at what people are paying right now to run the AC in the summer? Just the electricity costs alone make AI unprofitable - and that's before 20 other costs are thrown in.
theapplehead
macrumors 6502a
Let me start by saying I get that you (and most people for good reason) might not like AI and it becoming so prevalent and even being this buzzword that everyone won't quit talking about. And that is fair - I'm not crazy about it either.
However that does not change the reality that we are dealing with. Demand being tied up through the end of 2027 is reality. Working in the industry, I see it firsthand and that is already a guarantee - profitably or not.
And I do agree that these "AI guys" are operating at a loss at the moment with no profit to speak of. Interestingly that's not as important right now as the goal is making it happen rather than making money. Sure that won't last forever and I don't think anybody expects it to. As I mentioned in my last post, there will come a time that new competitors will spring up that can offer artificial intelligence systems at a lower cost point with the same or comparable features as the big guys (Google, Anthropic, etc.) so those big players will have no choice but to either exit the market or lower their pricing (whether that means finding cheaper ways to do it or operating at a loss til they do).
Point is, the market will continually work to "correct itself" and eventually there will be a downcycle, just not in the "AI is a bubble" way that so many people hope for. I know there are many folks literally betting against AI so they are hoping for this downturn sooner than later so they can capitalize on their option puts, but the proof of what they're and your saying is just not there.
And as far as affording it goes, whether or not it is right or makes sense, these hyperscalers are going all in on AI by borrowing significantly against their CapEx which (while it may not seem sustainable or even wise) shows how committed they are. So if AI is a bust as you and others like to imply, then major companies like Microsoft, Google, Amazon, etc. would "go under" because it would make the very thing they leveraged against worthless. The amount of billionaires funding these devlopments won't let that happen anyways and before that could even come about, it's much more likely that user data could be inflated or manipulated so that things don't "crash."
Anyways, I digress. The point is that nothing will happen overnight. Sure, there could be a slow fade and a rollback from how on fire things are today. But AI will not be going anywhere anytime soon, like it or not. I for one would like a return to simpler times before everything required megatons of water and power, but that's no longer a reality for our world, so we can either accept it or fall behind, the choice is yours.
SteveJay10
macrumors newbie
gleepskip
macrumors 6502a
Do you deny the fact that all this money pumped into AI by companies where even the people creating the AI honestly don't have any idea what AI is going to be and more importantly don't know how much money they're going to generate is an incredibly dangerous situation for the economy? I'm not rooting for a massive market crash. I hate the idea of rooting for failure, but this glazing and deification of AI that we've been seeing is insanely naive.
Except for the fact that with this AI development is that trillions are being pumped into the development of AI without any real way to see any kind of return of investment. This runs the risk of seeing a financial crisis that would make 2008 look tameYou can come back to this thread every year for the next decade and people like Ed will always be wrong but whenever there is a 10-20% down move in the market he and his echo chamber will pat themselves on the back and act like prophets even though 10-20% down movements happen almost every year. They are called profit taking, manipulation, corrections and stock rotations.
Samsung Electronics second-quarter operating profit beats estimates on soaring AI chip demand
Samsung Electronics extended its record run with second-quarter operating profit topping analysts' estimates.
“Samsung said it was seeing exponential growth in demand for general-purpose computing alongside AI, and was closely monitoring the relative pace of demand for HBM and server DRAM while maintaining an optimal product mix to support long-term AI demand.
The company expects industry supply constraints to persist next year and said it will manage its HBM and DRAM businesses in a balanced manner to achieve an HBM market share in line with its conventional DRAM business.
For the second half of the year, the company expects robust demand for memory centered on servers and continuing AI infrastructure capex amid a broader adoption of agentic AI.
It also expects to see accelerating demand for server DRAM, eSSDs, and HBM. Enterprise SSDs are high-performance storage devices designed to deliver data quickly and are essential to AI infrastructure.
Samsung said it expects supply constraints to tighten further in 2027, with rapidly growing AI token generation driving exponential demand over the medium to long term. The company said more customers are seeking multi-year supply agreements to secure AI infrastructure capacity, giving Samsung a better visibility into future demand and greater flexibility in planning investments.
The South Korean tech giant added it has finalized agreements with top five global data center customers and is in the final stages of talks with five additional major accounts to support their AI related demand.“
gleepskip
macrumors 6502a
The good thing is it doesn't matter how much the detractors hate it, AI will benefit them in the end anyway. They may cry they whole time getting there, but when they see the result, they'll embrace the new world.
Are you honestly not concerned when you see stories about an AI model going rogue and hacking others? Are you that naive? Or are you so wrapped up in the FOMO that you don't give a ****?
The idea that "AI" is going to be a world-changing panacea is completely misguided. It's a tool. Nothing more.