Lightning AI

GPU compute costs repriced on the 'Compute costs' table: T4 spot rose $0.52->$0.53/GPU/hr, L4 spot $0.62-$0.75->$0.68-$0.81, RTXP 6000 spot $1.46-$1.75->$2.61-$3.13 (~79% up). H100 on-demand dropped $4.68->$4.50 and gained a new 'New'-badged spot tier ($3.82-$4.58) that was previously unavailable, alongside a new banner promoting guaranteed on-demand H100 access.

H100 GPU compute rate increased from $3.50/GPU/hr to $4.68/GPU/hr (+34%); the 'Beta' qualifier was also dropped from the GPU's name.

Interruptible (spot) GPU prices increased for T4 ($0.62→$0.63), L4 ($0.58–$0.69→$0.62–$0.75), and A100 80GB ($3.06–$3.67→$3.32–$3.99) per GPU/hr.

A100 (40 GB) price increased from $1.29 to $2.19/GPU/hr (+70%); H100 (80 GB) price increased from $1.99 to $3.50/GPU/hr (+76%). Significant compute cost increases across high-end GPU tiers. T4 GPU interruptible (spot) price range decreased from $0.53–$0.64 to $0.52–$0.63 per GPU/hr. Small reduction in spot pricing floor and ceiling.

T4 GPU interruptible (spot) pricing increased from $0.52-$0.62 to $0.53-$0.64 per GPU/hr (low end +1.9%, high end +3.2%). TPU compute option added as a new tab in the GPU selector alongside CPU, 1 GPU, 2 GPU, 4 GPU, 8 GPU, and Clusters.

Multiple GPU pricing changes across T4, L4, L40S, RTXP 6000, and H200 instances. H200 cost/GPU/hr increased 86% from $3.50 to $6.53. L40S regular price cut from $2.89 to $2.14; L4 and RTXP 6000 interruptible rates raised.

L40S interruptible GPU price increased from $1.79/$2.15 to $1.80/$2.16 per GPU/hr (both standard and spot interruptible rates raised by $0.01).