Author: Du Yu Source: Hard AI Apple's Worldwide Developers Conference WWDC will kick off with a keynote speech at 1 pm US Eastern Time on June 10th (1 am Beijing Time on June 11th). It is widely regarded as a critical moment for Apple to integrate AI features into various hardware operating systems. On a software-centric release event, WWDC has traditionally focused on software, with the exception of last year's Vision Pro, Apple's first mixed reality head-mounted device. Analysts say that this year Apple may focus on AI updates to promote software functionality to catch up with competitors, such as adding artificial intelligence to emoji, photo editing, messaging, and email, without announcing anything truly revolutionary. It is expected that the updates for the new generation iOS 18 mobile operating system and macOS 15 computer operating system will be released next week. The operating system of smartwatches, iPad tablets, and Vision Pro mixed reality head-mounted displays should also incorporate more AI functionality, such as allowing Apple watches to use virtual assistant Siri to play music on other devices.

The weather is good today The weather is good today.

On Thursday, October 10th, AMD, which directly competes with Nvidia in the field of artificial intelligence accelerators, launched a new product, stating that the AI chip will be released for shipment soon and will surpass Nvidia in performance.

However, AMD's stock price briefly turned higher after erasing a 1.7% decline in early trading, only to reverse course again and fall more than 4% to a one-week low. Intel's stock price fell over 1% again after midday trading, while Nvidia maintained a more than 1% gain.

Some analysts believe that in the AI accelerator market, AMD lags far behind industry leader Nvidia, ranking second, and this situation is unlikely to change in the short term. Due to the demand for AI chips far exceeding the supply, AMD's new products are unlikely to impact Nvidia's datacenter revenue. AMD is expected to have datacenter revenue of $12.83 billion this year, while Nvidia's similar revenue is $110.36 billion.

What was announced? Servers CPU, AI GPU accelerators, and multiple network chips.

In the "Advancing AI Event" held in San Francisco, California, USA, AMD introduced a new fifth-generation Epyc datacenter CPU, codenamed "Turin", with computers featuring this processor set to be launched.

The new chip has up to 192 processor cores, priced at $14,813, using the Zen 5 architecture, providing up to a 37% speed increase for advanced AI data processing, aiming to further enhance AMD's leading position in the datacenter field.

AMD CEO Lisa Su stated that the performance of the Turin chip will surpass the latest Intel chips. Since its launch, AMD's EPYC CPUs have captured a 34% market share in the server domain, far exceeding the 2% share in 2018.

Additionally, AMD has introduced the Instinct MI325X AI GPU accelerator, directly competing with nvidia's Blackwell. Lisa Su mentioned that this new AI chip outperforms its nvidia counterparts in inference and uses new memory technology to accelerate AI calculations.

She also pointed out that the demand for AI accelerators exceeds expectations, with the AI accelerator market expected to reach $400 billion by 2027 and further rise to $500 billion in 2028.

Some analysts refer to the MI325X as an "upgrade in the middle of the cycle," still adopting the same basic design and CDNA 3 GPU architecture introduced last year with the MI300X. However, the MI325X is the "first AI GPU with 256GB of HBM3e memory." In the second half of 2025, AMD still plans to launch the MI355X chip with 288GB of HBM3e memory and CDNA 4 architecture, with inference performance expected to increase by 35 times.

According to reports, the MI325X will improve the capability of the NVIDIA H200 GPU by 1.8 times and increase the bandwidth by 1.3 times, with theoretical FP16 and FP8 computing performance at peak times also 1.3 times that of the H200 chip. The first batch of MI325X chips will be mass-produced and shipped in the fourth quarter of this year, and will be more widely available in the first quarter of next year through server manufacturers such as Dell, Eviden, Gigabyte, HPE, Lenovo, and Super Micro Computer.

In addition, AMD has also released multiple networking chips to speed up data movement between chips and systems within data centers, such as the Pensando Salina DPU and Pensando Pollara 400. The latter is the industry's first AI NIC that supports the Ultra Ethernet Consortium (UEC). On the software front, AMD has introduced the latest ROCm 6.2 ecosystem, with an average performance improvement of 2.4 times, up to 2.8 times in a range of AI inference workloads, and an average training performance improvement of 2.4 times.

