AMD Nvidia launched a new artificial intelligence chip on Thursday aimed squarely at Nvidia's data center graphics processors, known as GPUs.
The Instinct MI325X, as the chip is called, will begin production before the end of 2024, AMD said Thursday during an event to announce the new product. If developers and cloud giants look to AMD AI chips as a close replacement for… Nvidia products, could put pricing pressure on Nvidia, which has enjoyed gross margins of roughly 75% while demand for its GPUs has been high over the past year.
Advanced generative AI, like OpenAI's ChatGPT, requires massive data centers full of GPUs in order to do the necessary processing, which has led to increased demand for more companies to provide AI chips.
In the past few years, Nvidia has controlled the majority of the data center GPU market, but AMD has historically held second place. Now, AMD is aiming to take share from its Silicon Valley rival or at least capture a significant portion of the market, which it says will reach $500 billion by 2028.
“Demand for AI has continued to rise and has already exceeded expectations. It is clear that the rate of investment continues to grow everywhere,” AMD CEO Lisa Su said at the event.
AMD did not reveal major new cloud or Internet customers for its Instinct GPUs at the event, but the company previously revealed that both Meta and Microsoft are buying its AI GPUs and that OpenAI is using them in some applications. The company also did not reveal pricing for the Instinct MI325X, which is typically sold as part of a complete server.
With the launch of the MI325X, AMD is accelerating its product schedule to release new chips on an annual schedule to better compete with Nvidia and capitalize on the AI chip boom. The new AI chip is the successor to the MI300X, which began shipping late last year. The company said that AMD's 2025 chip will be called the MI350, and its 2026 chip will be called the MI400.
The introduction of the MI325X will pit it against Nvidia's upcoming Blackwell chips, which Nvidia has said will begin shipping in bulk early next year.
The successful launch of AMD's latest data center GPU could attract interest from investors looking for additional companies on track to capitalize on the AI boom. AMD is up just 20% so far in 2024 while Nvidia stock is up more than 175%. Most industry estimates suggest that Nvidia has more than 90% of the data center AI chip market.
AMD stock fell 3% during Thursday trading.
AMD's biggest hurdle in gaining market share is that its competitor's chips use its own programming language, CUDA, which has become standard among AI developers. This essentially locks developers into the Nvidia ecosystem.
In response, AMD said this week that it is improving its competing software, called ROCm, so AI developers can more easily shift more of their AI models to AMD chips, which it calls accelerators.
AMD has framed its AI accelerators as being more competitive in use cases where the AI models are creating content or making predictions rather than the AI model processing terabytes of data to optimize it. This is partly due to the advanced memory AMD uses on its chip, allowing it to server the Meta's Llama AI model faster than some Nvidia chips.
“What you see is that the MI325 platform delivers up to 40% greater inference performance than the H200 on Llama 3.1,” Su said, referring to the MI325 platform. dead Big language AI model.
Taking into account Intel as well
While AI accelerators and GPUs have become the most watched part of the semiconductor industry, AMD's core business has been central processors, or central processing units (CPUs), which lie at the heart of almost every server in the world.
AMD's data center sales during the June quarter more than doubled last year to $2.8 billion, with AI chips accounting for only about $1 billion, the company said in July.
AMD accounts for about 34% of the total dollars spent on data center CPUs, the company said. This is still less than Intelwhich remains the market leader with its line of Xeon chips. AMD aims to change that with a new line of CPUs, called EPYC 5th Gen, which it also announced on Thursday.
These chips come in a number of different configurations ranging from a low-cost, low-power octa-core chip that costs $527 to 500-watt 192-core processors intended for supercomputers that cost $14,813 per chip.
AMD said the new CPUs are particularly good for feeding data into AI workloads. Almost all GPUs require a CPU on the same system in order to run the computer.
“AI today is really about CPU capabilities, and you can see that in data analytics and a lot of these types of applications,” Su said.
Watch: Technology trends are set to continue over the years, and we're still learning with AI, says AMD CEO Lisa Su
