The boss of NVIDIA is hallucinating. Jen-Hsun Huang, CEO of the company, admires Elon Musk’s speed in installing artificial intelligence servers for your new project. All this, yes, in a completely different context from the way Apple is focusing in this area with its imminent Apple Intelligence.
The context. Elon Musk was a fundamental part of OpenAI, one of the undisputed leaders of the recent rise of generative artificial intelligence. However, she abandoned the project shortly after its birth and, given the current panorama, decided to create xAI in 2023. This company was born with the firm objective of leading a sector in which, in addition to the aforementioned OpenAI, d Other companies are already fully involved. big tech like Apple, Google and Meta.
What is NVIDIA doing in all this?. Servers and more servers. The training and processing of AI requires high resources and for this NVIDIA has established itself as another of today’s strong companies (it even managed to be the second most valued on the stock market after Apple). What the company offers are very powerful GPUs like the H100, which are the ones that Elon Musk commissioned to set up the xAI servers.
It’s a process that takes years. Typically, setting up a field of servers like the one Musk set up for xAI would take years. At least that’s what one of the experts on the subject says, who is none other than the CEO of Nvidia. Jen-Hsun Huang says at least three years of planning are necessary and an additional one to assemble the equipment that will form the data center.
And it’s not just because of the installation of graphics cards like the NVIDIA H100, but also because of the air conditioning of the place. Large industrial warehouses are needed with optimal conditions in areas such as safety and, above all, refrigeration. Not to mention that each data center requires a second like backup.
“Only one person could do it”Hsun Huang said. This is Elon Musk, to whom he gives immense credit for being able to set up the xAI data processing center in just 19 days. This is something that has never been done before, let alone to the extent that Tesla’s CEO has also done it: over 100,000 units of NVIDIA H100.
Apple does it differently. It’s neither better nor worse. Apple has simply adopted this old slogan. Think differently to concentrate the majority of its artificial intelligence on the power of the processors that make up its iPhone, iPad and Mac.
The M series chips as well as the A17 Pro as well as the recent A18 and A18 Pro are capable of performing local processing of queries which will be generated with Apple Intelligence. This saves the company a lot of work and money when setting up server farms. In the process, privacy prevails by “staying all at home”.
Apple is not cheap. In the fine print of Apple Intelligence, we see that there is more than just local processing. Without specifying to what extent they act, the company already explained during WWDC 2024 that part of this AI also requires its own servers. Well protected and with encryption that guarantees the privacy of users, but nevertheless of the servers.
In fact, we know that Apple is also turning to NVIDIA with around 3,000 units of the H100. And in addition to what it cost the company to set up, experts speak of an annual maintenance cost that runs into billions of dollars.
All is not rosy at Apple Intelligence. And no, we’re not talking about how long it will take to reach countries like USA, which too. We are talking more about the sacrifices represented by this hybrid AI system from Apple, favoring local processing. This will mean that no matter how powerful the iPhone and company’s SoC are, they will have less ambitious features.
This is particularly visible in sections such as image or text generation. In fact, Apple does not have the latter and although they have published studies in which they clarify their position of not trusting too much in the logic of this type of model, the truth is that if at some point this improves and they want to implement something native, they will have no choice but to significantly expand their processing centers.
By | Xataka
In Applesfera | Apple Intelligence, first contact: this is how the incredible AI that takes Siri to another level works
In Applesfera | Ideogram AI: how to use the viral artificial intelligence that creates names with 3D designs, from your iPhone and for free
