share_log

AI临界点到来!不是英伟达买不起,是AMD更有性价比

The AI tipping point has arrived! It's not that Nvidia can't afford it; it's AMD that's more cost-effective

Zhitong Finance ·  Feb 23 13:51

Source: Zhitong Finance Author: Yu Jing

$NVIDIA (NVDA.US)$After releasing a blowout fourth quarter earnings report last night, investors have repriced their rivals in the chip industry.$Advanced Micro Devices (AMD.US)$Shares rose higher on Thursday, continuing their impressive gains since this year.

Under the leadership of CEO Su Zifeng, AMD has become Nvidia's only real competitor to dominate the global market in the field of artificial intelligence chips with its new MI300X chip.

At the end of last year, the group launched the MI300 series designed to support generative artificial intelligence technology and launched the Instinct M1300A, a next-generation semiconductor focused on supercomputing.

Analysts said the MI300X could challenge Nvidia's H100 graphics processing unit chip, which dominates the large language model artificial intelligence market.

“We believe AMD's MI300 is one of the products most capable of challenging Nvidia,” said Purdy Ho, an analyst at Huatai Research. Notably, Lamini AI has announced the use of the MI300 and plans to use it to train large models.”

He added: “Tech giants such as Microsoft, Meta, and Oracle have also promised to use the MI300 series.”

Last month, AMD said the new chips could generate around $3.5 billion in sales over the next year because of the group's ability to use its new products to counter Nvidia and meet surging global demand.

AMD: Artificial intelligence market could reach $400 billion

AMD CEO said that the AI chip market could reach 400 billion US dollars in the next four years, double what Su Zifeng expected in August last year. This is a larger upgrade to previous predictions.

AMD Chief Technology Officer Mark Papermaster said at an event earlier this week, “When we think about $400 billion, this is an accelerator (potential market).” “You can think of it as an expansion of the Internet when it first launched. You need to do the entire expansion, not only computing, but also the network, infrastructure, etc. It's a new platform.”

However, to meet surging demand, AMD's production capacity is already tight, although Su Zifeng told investors last month that AMD has “made significant progress with supply chain partners and has obtained additional production capacity to support upward demand.”

Although this statement doesn't directly mention its larger competitor, it does highlight one of the major resistances Nvidia has faced recently.

Chip-on-wafer-on-substrate (CHIP-on-wafer-on-substrate, coos for short) is a key component of artificial intelligence chip manufacturing, and it has been hampering Nvidia's delivery for much of the past year.

TSM.US, the world's largest chip contractor, has been struggling to meet the surge in demand for its high-end stacking and packaging technology, and warned investors “this may continue until next year.”

Supply chain issues cast a shadow over Nvidia

Nvidia said it expects its next generation chips, including the higher-priced B100 Blackwell, to be “in short supply” in the short term.

Nvidia CEO Hwang In-hoon told investors on a conference call late Wednesday: “We expect demand to continue to outweigh our supply this year, and we will do our best.” “The cycle is improving and we will continue to do our best.”

Nvidia said it will launch an H20 chip later this year that complies with the new export regulations.

Nvidia's monopoly position in the wider artificial intelligence chip market is unlikely to be challenged by AMD for the next year or more. But as companies (and more and more governments) look to harness the potential of artificial intelligence technology, they'll look for new GPU sources to avoid falling behind other high-paying customers.

This is likely to significantly accelerate AMD's growth rate in the near future.

FOMO could boost AMD sales

“As AMD pushes its MI300x to the cloud provider, we are encouraged by the future of data centers (graphics processing units),” CFRA analyst Angelo Zino said in a recent customer report. “We think expectations for GPUs are conservative, and there is room for improvement in 2024 and 2025 data.”

Zino added that AMD will soon launch new Zen 5 chips for the PC market, enabling them to manage artificial intelligence tasks directly on laptops.

AMD's stock price is also cheaper than Nvidia, with an expected price-earnings ratio of 44 times (Nvidia's expected price-earnings ratio is 57 times), but this also reflects that Nvidia's sales are growing much faster and the overall profit margin is higher.

Despite this, AMD's shares are up 65%, while Nvidia's shares are up 152%. In May of last year, Nvidia released a shocking financial report for the first fiscal quarter, which actually kicked off the global AI investment competition.

John Vinh, analyst at KeyBanc Capital Markets, said, “We strongly believe in AMD's strategy to revive the product roadmap, and its product appeal is convincing.” He rated AMD stock as an excess, with a target price of $270.

“However, the market's expectations for higher and higher share prices are high. We are concerned that any moderate decline expected could pose a huge risk to the stock, based on valuations far above those of our peers.”

edit/lambor

The translation is provided by third-party software.


The above content is for informational or educational purposes only and does not constitute any investment advice related to Futu. Although we strive to ensure the truthfulness, accuracy, and originality of all such content, we cannot guarantee it.
    Write a comment