chiprook
← AI
AIOctober 20, 2026, 16:45

GPU prices rose 20–25% while token prices fell up to 80%

Smart Inference analyzed 11 million GPU rental price points and 3.3 million token prices from April to October. The median hourly price of H100 rose from $3.25 to $3.89, H200 from $4.00 to $5.02 and B200 from $6.11 to $7.42, while DeepSeek V4 fell 77% and GLM-5 dropped from $1.00 to $0.15 per million input tokens.

GPU prices rose 20–25% while token prices fell up to 80%
#Nvidia#DeepSeek#GLM
Read next
AI

Semi Analysis: n-gram, AgentX, TPU v7 and Vera Rubin in the race for fast tokens

Hardware

d-Matrix halves AI token latency with hybrid GPU-XPU inference at OCP

Business

Scaleway opens GPU zones with AMD MI355X in Stockholm and NVIDIA B300 in Copenhagen

Hardware

Nvidia GPU prices in Japan up to 41% higher than Amazon US