Large ser sales: Apple before AI mega order at Nvidia?

Apple apparently plans a billion dollar investment in more and newer AI technology for Apple Intelligence. The supplier is said to be the GPU specialist Nvidia. This is reported by the analyst Ananda Baruah from the American investment company Loop Capital, citing sources in the supply chain. Accordingly, it is about an order volume of around $ 1 billion. Up to 250 new servers are to be purchased.

AI server specializes in Reasoning

Like Baruah according to the stock market leaf IBD writes in an investment reportit is about Systems of the GB300 NVL72 type. These are systems that are not yet officially on the market and are particularly suitable for Reasoning AI applications. A large voice model opens the work steps over which there was a conclusion. Corresponding models are available from the large providers such as Openai in the USA or Deepseek in China. So far, Apple has not operated a real chat bot and does not yet use Reasoning in products that are accessible to customers.

A GB300-NVL72 system is based on 72 GPUs of the Blackwell Ultra plus type 36 Grace CPU “in a single platform that is optimized for test-time scaling inference”, as Nvidia describes. The systems can also be used to build up so-called AI factories, in which the machines are connected to Quantum-X800-Infiniband or Spectrum-X Ethernet. When using a ConnectX 8 supernics, the inference output of Reasoning models should be able to endure the time compared to the previous Hopper platform.

Large server cluster for gene Ai at Apple

Baruah further announced that Apple is planning to spend between $ 3.7 and $ 4 million per server. In addition to Nvidia, Dell and Super Micro are also part of Apple's suppliers of AI hardware. Apple is now “officially in the game with the Large server clusters in the area of ​​gene AI,” said the analyst. “Gen AI” stands for generative AI systems, i.e. large voice models (Large Language Model, LLM) with and without reasoning and image generators.

Apple drives a double strategy. For example, the company also develops its own AI server on the basis of the M4 to implement its private cloud computation concept, in which AI applications run out of the cloud without the group being able to look inside.