AMD shares climbed roughly 7% on Tuesday after the chipmaker issued upbeat projections for its data-center business at the Citi Global TMT Conference, forecasting a total addressable AI market of $2 trillion by 2030.
AMD management said inference has become the majority driver of AI computing over the past twelve months, with workloads shifting from basic chatbots toward independent, agentic AI systems. The company projects its data-center segment will more than double next year, while server CPU revenue is expected to expand more than 80% year-over-year in the second half of this year and over 70% throughout next year.
The firm also provided an update on its MI450 GPU, a next-generation AI accelerator set to launch this quarter with production ramping through the fourth quarter and into 2027. Volume expectations for its Helios rack-scale systems for 2027 are already outpacing initial projections, AMD said. The company disclosed it is in ongoing technical engagements with its three leading customers on its subsequent MI500 and MI600 chips.
Anchor clients for Helios systems include Meta Platforms and two unnamed AI labs, both of which provided demand forecasts larger than their initial purchase agreements. AMD also pointed to newly emerging demand from so-called "neo-cloud" providers and announced a partnership with Cerebras.
Profitability in AMD's data-center GPU segment currently trails the company's corporate average margins, but the decline has been offset by strong double-digit growth in high-margin server and embedded segments during the second and third quarters, keeping overall profitability intact, according to Citi's summary of the conference.
