to leave a comment.
The Information reports Apple targets AI inference with M8 Ultra cluster...aims for 2029 launch
Apple is considering re-entering the enterprise server market after 18 years and is discussing ways to adopt server connection technology with Nvidia, IT media outlet The Information reported on the 16th (local time), citing sources.
According to the report, Apple has been discussing building its own M8 Ultra chip-based server clusters using Nvidia's open server connection technology, 'NVLink Fusion'.
The clusters, configured with two or four chips, are targeting the artificial intelligence (AI) inference market, with a target launch time of 2029.
The plan is in its early stages and could fall through, the media outlet reported.
If Apple returns to the server market, it will be approximately 18 years since it discontinued its server product line, 'Xserve', in 2011. At the time, project participants said it largely failed due to a lack of interest in the enterprise server market and insufficient customer support.
NVLink Fusion is an open interconnect technology unveiled by Nvidia in May last year, which allows even non-Nvidia chips to connect between servers through Nvidia's networking technology.
Amazon Web Services (AWS) decided to adopt this technology in December last year and announced last month that it would expand its application.
Dion Harris, Senior Director of High-Performance Computing and AI Infrastructure at Nvidia, explained that the company's networking equipment strategy is expanding from self-production to open collaboration.
Discussions about this collaboration come amid observations that the previously strained relationship between Apple and Nvidia has recently entered a thawing phase.
The two companies had a strained relationship, stemming from a conflict in 2008 over sharing after-sales service costs due to graphics card defects, to the point where Nvidia graphics processing units (GPUs) disappeared from Mac products after 2019.
It is said that the animosity was so deep that there is an anecdote about Steve Jobs treating an Nvidia executive as if he wasn't there during a meeting at the time.
However, it is reported that an opportunity for improving relations was created when Apple decided to utilize Google Cloud's Nvidia Blackwell chips during the revamp of its voice assistant Siri, which was unveiled in June.
Apple's consideration of re-entering the server market is also linked to the expansion of its own hardware business due to a surge in demand for AI computation.
Apple's Mac sales in the third quarter of the last fiscal year increased by 29% year-on-year to $10.4 billion (approximately 14.3 trillion won), and this is analyzed to have been influenced by increased demand for Mac Studio and Mac mini for AI inference.
The Information pointed out that if Apple wants to strengthen its competitiveness in the enterprise server market in the future, it will need to hire more personnel and provide software tools like Nvidia's CUDA.
Awni Hannun, the researcher who led Apple's on-device AI framework 'MLX', recently left Apple to join Anthropic.
OpenAI is also reportedly developing its own AI chip, 'Jalapeno', indicating that the competition among big tech companies to build their own AI infrastructure is heating up.
Along with this, it is ironic that Taiwan's Foxconn, which has been entrusted with iPhone contract manufacturing by Apple, is poised to become a potential competitor to Apple with its AI server business, which has grown larger than the iPhone itself.
Foxconn is already Apple's largest hardware partner and the world's number one AI server manufacturer. If Apple enters the self-developed AI server market, it might again entrust assembly to Foxconn, potentially leading to a scenario where Foxconn's AI servers for other big tech companies compete with Apple's.
Newsletter
Get key news delivered to your email every morning
to leave a comment.