AMD Forecasts Over 70% Server CPU Growth in 2027 After Record $11.5 Bn Quarter

AMD reported a record second-quarter revenue of $11.5 billion, up 50% year-over-year (YoY), as data centre revenue more than doubled to $6.7 billion on strong demand for EPYC server processors and Instinct AI accelerators. 

The data centre business accounted for 58% of total revenue, up from 42% a year earlier. 

The company also guided for approximately $13 billion in third-quarter revenue, representing 41% YoY growth. 

Its quarterly profit rose 163% YoY to a record $2.3 billion. 

AMD materially raised its long-term outlook as it sees AI infrastructure demand expanding faster than previously expected. 

CEO Lisa Su said the company now expects revenue growth to be substantially above the long-term target of more than 35% outlined at its Financial Analyst Day.

The company also forecast that its data centre business will more than double in 2027, supported by server CPU growth of more than 70% and a faster ramp in AI accelerators.

AMD also increased its view of the AI infrastructure opportunity, estimating the data centre AI accelerator market will grow to approximately $1.4 trillion by 2030.

Su said “agentic AI is creating a new growth vector for server CPUs”, prompting AMD to raise its estimate for the server CPU market to $220 billion by 2030 as AI workloads increasingly require more CPU infrastructure alongside accelerators.

The company said cloud and enterprise server CPU sales each grew more than 70% during the quarter, marking its fifth consecutive quarter of record server CPU revenue. 

Unveiled at AMD’s Advancing AI 2026 event last month, the sixth-generation EPYC Venice processors were positioned by the company as outperforming Intel Xeon, NVIDIA’s Vera and rival Arm-based CPUs across AI and data centre workloads.

Su added that customer demand for Venice processors is stronger than for any previous EPYC generation, while adoption continues to expand across hyperscalers including AWS, Microsoft, Google Cloud and Oracle.

On the AI infrastructure side, AMD said its Helios rack-scale platform is now in production, with initial shipments scheduled to begin later this quarter before ramping through the fourth quarter and into 2027. 

Helios integrates sixth-generation EPYC Venice CPUs, Instinct MI455X GPUs and AMD Pensando networking into a rack-scale AI system designed for frontier AI training and inference.

Su said customer demand for Helios is running ahead of AMD’s initial forecasts, referring specifically to expected deployment volumes. 

The company also said supply constraints that affected server CPUs during the first half are easing as additional wafer, packaging and substrate capacity comes online to support future growth.

The momentum builds on AMD’s recent announcement that Anthropic will deploy up to 2 gigawatts of Instinct MI450 GPUs in Helios from 2027, following similar large-scale AI infrastructure commitments from OpenAI, Meta, Microsoft and Oracle.

Despite the strong earnings beat and an upbeat long-term outlook, AMD shares fell nearly 8% in after-hours trading.