Shares of Advanced Micro Devices (AMD 3.39%) are down nearly 12% (as of Feb. 10), despite a strong fourth-quarter fiscal 2025 earnings report on Feb. 3. The company seems to have missed investors' sky-high expectations for current performance and future growth.
Despite near-term volatility, AMD is well-positioned to recover much faster than expected. Here are the three things long-term investors should understand about the company.
Image source: Getty Images.
Instinct GPUs are ramping up, and Helios is the new opportunity
AMD's artificial intelligence (AI) data center business is showing solid traction. In the fourth quarter, the company's data center segment saw revenue rise 39% year over year to $5.4 billion, driven by rapid deployments of the Instinct MI350 Series GPUs and increasing adoption of Epyc CPUs in servers. Management claimed that 8 out of 10 top AI companies now use its Instinct GPUs in real-time deployments.
The chipmaker has also highlighted the launch of its MI400 Instinct accelerator series and Helios rack-scale platform (fully integrated AI system comprising CPUs, GPUs, networking, and software solutions) as a "major inflection point." Going beyond selling individual Instinct accelerators, AMD will now focus on delivering Helios systems, backed by MI450 chips. Management expects MI450 revenue to begin in the third quarter of 2026 and further ramp up in the fourth quarter, with the vast majority of volume delivered as rack-scale Helios systems.
AMD has already entered into a multi-year, multi-generation partnership with OpenAI, under which the latter will deploy up to 6 gigawatts of Instinct GPUs, starting with MI450 in the second half of 2026. The company is also in discussions with other customers regarding the MI450 and Helios multi-year deployments. With impressive demand visibility, the MI400 series and Helios are well-positioned to capture a significant share of the global AI infrastructure business.
NASDAQ: AMD
Key Data Points
Epyc CPUs are an underappreciated AI-powered catalyst
AMD's Epyc CPUs are also benefiting from AI demand. The company saw record server CPU sales in the fourth quarter, driven by increasing adoption of the fifth-generation Epyc "Turin" CPUs and robust demand for fourth-generation Genoa processors. Management expects CPU revenue to grow from the fourth quarter to the first quarter, despite the first quarter typically being seasonally weaker.
As AI systems scale, CPUs are increasingly required to handle intelligent jobs such as orchestration, memory management, data preprocessing, and agentic workloads involving large volumes of non-GPU tasks. AMD also claims that Epyc has become the "processor of choice" for modern data centers, as cloud providers expand AI infrastructure and enterprises build on-premise data centers to support new AI use cases. Additionally, demand for its next-generation Venice CPUs, launching in the second half of 2026, is already high.
Software execution will determine how big the upside gets
While AMD has historically lagged competitors in software, management is now focused on closing this gap. The company has expanded Radeon Open Compute (ROCm), its open-source software ecosystem, to provide day-zero support for Instinct GPUs and to integrate Instinct GPUs with popular inference engines (software for deploying AI models in production environments). The company has also launched a full-stack enterprise AI suite designed to simplify and accelerate large-scale deployments.
Increasing adoption of ROCm should further translate into a stickier client base and higher margins. This will also help expand deployments and GPU share gains.
