Citi: AI Inference Bottleneck Shifts From Chips to Power and Data Centers, A100 GPUs Up 11% in Six Weeks

According to Citi analysts, AI inference demand remains strong with the bottleneck shifting from advanced chips to older-generation GPUs and power infrastructure, the bank said on June 16 based on a report released June 14. A100 GPU rental prices rose 11% over six weeks, indicating sustained demand beyond cutting-edge hardware. The bank noted that data center capacity constraints now pose a major challenge: one private neocloud operator has signed for 4.9 GW of demand but has over 40 GW planned in its pipeline. Data center site selection increasingly depends on power costs around 9-12 cents per kilowatt-hour, renewable energy availability, and long-term power purchase agreements. Citi said AI infrastructure costs will continue rising as capital expenditures for equipment, power access, and early-stage infrastructure investments increase.
Disclaimer: The information on this page may come from third-party sources and is for reference only. It does not represent the views or opinions of Gate and does not constitute any financial, investment, or legal advice. Virtual asset trading involves high risk. Please do not rely solely on the information on this page when making decisions. For details, see the Disclaimer.
Comment
0/400
No comments