AMD’s AI Infrastructure Push Drives 57% Data Center Growth

extract

AMD’s AI Infrastructure Push Drives 57% Data Center Growth Strong EPYC and Instinct chip demand pushed revenue to $10.3 billion as inference workloads expanded AI infrastructure spending. AMD posted strong first-quarter results, with surging demand for AI infrastructure pushing data center revenue up 57% year over year and cementing the segment as the company’s primary growth engine. Total revenue reached $10.3 billion, up 38% from a year earlier, while net income nearly doubled to $1.4 billion. Drilling into AMD’s balance sheet, data center revenue climbed to $5.8 billion as hyperscalers and enterprise operators continued expanding AI capacity. The company also pointed to growing deployment activity for inference and agentic AI workloads, alongside Meta plans for up to 6 GW of AMD Instinct GPU infrastructure. “We delivered an outstanding first quarter, driven by accelerating demand for AI infrastructure, with data center now the primary driver of our revenue and earnings growth,” CEO Lisa Su said in the earnings release. Data Center Becomes the Growth Engine AMD’s data center segment generated more than half of total company revenue during the quarter, driven by EPYC server CPUs and the continued ramp of Instinct GPU shipments. Matt Kimball, vice president and principal analyst at Moor Insights & Strategy, said AMD is benefiting from both broader AI GPU demand and from shifts in AI deployment patterns. “AMD is certainly benefiting from being the second source in this AI GPU market,” Kimball said. “At the same time, it is making its own strides as inference comes more into focus and that CUDA moat becomes less treacherous to cross.” Kimball said AMD’s MI accelerator portfolio, ROCm software framework and expanding ecosystem support are helping the company gain traction as AI workloads move toward inference. “That CUDA barrier comes down – token economics emerge as a key metric, and performance-per-dollar wins out against raw performance in many cases,” he said. Stephen Sopko, analyst at HyperFrame Research, said the shift toward inference and agentic AI workloads could significantly expand AMD’s server opportunity beyond accelerators alone. “Inference and agentic AI aren’t just incremental – they’re structurally shifting the CPU-to-GPU ratio higher and expanding the total addressable server CPU market,” Sopko said. He pointed to AMD’s recent decision to double its long-term server CPU market outlook to more than $120 billion by 2030. “The Data Center franchise is no longer GPU-only,” Sopko said. “High-performance CPUs are becoming the indispensable foundation for orchestration and real-time inference workloads.” Hyperscale and Sovereign AI Deployments Expand Meta plans to deploy up to 6 GW of AMD Instinct GPUs, including a first 1 GW cluster built on the upcoming MI450 platform. Cloud providers, including AWS, Google Cloud and Microsoft Azure, also expanded EPYC-based instances across multiple workload types. AMD is also working with Tata Consultancy Services on rack-scale ‘Helios’ AI infrastructure aimed at enterprise and sovereign AI deployments. Sopko described the Meta deployment as a major validation of AMD’s ambitions for rack-scale AI. “It’s a multi-year, co-engineered commitment using custom MI450-based Instinct GPUs, 6th-gen EPYC Venice CPUs and the Helios architecture at power densities approaching 120-140 kW per rack with liquid cooling,” Sopko said. Kimball said the enterprise and sovereign AI partnerships may ultimately matter more strategically than headline hyperscale wins. “Meta is one customer that generates significant revenue,” Kimball said. “But by establishing a strong presence in the volume market, AMD is building an AI practice that is far more durable.” He specifically pointed to the Tata Consultancy Services partnership around Helios systems as a meaningful signal for enterprise adoption. “TCS designing Helios systems – this is a big win for AMD,” Kimball said. Sopko said the India and Korea partnerships reflect a broader geographic shift in AI infrastructure investment. “Nations aren’t content to rely solely on US hyperscalers or merchant GPUs,” he said. “They want sovereign control over compute, data and models.” Margin Pressure Reflects Scaling Phase Despite strong year-over-year growth, sequential results were mixed as AMD continued scaling AI supply. Revenue was essentially flat compared to Q4, while operating income declined 16% sequentially. Gross margin also dipped slightly to 53%, down one point from the prior quarter. CFO Jean Hu pointed to expanding revenue and “record quarterly free cash flow,” highlighting operating leverage even as expenses rise. Operating expenses increased 34% year over year, reflecting continued investment in AI products and scaling infrastructure. Kimball said he was not concerned by the sequential fluctuations. “The data center numbers – no concerns whatsoever,” he said. “This was a killer quarter.” Sopko similarly characterized the margin pressure as normal during a major product ramp, but said investors should closely watch customer concentration. “The bigger watchpoint isn’t the margin dip itself, but whether the same two anchor customers continue driving revenue, rack-scale credibility and margin recovery in parallel,” Sopko said. Outlook Points to Continued Expansion AMD expects second-quarter revenue of about $11.2 billion, representing roughly 46% year-over-year growth and a 9% sequential increase. The company also guided to a non-GAAP gross margin of about 56%. Su said customer forecasts for upcoming MI450 accelerators and Helios systems are already exceeding initial expectations, with a growing pipeline of large-scale deployments improving visibility into future growth. Sopko said the larger shift may be AMD’s effort to reposition CPUs as a core layer of AI infrastructure rather than a secondary component behind accelerators. “While the narrative for the last few years was the GPU race, AMD is repricing its CPU opportunity in a way the Street hasn’t fully absorbed yet,” he said.