Demand for AI infrastructure shows no signs of slowing down: News Street
According to New Street Research, artificial intelligence infrastructure has no signs of slowing down as more spending on computing power leads to better performance models.
It is like a self-perpetuating system.
"OpenAI has shown how model size puts significant pressure on inference costs," said Pierre Ferragu, an analyst at New Street Research, in an industry report.
"As the models mature, inference costs decrease very rapidly, but the next generation models will be significantly more expensive than the previous versions (which only worsens with the introduction of inference functionality)," he added.
This cycle ensures high demand and sales for companies that provide hardware to power AI, but raises questions about whether the increasing cost of next-level models is worth it.
"Generative AI is a business (latest model) with terrifyingly high incremental costs," Ferrag pointed out. "This is the complete opposite of software, which has zero incremental costs. It will bring serious disruption to the structure of the value chain, although it is uncertain how it will unfold."
Companies such as Meta Platforms (NASDAQ:META) and xAI, a startup founded by Tesla (TSLA) CEO Elon Musk, continue to raise the stakes in the AI hardware race. Last week, Musk announced an AI training cluster with 0.1 million units of Nvidia's H100 GPU. He plans to double that in the coming months. Similarly, Meta CEO Mark Zuckerberg plans to deploy 0.35 million units of H100 in the company's AI efforts by the end of the year.
"As we deepen our understanding of the economics of AI on both the supply and demand sides, we see a strong case for the sustainability of current spending levels and further material growth," said Ferrag. "This does not mean there won't be pauses or corrections, but it does mean we are far from the peak of an unsustainable bubble."
Nvidia is currently winning the competition to provide the processors needed to power a range of evolving AI models. The main reason is Nvidia's latest GPU's full-stack and system-scale integration capabilities, enabling larger interconnect bandwidth between larger domains.
According to New Street Research, this is what makes Nvidia superior to its competitor AMD (NASDAQ:AMD).
"However, its architecture is less versatile and scalable, but it may still be competitive, albeit with a much more limited use case set, without sacrificing dedicated optimization efforts that would disqualify AMD," added Ferrag.
However, AMD understands this and is working to challenge Nvidia. This is one reason why AMD recently acquired ZT Systems.
Details about Advanced Micro Devices, Broadcom, etc.
Nvidia: Blackwell Delay and Its Consequences
Nvidia Q2 provides some clues for sustainable future growth.
The power of expectation: Nvidia's earnings and market reaction.
Today's biggest stock movers: AVGO, PATH, etc.
According to analysts, the post-revenue drop of Broadcom provides a buying opportunity.
Disclaimer: Community is offered by Moomoo Technologies Inc. and is for educational purposes only.
Read more
Comment
Sign in to post a comment
Kimihiko OP : Comments on signs of slowdown are from blackrock and JP Morgan, two buy-side companies.
コレ君 : Somehow, the current situation still feels like a startup. It's still early days for NVIDIA!...