Apple couldn’t even justify building an M2 Extreme chip for Mac Pro. They aren’t entering the GPU market when they don’t even sell CPU. Nvidia CUDA has a 20-year head start and the hardware is extremely powerful.
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.
iOS 18: AI Server Industry Aiming to Win Business From Apple
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple couldn’t even justify building an M2 Extreme chip for Mac Pro. They aren’t entering the GPU market when they don’t even sell CPU. Nvidia CUDA has a 20-year head start and the hardware is extremely powerful.
LOL, yeah - sell em some AI servers. The best ones, of course. Equipped with NVidia A100 😂😂😂😂
Nvidia, you know. The same company Apple refused to sign their macOS GPU drivers for a decade.
Nvidia, you know. The same company Apple refused to sign their macOS GPU drivers for a decade.
Refined and not sub-standard.... I know you're not referring to Siri, right? 🤣
You mean enthusiasts get bored. The average consumer is just fine
nah. I know a lot of average consumers that switched to samsungs this year. Usually comes down to iPhones all looking the same and doing the same things they’ve always done. Other people keep em til they die at this point seeing no need to upgrade. I personally am an enthusiast and im still frustrated how long it takes to get a feature that was rumored years before or that Samsung had forever like their zooms. Apple really doesn’t have much excuse for being so far behind the industry with AI
I think without a doubt 100% Apple Silicon for its needs. Besides Nvidia chips are not as efficient as Apples and are for the cloud mostly not for AI on the Edge which is where AI is headed
No one in the Cloud AI does it but on device AI or Edge AI is a completely different AI Animal and perfect for Apple
For initial raw power nvidia chips exceed Apple but in efficiency no way! And in the long run its all about efficiency.
For training, yeah. Good point. I was purely thinking of the end user experience
Is Apple really behind? Google just released a laughable LLM that is so biased it thinks that certain people do not exist! Also, everything Microsoft does with AI comes from Open AI! And Microsoft just back stabbed Open AI by investing in the Mistral LLM.
all companies / businesses depend on other companies / businesses.
no business is 100% independent / non-reliant on another business.
Apple has repeatedly said that they want to own/control the core technologies for their business. Their development of AppleSilicon (which many, many people said would NEVER happen because how could lowly little Apple compete with mighty Intel) is a prime example of that.
If AI is going to become core to Apple's business (and it is), then Apple should own/control it. Also, Nvidia makes 70% margins on their GPUs. If apple was unwilling to pay Intel their 60% margins, they shouldn't be paying Nvidia their 70% margins.
Apple builds GPUs and NPUs right now, today. They have the software via metal, coreML, etc right now, today. If Apple is going to spend literally billions of dollars on servers for AI training, then they ought to be building those servers themselves, using their own silicon and software. And then they can make those computational resources available to developers and end-users. It would be kind of crazy for Apple not to do this. It's so totally within their wheelhouse -- much more so than music, movies, or tv shows.
CUDA outperforms Apple GPU by a large margin. Apple will bring it in-house only if they can do it better. Qualcomm is a good example. AI is a race and Apple is already two years behind. They’re not going to develop a new GPU just for in house use, hope to match Nvidia, and risk falling behind.
We already know the current stuff Apple uses is Nvidia or else Supermicro wouldn’t be soliciting. We would be hearing something from TSMC if Apple were doing anything in house.
Bruh...he left leather jacket for finewoven jacket
Anecdote. Got any actual evidence of this?
Tech progress is actually slowing down quite a bit for everyone. Hence the more and more ridiculous prices and stagnation. But yes samsung manages to still find some ways to add those little fun features.
I don’t think it’s long before climate collapse now so this iPhone 15 pro max will probably be my last hedonistic splurge to have some fun and spend it all before inevitable loss of geopolitical stability and nukes.
Do you think Apple uses Linux or Windows based systems? Since they refuse to sign Nvidia drivers, I guess there are no drivers available. Maybe Apple already coded a CUDA layer? Sometimes fate hits you very hard, Apple - but i can't feel sorry for you.
Stats show more people are switching to iPhones vs android so yeah.
I’m ok with apple taking longer. Because like every android OEM they release something first half baked and ironically only better when apple does.
You use a different product for running user-facing AI than for training the AI.
Training is FAR MORE computationally expensive than running in the "forward mode". Even large AI models could be run on a big enough home computer. But you need a very expensive server room to train the model.
So I imagine Apple will have customer-facing data centers and developer-facing centers and they will use different hardware. It is hard to see which data centers will need to be larger.
For training, you need the "real GPU" that can do at least 32-bit floating point math. But for running the trained model we can use much more efficient 16-bit floating point or even 16-bit or even 8-bits integers.
So,... Apple will buy a range of different Harvard. But all of it will likely run on Intel servers running Linux. But then ARM is moving into this spec. There are some very impressive ARM server ships. And then to that you add the Nvidia GPS or TPUs.
BTW, I really wish Apple would use those ARM server chips in the Mac Pro. The MP has poor performance compared to iter top-of the line ARM-based systems.
No one uses Window for this. Not even Microsoft.
yes, you are right
I‘m using a Mac ….to log into my Linux workstation fur such tasks 😂😂😂😂
It has been a few years since Apple discontinued the XServe; they should consider going back into the server business and release a first-party AI solution. Apple has their own top-of-the-line chips with the M series
AI server makers are hoping to obtain orders from Apple ahead of its highly anticipated unveiling of new AI features later this year, according to Taiwanese research firm TrendForce.
Server manufacturer Supermicro is reportedly "aggressively expanding" its efforts to secure AI-related orders from Apple. The company has the potential to double its shipments of AI servers this year, supported by orders from existing clients including CoreWeave and Tesla.
Late last year, Apple analyst Ming-Chi Kuo reported that Apple is expected to spend around $4.75 billion on servers to support its AI technology in 2024, up from $620 million in 2023. The company could purchase up to 20,000 servers this year.
Apple is widely expected to announce a range of new AI features at WWDC later this year as part of the preview of iOS 18 and its other software updates. At its annual shareholder meeting yesterday, Apple CEO Tim Cook said that the company will "break new ground" in generative AI in 2024. "We believe it will unlock transformative opportunities for our users," he added.
Article Link: iOS 18: AI Server Industry Aiming to Win Business From Apple
Apple could give the industry a run for its money
