Meta, Microsoft, and OpenAI announced at an AMD investor event on Wednesday (December 6th) that they will use AMD's latest artificial intelligence (AI) chip, Instinct MI300X, in the future.
A graphics processor (GPU) is crucial for creating and deploying artificial intelligence programs such as ChatGPT for OpenAI. At a time when Nvidia's GPU dominates the artificial intelligence market, many technology companies are also actively seeking alternatives to reduce costs. The latest statements from companies such as Meta and Microsoft are clear examples so far.
The first product launched by AMD at the conference was the Instinct MI300X accelerator, which consists of 8 MI300X GPUs and can provide up to 1.5TB of HBM3 memory capacity.
Compared with Nvidia's H100 HGX, the Instinct MI300X accelerator has significantly higher throughput and latency performance when running large language model inference, and is also more outstanding in various AI and HPC projects.
Analysis suggests that if AMD's latest high-end chips start shipping early next year and their performance is sufficient to meet the needs of technology companies and cloud service providers building artificial intelligence models, it will inevitably create competitive pressure on Nvidia's soaring sales growth of artificial intelligence chips.
AMD CEO Su Zifeng also stated on Wednesday, "All external interest is focused on large processors and GPUs for cloud computing."
The problems faced by AMD
The main issue AMD is currently facing is whether the company, which has always relied on Nvidia chips as its foundation for computing power, will invest time and money to increase its procurement from another GPU supplier. Su Zifeng believes that AMD still needs to make efforts in this regard.
Meanwhile, price is also important. AMD did not disclose the pricing of the MI300X at the meeting on Wednesday. However, Su Zifeng told reporters that AMD's chips must be priced lower than Nvidia's chips in order to persuade customers to purchase. It is reported that Nvidia's GPU chips are priced at approximately $40000 per chip.
In terms of software, AMD informed investors and partners that the company has improved a software suite called ROCm, addressing a critical flaw that is currently one of the main reasons why artificial intelligence developers prefer Nvidia.
The AMD conference announced the latest version of the ROCm 6 open-source software platform, which has made significant progress in improving AI acceleration performance.
In addition, AMD also revealed that it has signed agreements with some companies that require GPUs the most to use the chip.
Numerous supporters
At AMD's investor event, many companies came to support, including Microsoft, Meta, OpenAI, Oracle, and others.
Meta stated at the meeting that it will use the MI300X GPU to handle the workload of artificial intelligence inference.
Microsoft Chief Technology Officer Kevin Scott stated that the company will develop access to the MI300X chip through Azure network services.
OpenAI stated that it will support AMD's GPU in a software product called Triton. Triton is not a large language model like GPT, it is an open-source programming language of Python.
In addition, Oracle also claims that its cloud computing will use this chip.
Interestingly, just recently, a latest report from market research firm Omdia Research showed that Meta and Microsoft each purchased 150000 H100 GPUs from Nvidia this year, ranking first in terms of procurement volume.
This shows the fierce competition in the artificial intelligence GPU market.
AMD predicted on Wednesday that the total market for artificial intelligence GPUs may climb to $400 billion over the next four years (until 2027), which is twice the company's previous forecast.
Su Zifeng also told reporters that AMD does not believe that it needs to defeat Nvidia in order to achieve good results in the market, but can occupy a good portion of the market share in the future of $400 billion on its own.