According to information provided by AMD, the MI 400 accelerator chip will be launched in 2026, using a more advanced CDNA Next architecture at that time. Some commentators say AMD is trying to break NVIDIA's dominance in the field of AI accelerators. Like NVIDIA, AMD is also committed to releasing a new accelerator every year to accelerate innovation.

Forrest Norrod, the executive responsible for AMD's datacenter solution business, stated: AMD will continue to follow the product roadmap, providing customers with the required performance and desired choices, to quickly bring AI infrastructure to market in a large-scale manner.

With the continuous development of the new AMD Instinct accelerator, Epyc processor, and AMD Pensando network engine, as well as the open software ecosystem, and the ability to integrate all of this into AI infrastructure optimization, AMD highlights the crucial expertise in building and deploying world-class AI solutions.

Su Zifeng also stated that there are currently no plans to hire any foundries other than Taiwan Semiconductor, but they are open to the prospect of working with other foundries in the future, hoping for more geographic diversity among company production partners.

How does Wall Street view it? The new product launch was disappointing, but MI300X may ship 0.5 million units this year.

Analysts generally recognize AMD's efforts to catch up with Nvidia in the profitable AI GPU market, continuously surpass Intel in the server CPU field, and react far better than Intel in developing AI accelerators. Moreover, AMD's market cap has also surpassed Intel.

However, from the stock price performance perspective, investors were somewhat disappointed with AMD's AI chip new product launch event on Thursday. Lynx Equity Strategies believes this may be because MI325X cannot significantly challenge Nvidia's dominant position:

Due to investors' expectations for clear details of MI325 and its subsequent products driving AMD's stock price to a slight increase last week, we are skeptical of AMD's ability to narrow the gap with Nvidia unless AMD can demonstrate significant market share growth driving force. Otherwise, the stock may retest its annual low.

In recent months, AMD's customer appeal has been relatively weak. Although the company's management may promote that its chip benchmark test results are superior to Nvidia's Hopper series, we believe that the MI300X still lags behind Nvidia H100 in running actual workloads. Besides excellent HBM memory density, software stack, backplane network, and power consumption are equally important.

The aforementioned institutions also mentioned that the support from Meta, which backed AMD's launch of MI300X last year, may have diminished. Meta may potentially replace Nvidia's products with internally developed chips, and Amazon AWS also has this intention. Therefore, there are doubts about whether AMD can be seen as a viable alternative to Nvidia. The MI chip series and its software stack may not be the 'plug-and-play' solution as AMD advertises.

But KeyBanc's analyst John Vinh believes that even far behind nvidia, AMD's development trajectory is clearly optimistic, expecting to ship 0.5 million pieces of Instinct MI300X AI accelerators this year, increasingly gaining a foothold in the AI market.

"Thanks to its outstanding performance and cost-effectiveness, AMD's Instinct MI300X AI GPU is a 'hot' product in the field. Praises from partners like Lenovo also help bring in new business.

AMD is more competitive in the AI market than a few months ago, implementing a thorough business transformation such as unifying its datacenter and consumer GPU architectures, and expanding its AI business through partnerships with companies like Samsung, Microsoft, and Oracle.

Thursday's AI push event will reshape AMD's AI business as the company will introduce new architectures, showcase its future plans, which could further intensify industry competition.

AMD will announce its third-quarter performance in a few weeks. In July, the sales guidance for this year's AI accelerator chips was raised from the previous 4 billion to 4.5 billion dollars, representing rapid growth. Last year, demand for the MI300X chip surged due to enterprises' enthusiasm for developing and deploying generative AI products.

Before Thursday's event, Wall Street has given AMD 13 'buy' ratings and 3 'hold' ratings, with an average target price of $190.56, representing about 11% upside potential. AMD has risen nearly 11% this year, hitting a new high in March.