Advanced Micro Devices Sees Agentic AI Doubling CPU Opportunity as MI450 Nears

Key Points

• Agentic AI is boosting AMD’s server CPU outlook, with executives saying the shift from AI experimentation to deployment is driving much higher demand for CPU-heavy workloads like orchestration, database access and tool execution.
• AMD said its CPU market opportunity has been revised sharply higher, from a prior estimate of $60 billion by 2030 to more than $120 billion, with agentic AI expected to make up over half of that long-term market.
• The company said its MI450 GPU launch remains on track for the second half of the year, while supply remains tight and AMD is working closely with TSMC and major customers to secure capacity through 2027 and beyond.

Advanced Micro Devices NASDAQ: AMD executives said demand tied to agentic artificial intelligence is accelerating faster than expected, driving a sharp expansion in the company’s server CPU opportunity and shaping its plans for GPU systems, supply chain commitments and customer engagements.

Speaking at the BofA Global Technology Conference with Vivek Arya, managing director at Bank of America, AMD Chief Financial Officer Jean Hu and Matt Ramsay, head of investor relations, said the shift from AI experimentation to broader deployment is creating new demand for compute across CPUs, GPUs and other accelerators.

Agentic AI Drives CPU Demand

Hu said the “biggest change during the last few months” has been the rise of agentic AI, which she described as moving beyond answering questions into orchestration, database access and tool execution. Those workloads, she said, require significant CPU performance.

AMD reported what Hu called “record CPU performance” in the first quarter, with CPU revenue growing more than 50%. The company guided for second-quarter CPU revenue growth of more than 70% year over year. Hu said roughly two-thirds of that growth is coming from unit expansion rather than higher average selling prices.

Ramsay said the AI capital spending environment is shifting from being driven primarily by large-model training toward inference, and that chatbot inference is evolving into agentic inference. Between inference tasks, he said, there is substantial CPU work, including post-processing data, determining next steps and retrieving information from cloud systems, ERP platforms, payment systems or CRM systems.

That workload is helping drive demand for AMD’s Turin server CPUs, Ramsay said. He also pointed to the company’s upcoming Venice parts, which he described as 256-core, 2-nanometer products expected to become a key workhorse next year.

AMD Raises Its View of the CPU Market

Hu said AMD previously outlined three CPU market segments: traditional general compute, head-node CPUs that communicate with GPUs, and agentic AI systems. At its financial analyst day last November, she said AMD projected the CPU total addressable market would grow at an 18% compound annual rate to $60 billion by 2030. Since then, the company has updated its view to more than $120 billion, citing faster agentic AI adoption.

Hu said the traditional general-purpose CPU market is currently viewed by third parties as a roughly $25 billion to $30 billion opportunity in 2025, with steady growth. She said the head-node segment is expected to grow faster as CPU-to-GPU ratios change. The most significant opportunity, she said, is agentic AI, which AMD expects to represent more than 50% of the larger long-term CPU market opportunity.

On the question of x86 versus Arm, Hu said AMD focuses less on instruction set architecture and more on delivering the best total cost of ownership for customers. Ramsay added that traditional enterprise workloads have an affinity for x86 because of the existing code base, while head-node and agentic AI systems are more focused on performance, utilization and security features.

Supply Remains Tight, With TSMC Central to AMD’s Plans

Hu said AMD has been pleased with its server CPU share gains, noting that the company reached about 46% value share in the first quarter. She said AMD’s success is closely tied to its long-standing relationship with Taiwan Semiconductor Manufacturing Co., including chiplet design and packaging technologies.

Hu said AMD has been planning the current ramp since last year and is working with TSMC to increase supply. Demand remains “tremendous” and supply is still tight, she said, adding that AMD expects to be able to place all available supply with customers this year and next year. Longer-term planning is already focused beyond 2027.

Ramsay said supply constraints remain in areas such as 3-nanometer capacity, but said AMD had already planned for substantial growth in 2026 and 2027. He said TSMC has been disciplined in bringing capacity online and in assessing demand deeper into the customer base.

MI450 Launch Remains on Track

On GPUs, Hu said AMD is on track with its MI450 product, which has been sampled and is expected to launch starting in the second half, with activity beginning in the third quarter and continuing into the fourth quarter. She said AMD has been preparing for the launch for a long time, including through its 2024 acquisition of ZT Systems to add system-level design capabilities.

Ramsay said several customers already have full Helios racks in their own data centers running production workloads for testing. He said AMD expects a significant revenue increase in the fourth quarter, followed by another sizable increase in the first quarter, while acknowledging the complexity of ramping a rack-scale system.

Hu said AMD is pleased with its long-term partnerships with OpenAI and Meta, adding that forecasts from those customers are above AMD’s original plan for 2027. She said AMD expects other multi-gigawatt customers and is working with hyperscalers, model companies and AI-native companies to expand adoption of MI450.

Memory Costs and Visibility Into 2027

Hu said rising memory costs are a significant industry issue and that AMD’s priority is to secure capacity through longer-term supplier agreements. She said AMD does not want to use memory shortages simply to raise prices, but expects customers to help absorb cost increases as part of strategic, long-term relationships.

Ramsay noted that in server and PC markets, most DRAM is purchased by OEMs, ODMs and hyperscalers rather than flowing through AMD’s profit and loss statement. He said AMD is working with customers to match available memory supply with CPU and GPU supply to avoid mismatches or hoarding.

Hu said the tight supply chain environment is causing customers to plan further ahead. AMD has “very good visibility into 2027 and beyond,” she said, with confidence in both demand and supply visibility for 2027, while planning for 2028 and later continues.

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.

This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to contact@marketbeat.com.