No worry. The Machines has access to over 7 billion people as a power source.😱
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.
Kuo: Apple Could Spend $4.75 Billion on AI Servers in 2024
- Thread starter MacRumors
- Start date
- Sort by reaction score
Oh wait I was given an outdated article from 2018. Never mind I'm erasing that last bit
I think Kuo needs to understand how much $1 billion dollars US is, or maybe he's talking about HK$ & forgot to put the HK in front 🙄
Apple is expected to spend several billion on hardware to support its artificial intelligence development in 2024, according to speculation from Apple analyst Ming-Chi Kuo. Kuo expects Apple to spend "at least" $620 million on servers in 2023 and $4.75 billion on servers in 2024.
Apple could purchase between 2,000 and 3,000 servers this year, and up to 20,000 next year. Kuo thinks that Apple is purchasing servers equipped with Nvidia's HGX H100 8-GPU for generative AI training, with the company planning to upgrade to B100 next year. Nvidia calls its H100 an AI supercomputing platform, and each one is priced at around $250,000.
Kuo appears to be guessing at Apple's purchasing plans here, and he says that he expects Apple will use AI servers it is purchasing and installing itself to train large language models rather than virtual hosting from other cloud service providers for improved security and privacy. He does say that Apple could develop its own server chips to save on server costs, but he has seen no evidence that Apple is doing that at this time.
While Apple appears to be making a major investment into AI, Apple's server purchasing will fall behind other companies like Meta and Microsoft. Apple will also need to invest in labor costs, infrastructure, and more, and Kuo suggests that Apple will need to spend several billion dollars annually to have a chance of catching up with competitors. Kuo claims that he is "genuinely concerned" about the future of Apple's generative AI business if Apple spends just a billion dollars a year as suggested by Bloomberg's Mark Gurman.
Over the weekend, Gurman said that Apple is on track to spend $1 billion per year on its AI efforts. Gurman says that Apple is working on a new, smarter version of Siri and is aiming to integrate AI into many Apple apps.
Article Link: Kuo: Apple Could Spend $4.75 Billion on AI Servers in 2024
Better hope China does not invade Taiwan - where do you think Nvidia fabricates all these chips at? With the AI-boom, TSMC is even more a critical supplier than before. I'm glad to see several chip fabs are being built in the US again.. but it will take years to rival TSMC's super fabs sheer scale and wafers per day output.
But this articles say they are spending Billions on the hardware alone. That does not include the cost to port all those applications to use the new information source from the hardware. Those are 2 very different things. If Apple spends 4.8 Billion to buy the hardware...it will cost another billion maybe in development costs to make it all work with the applications you listed. This is a very delicate and time consumer effort. There is testing...then dev...then QA then UAT.
It's a load of rubbish speculation for clicks.
Apple is expected to spend several billion on hardware to support its artificial intelligence development in 2024, according to speculation from Apple analyst Ming-Chi Kuo. Kuo expects Apple to spend "at least" $620 million on servers in 2023 and $4.75 billion on servers in 2024.
Apple could purchase between 2,000 and 3,000 servers this year, and up to 20,000 next year. Kuo thinks that Apple is purchasing servers equipped with Nvidia's HGX H100 8-GPU for generative AI training, with the company planning to upgrade to B100 next year. Nvidia calls its H100 an AI supercomputing platform, and each one is priced at around $250,000.
Kuo appears to be guessing at Apple's purchasing plans here, and he says that he expects Apple will use AI servers it is purchasing and installing itself to train large language models rather than virtual hosting from other cloud service providers for improved security and privacy. He does say that Apple could develop its own server chips to save on server costs, but he has seen no evidence that Apple is doing that at this time.
While Apple appears to be making a major investment into AI, Apple's server purchasing will fall behind other companies like Meta and Microsoft. Apple will also need to invest in labor costs, infrastructure, and more, and Kuo suggests that Apple will need to spend several billion dollars annually to have a chance of catching up with competitors. Kuo claims that he is "genuinely concerned" about the future of Apple's generative AI business if Apple spends just a billion dollars a year as suggested by Bloomberg's Mark Gurman.
Over the weekend, Gurman said that Apple is on track to spend $1 billion per year on its AI efforts. Gurman says that Apple is working on a new, smarter version of Siri and is aiming to integrate AI into many Apple apps.
Article Link: Kuo: Apple Could Spend $4.75 Billion on AI Servers in 2024
What's really amusing is that most people who are desperate for AI, don't realise it's going to replace them 🤣
It comes down to the value of the tie-in. Apple is all about their ecosystem and if their AI glues everything together, that's their main selling point.
Apple is already spending $1B on car with nothing to show for it.
Microsoft will spend 25 billion.
Meta will spend more them 10 billion.
If apple manages to bring AI to the masses, they will prevail once again.
Meta will spend more them 10 billion.
If apple manages to bring AI to the masses, they will prevail once again.
Apple has been softening on it's Nvidia stance for the last year or so. One of the interviews when the Mac Pro was updated for Apple Silicon literally praised Nvidia as the best platform for creating AI models. I wouldn't be surprised if we see Nvidia return to the Mac Pro for specialty compute and AI workloads. The bad blood between Apple and Nvidia may finally be dissolving. I wouldn't be surprised if we see Apple produce Nvidia GPUs on their reserved TSMC capacity if there are supply chain constraints.
