AMD Projects Data Center Revenue to Double on AI Chip Demand

AMD sees AI demand expanding across product lines

Advanced Micro Devices is preparing for what its chief financial officer calls "the most consequential technology transformation" as artificial intelligence workloads spread beyond traditional graphics processors to encompass server CPUs, AI-enabled personal computers, and adaptive computing platforms.

Speaking at the Citi Global TMT Conference, CFO Jean Hu outlined how the AI market has evolved as inference demand—the process of running trained models—has overtaken training demand. This shift is driving increased requirements for both GPUs and CPUs, particularly as agentic AI systems that can take autonomous actions gain traction.

AMD now estimates its addressable market opportunity at $2 trillion by 2030, an update the company shared at its recent Advancing AI event. The chipmaker expects its data center business to double in the coming year, powered by demand for both GPU accelerators and server processors.

MI450 production timeline and customer commitments

The company plans to begin production shipments of its MI450 accelerator products at the end of the third quarter, with revenue recognition starting in that period. Hu projected a "very significant step up" in the fourth quarter, followed by another increase in the first quarter of 2027 and continued scaling throughout that year.

AMD has secured three anchor customers—Meta, OpenAI, and Anthropic—for multigigawatt-scale deployments spanning multiple product generations. The company also reported strong uptake of its current MI350 platform among model builders, emerging AI companies, and enterprise customers.

Deploying AMD's Helios rack-scale systems involves complexity beyond component availability, Hu noted. The company is collaborating with original design manufacturers, supply-chain partners, and customers on manufacturing processes, mechanical systems, and software integration.

Supply expansion to meet 2027 demand

"The demand and the volume for 2027 has certainly gone above our original initial expectation," Hu said, indicating AMD is working to expand supply capacity to meet customer requirements.

Matt Ramsay, AMD's corporate vice president of financial strategy and investor relations, said technical engagement with large model-building customers is informing the design of future Instinct MI500 and MI600 products as well as subsequent rack configurations. AMD also plans to work with neocloud providers to serve customers beyond its largest strategic accounts.

Why it matters

AMD's revenue projections and production timeline offer a concrete signal that AI infrastructure spending is broadening beyond NVIDIA's dominant position in the accelerator market. The shift from training to inference workloads creates opportunities for differentiated chip architectures, while multigigawatt commitments from major AI labs suggest capital expenditure cycles extending well into 2027. For enterprises evaluating AI infrastructure investments, AMD's ramp schedule provides visibility into when alternative accelerator options will reach production scale.

These details were first reported by AI Watch based on AMD executive remarks at the Citi Global TMT Conference.