Advanced Micro Devices (NASDAQ:AMD) expects continued rapid growth in its server CPU and data center businesses as demand for AI inference systems, including agentic AI workloads, expands, according to Matt Ramsay, the company’s corporate vice president of financial strategy and investor relations.
Speaking at a KeyBanc Capital Markets event, Ramsay said AMD’s server business grew more than 50% in the first quarter and more than 70% in the second quarter. Both cloud and enterprise server revenue increased by more than 70% during the second quarter, he said.
Inference Shift Drives Server Demand
Ramsay attributed the growth outlook in part to a shift in AI spending from training large models toward inference. He said the transition is occurring alongside the emergence of agentic inference, in which automated agents repeatedly access data, run code and send tasks back to AI models.
Those processes create demand for both accelerators and high-core-count CPUs, he said. While GPUs or other accelerators perform the inference work, CPUs can handle varied tasks such as retrieving data from cloud, enterprise and web sources, reorganizing that information and executing generated code.
“The computing that these workers and agents do is very diverse,” Ramsay said. He said this has increased demand for CPUs with high thread counts and performance across multiple workloads.
AMD said its forthcoming 2-nanometer Venice CPUs are being sampled to customers and will ship in its Helios AI racks, while also being deployed broadly across server markets. The company has also outlined its Florence CPU lineup for 2028.
Supply Chain Focus Includes Packaging and Memory
Ramsay said AMD has worked with Taiwan Semiconductor Manufacturing Co. to secure additional supply and described TSMC as a key partner. He said near-term supply remains tight, but AMD and its partners have had more time to adjust capacity for expected growth in 2027 and 2028.
Advanced packaging is another area of focus, according to Ramsay. He said Venice will be the first server product in the market to use advanced packaging, and referred to AMD’s announced $10 billion ecosystem investment in Taiwan, much of which is directed toward backend capacity.
The company is also coordinating with original equipment manufacturers, original design manufacturers and hyperscale customers to ensure sufficient DRAM availability for server deployments, he said.
Helios Ramp Expected to Begin in September
AMD plans to begin shipping MI450 accelerators, Venice CPUs and certain Pensando networking products to ODM partners building Helios systems in September, Ramsay said. He expects a significant revenue ramp in the fourth quarter and another sizable increase in the first quarter.
He said AMD expects data center revenue, including AI, to grow by “much more than 100%” in 2027. Customer feedback on Helios has been positive, Ramsay said, noting that customers are running model code on sampled systems.
Ramsay identified OpenAI, Meta Platforms and Anthropic as major customers for the rack-scale platform, saying each intends to pursue gigawatt-scale deployments. AMD has commitments for six-gigawatt arrangements, including one-gigawatt commitments from OpenAI and Meta, as well as a one-gigawatt commitment and two-gigawatt ambition from Anthropic, he said.
He cautioned that the pace of deployments will depend on factors including land, power, facilities and capital commitments. AMD’s objective is to deliver stable systems capable of running production code, rather than simply shipping racks, he said.
To reduce execution risk, Ramsay said AMD acquired ZT Systems to add system-level expertise and plans to initially concentrate production with a limited number of ODM partners before expanding more broadly.
ROCm and CPU Strategy
Ramsay said AMD has made substantial progress with its ROCm AI software stack over the past 18 months, accelerated in recent months by the use of AI development tools. He said Anthropic earlier rented a cluster of MI355 systems and brought its primary inference model online and tuned it over a weekend.
AMD is also working with Anthropic to help ensure code automated through Claude can run on ROCm and AMD Instinct platforms, he said.
On CPUs, Ramsay said AMD’s strategy is to build products for a range of workloads rather than frame the market primarily as a competition between x86 and Arm architectures. He said the company sees distinct requirements for AI head nodes, CPU-only agent racks, traditional enterprise deployments and cloud workloads.
For agentic rack deployments, customers are emphasizing metrics such as “agents per megawatt” and “threads per megawatt,” Ramsay said. AMD believes its chiplet-based approach enables it to offer multiple CPU configurations, including Venice products with up to 256 cores and 512 threads, to address those varying needs.
About Advanced Micro Devices (NASDAQ:AMD)
Advanced Micro Devices, Inc (NASDAQ: AMD) is a global semiconductor company that designs and sells microprocessors, graphics processors, chipsets and adaptive computing solutions for a broad set of markets. The company’s product portfolio includes consumer and commercial CPUs under the Ryzen and Threadripper brands, data center processors under the EPYC brand, and Radeon graphics processing units for gaming and professional visualization. AMD also offers semi-custom system-on-chip (SoC) products for gaming consoles and other specialized applications, and provides supporting software and platform technologies for OEMs, cloud service providers and end users.
Founded in 1969, AMD has evolved from a supplier of logic chips into a diversified, fabless semiconductor designer.
