In the most recent earnings call, AMD revealed expectations of a deceleration in revenue growth. CEO Lisa Su explained the slowdown as a result of seasonal factors, yet she expressed continued optimism about the vast total addressable market for AI.
After releasing the financial results for the quarter that ended in December 2024, AMD held an earnings call, with Su and CFO Jean Hu responding to analysts' questions.
According to Su, AMD delivered record annual revenue in 2024, with net income rising 26% year-over-year and free cash flow more than doubling from 2023. The company's data center segment accounted for roughly half of the annual revenue, adding that its Instinct accelerators generated sales of over US$5 billion in 2024. The data center segment alone grew 69% year-over-year in the fourth quarter to a record US$3.9 billion.
Elaborations on guidance
Hu said that AMD expects first-quarter revenue to reach approximately US$7.1 billion, plus or minus US$300 million, marking a 30% year-over-year increase. Growth in the data center and client businesses is expected to offset a significant decline in Gaming and a modest drop in Embedded revenue. However, revenue is projected to decline 7% sequentially due to seasonality.
Hu added AMD will restructure its segment reporting, combining client and gaming into a single unit while maintaining separate revenue disclosures. AMD also forecasts a non-GAAP gross margin of 54%, operating expenses of US$2.1 billion, and a non-GAAP effective tax rate of 13%.
Su stated that the data center segment is anticipated to decline in line with the corporate average, while the client and embedded businesses are expected to see steeper declines due to seasonal factors. She confirmed that the company expects GPU revenue to remain roughly flat in the first half of 2025 compared to the second half of 2024, though this could vary depending on deployment timings.
The gaming segment, however, is projected to decline at a slower rate, which Su described as an atypical trend. She attributed this to inventory normalization following a year of adjustments, suggesting a stabilizing segment.
Data center market in the midterm
Su noted that the total addressable market (TAM) of Data Center accelerators is projected to surpass US$500 billion by 2028.
While AMD did not provide specific financial guidance for 2025, Su indicated that the company expects the data center business to grow into the tens of billions over the next few years. The company plans to offer more detailed segment-level guidance and qualitative insights as the year progresses.
Su said AMD had experienced substantial growth in its data center GPU business, with its MI300X accelerators being widely adopted by global cloud service providers. Meta uses these GPUs exclusively for its Llama 405B Frontier model, and Microsoft employs them for its GPT-4-based Copilot services, utilizing thousands of GPUs for AI training, inference, and HPC tasks. Additionally, AI service providers like IBM, DigitalOcean, and Vultr have begun deploying AMD's Instinct accelerators, signaling high demand for its AI-focused hardware.
Looking ahead to 2025, AMD plans to ramp up production of its MI350 series, initially slated for the second half of the year, to mid-2025 due to stronger-than-expected customer demand.
Regarding workload focus, Su mentioned that although AMD traditionally emphasized inference workloads, demand for training solutions is increasing. The company is investing heavily in training software and libraries, especially in networking, to enhance its capabilities. Su added that while ASICs are crucial, AMD is dedicated to offering a wide range of solutions, including CPUs, GPUs, and networking technologies. This comprehensive approach positions AMD uniquely in the AI and data center sectors.