Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Ain’t that the truth,looked at a 128Gig M5 Max and was like, nope can’t even remotely justify this, I can’t write it off, and I can’t bite on 5K for a laptop 🫠
The Studio works out a little cheaper than the MacBook, but it’s still a considered purchase. I had to steal myself when I placed the order last year, but I’m glad I did.
 
Right now companies are pouring billions and billions into AI (mostly In the construction of data centers, but also R&D), but not getting any money out of it.
In not too many years, AI hardware and software might improve to the point where the amount of data center resources needed (land, power, etc.) will shrink to the point where much smaller data centers will do the job, and much less space needed in the huge ones, at which point some of the space in the huge ones might be repurposed through adaptive use for warehouses, indoor farming (they already have climate control systems, water coming in), etc. Some local Planning Commissions are headed in this direction, trying to get the companies that want to build big data centers in their communities to agree to clauses that treat the buildings like local infrastructure that can be used for other purposes if the data centers reduce their use of space in them or leave altogether. Maybe that'll be a route to AI companies getting some of their money back, but the numbers are speculative.

It's also possible (likely?) that they'll just fill the space freed up by removing their older hardware, with the newer hardware.
 
Last edited:
  • Haha
Reactions: Agent007
The problem with AI has nothing to do with its capabilities (or lack thereof), but the economics of it. Right now companies are pouring billions and billions into AI (mostly In the construction of data centers, but also R&D), but not getting any money out of it. Google, OpenAI, Anthropic, Microsoft, etc. are desperately trying to monetize runs, but thus far have failed to do so. Eventually all of these companies are going to realize that there isn’t enough money in AI and drastically scale things back - at which point the bubble will burst.

AI isn’t just going to go away at this point, but sooner than later the hype train around it will fade and fall off a cliff.

CapEx is high, because the tech is new. Real engineers are starting to hit the AI field, and requirements are going to be less and less. Those data centers will still be built, but you'll be able to run even more stuff on it.

If you look at those companies, their annualized AI revenue is already in the billions (for some of them). Will they be able to keep those revenue streams? Well, I'm sure they have AIs working on that.
 
As a note, today I probably saved my wife's companies $35-50k a month after a random chat with Gemini about energy storage. Apparently there are renewable energy offset programs that "nobody" knows about in her home country. Makes the $10 gemini pro subscription through Verizon worth it for sure.
 
The problem with AI has nothing to do with its capabilities (or lack thereof), but the economics of it. Right now companies are pouring billions and billions into AI (mostly In the construction of data centers, but also R&D), but not getting any money out of it. Google, OpenAI, Anthropic, Microsoft, etc. are desperately trying to monetize runs, but thus far have failed to do so. Eventually all of these companies are going to realize that there isn’t enough money in AI and drastically scale things back - at which point the bubble will burst.

AI isn’t just going to go away at this point, but sooner than later the hype train around it will fade and fall off a cliff.

Lmao the runs have been heavily monetized. Profitability hasn’t been reached yet but the models have been monetised and revenues are shooting up. Anthropic has more than 10x its revenue each of the three years it has existed as a company.

They are now up to a $19bn ARR (almost doubled from the ARR at the end of 2025) and at the last update expected to reach break-even in 2028. Also, it is known in the industry that the margins on serving models (inference) is high.

This idea that is no revenue or no business model even is wrong and really misleads people as to what to expect in the future.

Model capabilities continue to increase, with both OpenAI and Anthropic making plain that they see no wall and expect the models to get dramatically better this year.

People need to start dealing with the implications of this technology and stop consoling themselves with fantasies where it “falls off a cliff” and becomes some kind of triviality or irrelevance.
 
There seem to be an awful lot of YouTubers clustering these things. Maybe Apple have just run out of stock, lol.

