AMD has further raised its full-year revenue forecast for AI data center GPUs to US$4.5 billion, underscoring the surging demand for computing power driven by generative AI. The chipmaker is ramping up its investments in both hardware and software solutions to capture this growing market.
For the second quarter of 2024, AMD reported data center revenue of US$2.83 billion, comprising 49% of its total revenue and marking a staggering 115% year-over-year increase. Despite a rise in R&D expenses, operating income soared to US$743 million, up 405% from the previous year, pushing the operating margin up 15pp to 26%.
CEO Lisa Su attributed the robust growth to the rapid adoption of the Instinct MI300 series GPUs, built on the CDNA 3 architecture. These GPUs generated over US$1 billion in quarterly revenue for the first time, with Microsoft expanding its deployment of MI300X accelerators to power GPT-4 Turbo and various Copilot services.
AMD is accelerating its Instinct product roadmap, planning to launch the MI325X later this year, which is expected to boost fourth-quarter revenue. Looking ahead to 2025, AMD will unveil the MI350 series, leveraging the new CDNA 4 architecture, promising a 35-fold performance increase over CDNA 3. The MI400 series, expected in 2026 with the CDNA Next architecture, is also making significant strides in development.
Su noted that AMD's Instinct GPUs and its ROCm software stack have received "extremely" positive feedback from customers, prompting the company to raise its data center GPU revenue outlook from the previously estimated US$3.5 billion to US$4 billion, to over US$4.5 billion.
Looking to the future, Su emphasized that generative AI and the development of increasingly powerful models are fueling the surge in computing demand. AMD is optimistic about its growth prospects over the coming years and is increasing investments in its software, hardware, and solutions portfolio. The company aims to release at least one flagship data center GPU annually, incorporate AI capabilities across all product lines, and strengthen the ROCm platform through collaborations with the open-source community.
In addition to its data center gains, AMD's client segment, which includes its PC business, also posted significant growth in the second quarter. Revenue rose 49% year-over-year to US$1.5 billion, with operating income reaching US$89 million, resulting in a 6% operating margin—a notable improvement from the negative 7% margin in the same period last year.
Su sees AI PCs as an emerging growth engine for the broader PC market and anticipates a wider price range for AI PCs by 2025, creating more business opportunities. However, when questioned about the potential impact of Arm-based PCs entering the market, Su maintained that while the PC market is large, AMD is still underrepresented and will approach competition cautiously.
Regarding the MI300 series supply chain, Su acknowledged ongoing challenges, with supply expected to gradually increase in the latter half of the year, though tight conditions are likely to persist into 2025.
On the competitive front with Nvidia, Su addressed concerns about AMD's position in rack-level performance, an area where the company may be lacking critical architecture. She highlighted that as systems grow more complex, particularly for large training clusters, customers will increasingly rely on suppliers for integration assistance. AMD's solution includes its Infinity Fabric technology, the foundation for UALink, and broader system-level integration capabilities.
AMD continues to invest heavily across its portfolio, from CPUs and GPUs to networking technologies acquired from Pensando, aiming to accelerate its customers' time to market.
