AMD launches AI chip to rival Nvidia’s Blackwell, aiming to capture a larger share of the booming AI market. AMD has introduced its latest artificial intelligence chip, the Instinct MI325X, in a move directly targeting Nvidia’s hold on the data center graphics processor market. This development is expected to heat up competition in the fast-growing AI chip industry, where Nvidia has maintained a dominant position.
The Instinct MI325X is set to begin production by the end of 2024, AMD announced during a recent product launch event. If this AI chip is embraced by developers and cloud providers as a viable alternative to Nvidia’s popular GPUs, it could place pricing pressure on Nvidia, which has seen significant demand for its products. Nvidia’s GPUs have been essential in powering advanced AI applications, such as OpenAI’s ChatGPT, contributing to the company’s 75% gross margin.
AI Market Poised for Growth
Nvidia currently commands over 90% of the data center GPU market. However, AMD has remained a strong second player in this field and is now aiming to capture a larger share. The AI chip market is projected to reach a value of $500 billion by 2028, with companies worldwide investing heavily in AI technologies.
AMD’s CEO, Lisa Su, emphasized the continued demand for AI solutions, noting that investment in AI has exceeded expectations globally. Despite this enthusiasm, AMD did not disclose any new major cloud or internet customers for the Instinct MI325X during the event. However, the company has previously revealed that both Meta and Microsoft use its AI GPUs, and OpenAI integrates them for some applications.
Performance and Competitors
With the Instinct MI325X, AMD launches AI chip to rival Nvidia’s Blackwell, challenging Nvidia’s dominance in data center GPUs. AMD has accelerated its product launch schedule, positioning the MI325X against Nvidia’s upcoming Blackwell chips, which will ship in early 2025. AMD also provided details about future AI chip releases, including the MI350 in 2025 and MI400 in 2026, as part of its strategy to compete more effectively in the market.
While pricing for the MI325X was not disclosed, the chip is likely to be sold as part of a larger server setup. If the product succeeds in gaining traction, it could attract investors looking for opportunities in the AI space.
Challenges Ahead
One of AMD’s biggest obstacles in taking market share from Nvidia is the widespread use of Nvidia’s CUDA programming language. CUDA has become a standard among AI developers, making it difficult for companies to transition to AMD’s ecosystem. In response, AMD has been working on improving its ROCm software, which enables developers to shift their AI models to AMD’s chips.
AMD emphasized that its AI accelerators are particularly competitive in scenarios where models generate content or predictions, rather than processing vast amounts of data. The company highlighted the performance of its advanced memory, claiming that the MI325 platform delivers up to 40% more inference performance on Meta’s Llama 3.1 AI model than Nvidia’s H200 chips.
Beyond the battle for AI accelerators, AMD continues to challenge Intel in the data center CPU market. While AI chips have captured headlines, CPUs remain critical to the operation of most servers. AMD’s data center sales have more than doubled in the past year, with AI chips accounting for $1 billion of its $2.8 billion in total sales in the June quarter.
AMD launches AI chip to rival Nvidia’s Blackwell, positioning itself as a stronger competitor in the $500 billion AI chip market. AMD currently holds a 34% share of the data center CPU market, though Intel remains the leader with its Xeon line. In a bid to narrow the gap, AMD has introduced its 5th Generation EPYC CPUs. These chips come in a range of configurations, from an 8-core, low-power model priced at $527 to a high-performance 192-core version intended for supercomputers, costing $14,813.