Jensen Huang: Inference Now Accounts for 60% of AI Compute Cycles
karlfreund · x · 2026-07-24
Speaking recently, AMD CEO Lisa Su cited a perspective from Nvidia's CEO Jensen Huang, noting that 60% of current AI compute demand is now driven by inference rather than training. This figure highlights a major shift in AI infrastructure, where the focus is rapidly moving towards large-scale deployment and real-time response capabilities.
More from Infra
- AMD’s enterprise AI lead takes the stage, and MI430X is said to ship in H1 2027 — ryanshrout · 2026-07-24
- CPU-only inference on a $100 Celeron SBC shows 0.6B models are usable — tre7744 · 2026-07-24
- DeepSeek founder Liang Wenfeng says CUDA’s moat is weakening and Ascend may emerge as an alternative — sudoraohacker · 2026-07-24
- Geekbench 7 gets pitched as a strong CPU benchmark for agentic workloads — BenBajarin · 2026-07-24
- Nebius tops MiniMax M3 serving benchmarks with 197.5 tokens/s and 12.04s latency — Arindam_1729 · 2026-07-24
- AMD unveils ROCm.AI, an AI-driven platform for developers and deployment — ryanshrout · 2026-07-24