Seems odd that those YouTubers were talking about some new technology to have these things connected together, with Studios provided by Apple to show off the new tech, with 2TB combined RAM, and now you can’t buy one.
Yes, Apple lent these YouTubers the Studio Ultra clusters so they could demonstrate how they could run big models locally using the new RDMA over Thunderbolt 5 feature in macOS Tahoe 26.2. So, Apple is very interested in exploiting this (accidental to them) market.

Funny thing is that not all the Studios Apple lent out were 512GB models. For example, one test, two Ultras had 512GB, and the other two had 256GB.
 
  • Haha
Reactions: rezwits
That would be devastating. But if that is the case, how can they ship lesser memory configurations at all — is it a question of chip size?

Yes. Apple uses LPDDR5X memory chips and those chips would need to meet Apple's specifications so they are probably not "commodity" parts. My guess is they are using four 128GB modules for the Ultra and that size of module is probably very much in-demand right now, which is both raising the price for Apple and constraining available stock to purchase.

Is this a speculation or a fact? (source?)

Speculation.
 
Last edited:
  • Like
Reactions: jole
May I just say that this sucks because I've always loved maxing out computers and adding them to the cart, wondering what I'd do if I won the lottery...

Also, for those of you using AI in any way, shape or form, I hope you get better soon because you're suffering from some rather extreme delusions. Stop killing the earth; this **** can't do basic arithmetic. The more intelligent you are, the less you're into AI. Because it sucks. And it obviously sucks.

Anyway. If I won the powerball, I'd max out four studios, plug them all together, surrounded by 8 32" XDR displays. I don't know what I'd do with it other than sit back and admire the sheer power of it. Maybe have you guys over to run your LLMs and we could sit back, relax, and enjoy the hallucinations.
 
AI needs to crawl up its own ass and die. While I'm terrified for what the inevitable AI bubble burst will mean for the economy (and for the thousands that will lose their jobs), it really can't happen soon enough.
It’s not thousands, it’s not even millions, billions will lose their jobs to AI worldwide. The 1% will use it to eliminate costs and pay executives and shareholders more and more money widening the wealth gap to an insurmountable point where we enter a completely different world.
 
May I just say that this sucks because I've always loved maxing out computers and adding them to the cart, wondering what I'd do if I won the lottery...

Also, for those of you using AI in any way, shape or form, I hope you get better soon because you're suffering from some rather extreme delusions. Stop killing the earth; this **** can't do basic arithmetic. The more intelligent you are, the less you're into AI. Because it sucks. And it obviously sucks.

Anyway. If I won the powerball, I'd max out four studios, plug them all together, surrounded by 8 32" XDR displays. I don't know what I'd do with it other than sit back and admire the sheer power of it. Maybe have you guys over to run your LLMs and we could sit back, relax, and enjoy the hallucinations.
I feel sorry for people like this. You're going to get left behind.

My coworker, who only knows a little bit of python (from 20 years ago), built a badly needed dashboard for our org to track professional development opportunities. They used Claude Code and built it in a day. It would not have been possible otherwise with generative AI. That one dashboard made life better for everyone at work, and my coworker's job easier. Is it perfect? No. Is it functional and useful? Absolutely.
 
Last edited:
Surprised to see this move. Not sure on when the next version will be released but definitely further price hikes are possible when it is launched.
 
  • Like
Reactions: mganu
That makes no sense.
Why prevent people from buying ?
Just raise the price to whatever is necessary.
They did raise the price for 2 days then removed it.

"Due to the supply shortage and more demand of the lower unified memory variants our R&D team has removed the 512Gb unified memory option from Apple Store. As a result, the Mac Studio now tops out at 256GB of unified memory, and the remaining memory upgrades have also increased in price. This change affects all current configurations and at this moment we do not have any additional information when it will be available again for sale at the Apple Store."

If they have supply of those 128GB chips they are using them internally either for their own servers on apple silicon or move to M5 Ultra production. Or maybe we are getting M4 Ultra and not M5. Or it has just dried up and just not worth bothering with if they only are selling 10 a week or whatever.
 
