AMD, a renowned semiconductor manufacturer, announced on Tuesday that its cutting-edge GPU for artificial intelligence (AI), the MI300X, is set to begin shipping to select customers later this year.

Speaking to investors and analysts in San Francisco, AMD CEO Lisa Su emphasized that AI represents the company’s most significant long-term growth opportunity.

Su highlighted the potential of the data center AI accelerator market, projecting its growth from $30 billion this year to over $150 billion by 2027, with a compound annual growth rate of over 50%. AMD sees GPUs as the driving force behind the advancement of generative AI, and the MI300X, along with its CDNA architecture, is purpose-built for large language models and cutting-edge AI applications.

One standout feature of the MI300X is its impressive memory capacity, supporting up to 192GB. This allows the chip to accommodate even larger AI models compared to competing chips like Nvidia’s H100, which supports 120GB of memory. Given the increasing calculations required by large language models, AMD’s MI300X has the potential to deliver superior performance. During a demonstration, AMD showcased the MI300X running a 40 billion parameter model named Falcon, while OpenAI’s GPT-3 model boasts 175 billion parameters.

AMD’s focus on expanding memory capacity aligns with the growing demand for AI models that require extensive memory usage. The increased memory in AMD’s chips eliminates the need for developers to rely on multiple GPUs, streamlining their operations and offering a more efficient solution.

Additionally, AMD announced the development of its Infinity Architecture, which combines eight of its M1300X accelerators into a single system. This move brings AMD in line with competitors like Nvidia and Google, who have introduced similar systems that consolidate multiple GPUs into a single box for AI applications.

To complement its AI chips, AMD unveiled its proprietary software, ROCm. This software stack aims to provide developers with a powerful toolset that integrates seamlessly with the open ecosystem of models, libraries, frameworks, and tools, enabling enhanced AI development capabilities.

As AMD expands its presence in the data center market with the MI300X GPU and accompanying software, industry experts are closely monitoring the company’s progress. With its focus on addressing the growing demands of AI applications, AMD is poised to compete against established players in the AI accelerator space, positioning itself as a key player in the evolving AI landscape.

AMD Reports Better-Than-Expected Q1 Revenue and Earnings, Expects Growth in H2