[News] AMD’s AI Chips Power Data Center Revenue: $100B Expected in 5 Years with 60% CAGR

Shortly after reporting record Q3 sales and forecasting even higher Q4 revenue on the AI boom, AMD painted a more bullish, concrete growth picture at its first analyst day in three years, according to Reuters and CNBC. The chipmaker now expects annual data center chip revenue to hit $100 billion within five years, with earnings set to more than triple, Reuters notes.

Reuters, citing CFO Jean Hu, adds that AMD forecasts 35% annual growth overall and roughly 60% in its data center business over the next three to five years. AMD also projects its earnings to surge to $20 per share within the same period.

CEO Lisa Su, according to AMD’s press release, said the data center market could hit $1 trillion by 2030, covering CPUs, networking chips, and AI accelerators. It is worth noting that, as per Reuters, rival NVIDIA CEO Jensen Huang offers an even bolder forecast, projecting the AI infrastructure market could balloon to $3–4 trillion over the same period.

Notably, AMD laid out an upbeat financial outlook while offering a closer look at its AI strategy, including updates on its product roadmap and recent acquisitions. Here are some other key takeaways from the event.

Roadmap: AMD Gears Up for MI400 and MI500 Launches

Among all products, AI GPUs stand out as the main focus. According to Wccftech, the company is maintaining its annual launch schedule, with MI400 and MI500 leading the effort.

The report highlights that AMD will use HBM4 memory in its 2026 Instinct MI400 series, increasing capacity from 288GB HBM3e to 432GB HBM4 and delivering a massive 19.6 TB/s bandwidth—over twice the 8 TB/s of the MI350 series.

Wccftech notes that AMD is positioning its Instinct MI400 GPUs to challenge NVIDIA’s Vera Rubin. The series reportedly debuts with two models: the MI455X for large-scale AI training and inference, and the MI430X for HPC and sovereign AI workloads, both using HBM4, with the MI430X adding hardware FP64 and hybrid CPU+GPU compute. Reuters adds that AMD also plans a full server rack, similar to NVIDIA’s GB200 NVL72.

Following MI400, AMD is set to debut its next-gen Instinct MI500 series AI accelerators in 2027.

Accelerates AI Push Through M&A

As reported by Reuters, AMD CEO Lisa Su highlighted AMD’s wave of AI-driven acquisitions—from ZT Systems, a major server builder, to a cluster of software startups—at the event, saying the company has effectively turned mergers and acquisitions into a core growth engine.

CRN notes that AMD spent $36 million on smaller acquisitions in 2025 to strengthen its data center business, which includes silicon photonics startup Enosemi to support next-gen AI photonics and co-packaged optics solutions, compiler startup Brium to deliver highly optimized AI solutions, and the team from AI chip startup Untether AI to boost AI compiler, kernel development, and SoC design capabilities.

According to Reuters, AMD’s latest move came Monday with the acquisition of MK1. The goal, said Chief Strategy Officer Mat Hein in an interview with Reuters, is to secure both the software stack and the talent needed to scale AMD’s AI ambitions.

Read more

• [News] AMD Flags Intel–NVIDIA Deal Risks; Key Team Blue AI Executive Reportedly Moves to Rival
• [News] AMD Forecasts Strong Q4 After Record Q3 Sales, Flags China and 2026 AI Chip Ramp-up

(Photo credit: AMD)