One thing to note is that Microsoft hosts a service to build AI models in the cloud. Apple wouldn't need nearly as many as Microsoft since they would be unlikely to offer it as a service for others.
One thing to note is that Microsoft hosts a service to build AI models in the cloud. Apple wouldn't need nearly as many as Microsoft since they would be unlikely to offer it as a service for others.
Last edited:
That’s what everyone is working on, because less resource consumption = more profit. It’s rather difficult though.
Top end NVIDIA DGX H100 servers cost $500k each. A billion dollars doesn't go far and I think Kuo's prediction is closer to the mark.
And I bet Apple is absolutely thrilled having to pay them that much money after their long drawn out spats over GPUs in Macs and many other things.
AMD has basically no development platform for AI. NVIDIA has a 5+ year head start (conservative estimate) with CUDA.
Apple isn't going to reinvent the wheel when time is key. It would be like using RISC-V instead of ARM.
That’s one way to look at it. Very optimistic. Not untrue but I’m not sure how much it helps. That was the Unix philosophy, keep it tight and focused in every way. I hope that still endures but it’s definitely being discouraged by modern software companies.
How far we have come in terms of how much power it takes to do the same thing. Or should I say how far we’ve fallen. From simple text to graphics to video and now ten times the electricity consumption to do the same things.
100% This. MS365 and Azure are MS' fastest growing business units.
Imagine leveraging all those binned M1/M2 SoCs that don't have enough RAM or CPU cores to make it into a Mac or iPhone. Stick them on a custom logic board that supports dozens at once. Network them together and you could have tens of thousands of them in a rack and stay within a reasonable power and cooling envelope.
These would be extremely affordable, power-efficient, and capable.
These would be extremely affordable, power-efficient, and capable.
Last edited:
Have you actually used one? It doesn't sound like you're very familiar since you said generative which is pretty broad instead of LLM which is what this will most likely be using if it will be similar to chatGPT. They are getting better every month and they are not going away. Generative also encompasses image and audio generation, and I'm not sure what there is to not trust with that? You tell it what you want it to make and it makes it.
For me, chatGPT has really sped up my workflow as I can have it usually write 3/4 of a script and then I can finish the rest. For easy scripts it can just do the whole thing. I check over the code and test it. This ranges from something like a performant find script to run across all sites on our server, to cracking out a WordPress plugin in record time. Sometimes it makes up a weird answer, but I notice that happening less frequently, and I've been correcting it as I go. It also depends on how you talk to it. It's strange, but it's kind of like a way of programming using language. If I frame things within scopes and explain what my end goal is clearly, it tends to do better. What's also odd is I you tell it something like "take a deep breath and try again" it refocuses and often comes up with correct responses. Very strange, and we're still figuring it all out. It's the Wild West right now!
There is no comparison to NFTs, which had zero value whatsoever. I was against NFTs from the start. But this? This is extremely valuable. In the future, something like this could replace an operating system. Not joking. You could literally have it spin up an OS and customize it on the fly based on your preferences and have it connect to other services online to do whatever you want. The future is gonna be absolutely wild. We're transitioning from fixed code to code that is reactive and flexible and alive and responds to our every desire. It's an extremely exciting time to be alive! This is a new era of computing, bigger than the invention one the internet, and will have a much larger impact on society. It is up to us to understand it, and help shape and steer it and laws/regulations to make sure that it is controlled and does more good than harm. It's scary but lately I've been leaning towards being more excited than scared.
Every new major breakthrough in tech can be scary, especially to older people, and I'm nearly including myself in that category as I'm in my late 30s. But we have to realize that these are extremely early days and the best is yet to come. Things we can barely imagine will be possible in a year or two, and things that we can't even comprehend will be possible 5-10 years from now. What we are entering is the most rapid development period in all of human history by orders of magnitude. The things that we will be able to put this to use to discover are unimaginable. We're already seeing some of the early applications in science. The ability to ingest huge amounts of data and see connections and patterns that researchers miss is amazing. Everything from materials science, to discovering new types of drugs, to figuring out how to make stable fusion reactors, to a mass increase in exoplanet discovery and potential life in other systems.
Smaller models also execute faster. I'm sure everyone is trying for that, however there are probably still differences when optimizing to run on devices vs running in the cloud. Apple is likely the only one to really push running on device.
Which is why AMD is buying Nod.ai, so they will have a software developer ecosystem.
AMD to Acquire Open-Source AI Software Expert Nod.ai
— Nod.ai to accelerate deployment of optimized AI solutions on AMD high-performance platforms and bolster AMD open-source software strategy ꟷ
It wouldn't be capable at all. AI training is massively parallel, something only GPUs offer.
M1/M2 basically has such a low level of graphics capability, it's around 100x slower compared to H100.
No by TechRepublic and Vox. After JPack replied I asked the Bing ChatGPT AI and the funny thing is that AI was what corrected my mistake pulling up more recent numbers.
I love that AI so much. It's made researching context so easy and has allowed me to fully ditch Google for Bing.
You saw Mother Nature giving Timmy's crew a hard time in their Apple event promo. She will he hopping mad if those servers aren't driven by good old Californian sunshine.
Eat, drink, and be merry for tomorrow we die. Eh?