Last edited:
Anthropic is killing it:

Key financial details earlier in 2026 from their funding round:
  • Revenue Growth: Revenue has grown more than 10x annually over the past three years.
  • Enterprise Focus: Over 500 customers spend more than $1 million annually on Claude.
  • Key Products: The Claude Code tool is driving significant growth, with a run-rate exceeding $2.5 billion.
  • Projections: Anthropic aims for positive cash flow by 2028. View attachment 2610539
Trajectory:
  • January 2025: ~ $1B
  • May 2025: ~ $3B
  • Mid-2025 (June/July): ~ $4B
  • August 2025: > $5B
  • October 2025: ~ $7B
  • End of 2025 (December): > $9B
  • February 2026: ~ $14B
  • March 2026: nearing $20B (~$19–20B reported) — I attribute this $5B uplift to Pete Hegseth 🙂

With this, I would guess that Anthropic is way ahead of the game in getting to black due to them planning costs with a significantly lower revenue estimate.

OpenAI might be in pickle through...

This has all the hallmarks of being written, at least in part, by AI.
 
  • Like
Reactions: AbeLincoln1984
I feel sorry for people like this. You're going to get left behind.

My coworker, who only knows a little bit of python (from 20 years ago), built a badly needed dashboard for our org to track professional development opportunities. They used Claude Code and built it in a day. It would not have been possible otherwise with generative AI. That one dashboard made life better for everyone at work, and my coworker's job easier. Is it perfect? No. Is it functional and useful? Absolutely.

Recent Windows critical updates have been partially written by AI, and look at all the rollbacks and crashes people have been having.

AI is like any other tool. It's only as good as the person wielding it, and if the person is either a) to lazy to enumerate the accuracy of the code themselves, of b) doesn't have basic programming skills, the result is going to be AI slop. From observations, this pretty much sums up "vibe coding".
 
Recent Windows critical updates have been partially written by AI, and look at all the rollbacks and crashes people have been having.

AI is like any other tool. It's only as good as the person wielding it, and if the person is either a) to lazy to enumerate the accuracy of the code themselves, of b) doesn't have basic programming skills, the result is going to be AI slop. From observations, this pretty much sums up "vibe coding".
Thank you and agreed.

When people take a hard look at vibe coding, it's not really helping.

I have no problem being left behind of the AI disaster. Pretty please, leave me behind.

It's always worth asking what something is actually good at; a lot of the AI enthusiasm seems to focus on incredible potential that's always just out of reach. But a rigorous evaluation of anything also asks where something breaks down, what it's not good at.

There's plenty of specific applications where machine learning can be a wonderful optimization tool. But people seem to want these omniscient, omnipotent chatbots that can take the average of all human data and spit out something novel and wise; that's foolish.

But if we're going to tinker with AI, we should absolutely do it in the comfort and privacy of our own homes. I have no problem with such projects. Just don't be delusional about silicon Jesus coming to save humanity or whatever.
 
  • Like
Reactions: dricci
Recent Windows critical updates have been partially written by AI, and look at all the rollbacks and crashes people have been having.

AI is like any other tool. It's only as good as the person wielding it, and if the person is either a) to lazy to enumerate the accuracy of the code themselves, of b) doesn't have basic programming skills, the result is going to be AI slop. From observations, this pretty much sums up "vibe coding".

You can find negative examples anywhere. Are you choosing to ignore the positive examples? Effectively 100% of Anthropic's code is now written by Claude code.

The dashboard my colleague made isn't slop. It does the job it needs to do. It's an internal tool only. We're not shipping it to anyone else. My colleague is an educator, not a software developer. But they are highly intelligent, and have a builder's mindset. This is a project that would have taken weeks or maybe even months without generative AI, and thus would have never come to fruition. This is the worst AI they'll ever use.
 
  • Like
Reactions: Schrödinger’s
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.