AMD (AMD) AMD Advancing AI 2026 Keynote summary

Industry and Market Trends

• AI demand is accelerating rapidly, with a 158x increase in tokens consumed monthly over the past two years and training compute (FLOPS) increasing 5x annually since 2020, driving a projected $2T total compute TAM by 2030.

• Inference workloads have surpassed training as the primary AI workload, with agentic AI expected to account for 60% of global AI compute in 2026 and further increasing compute needs through orchestration, more agents, and tool usage.

• The AI accelerator market is projected to reach $1.4 trillion by 2030, with data center AI accelerator and CPU TAMs growing at >45% and >50% CAGR respectively, and server CPU market forecasted to grow over 50% to $200B by 2030.

• AI workloads are becoming distributed across cloud, data center, client, and edge environments, with edge and personal devices becoming critical for real-time, distributed intelligence.

• AI is transforming diverse industries including healthcare, manufacturing, finance, education, science, software, agriculture, and media.

Product and Technology Announcements

• AMD Helios, a high-performance rackscale AI infrastructure, was launched, integrating Instinct GPUs, EPYC CPUs, and advanced networking, delivering up to 2.9 exaFLOPS FP4 compute, 31TB HBM capacity, and leadership in compute, memory, and networking.

• Helios delivers 15% more compute, 50% more memory capacity and bandwidth, and 50% more scale-out bandwidth than competitors, with up to 30% more tokens per dollar and shipments starting end of Q3 and ramping in Q4.

• The Instinct MI455X GPU offers up to 34x higher inference throughput and up to 18x lower token cost versus previous generations, while MI350P supports up to 260B parameter models and delivers 4x tokens per second per dollar versus competition.

• Venice, the 6th-gen EPYC CPU family, features up to 256 cores, 512 threads, 1.6TB/s memory bandwidth, and up to 1.8x performance, optimized for agentic, GPU-host, and general-purpose workloads.

• ROCm AI, an agentic AI platform, introduces AI-assisted GPU programming, enabling developers to optimize workloads and automate kernel generation, delivering up to 3.3x speedup over previous versions and supporting 3M+ models on Hugging Face.

Strategic Partnerships and Ecosystem Expansion

• Deep collaborations with Anthropic, OpenAI, Meta, Cerebras, AT&T, Cisco, Hugging Face, Microsoft, Oracle, HPE, and Lenovo, co-designing hardware/software and deploying AMD infrastructure at scale.

• OpenAI and Meta are deploying Helios and Venice at massive scale, with joint engineering optimizing software stacks and hardware for next-gen AI workloads.

• AT&T launched OTel 2.0, an open-source telco AI model trained on AMD, reporting significant cost savings and workflow transformation.

• Cisco and AMD deliver unified management and security for enterprise-scale personal AI deployments, with Halo devices and cloud control integration.

• Hugging Face partnership brings optimized open models and toolkits to Ryzen AI Halo, accelerating developer productivity and enterprise adoption.