Blended GPU compute costs remained stable 0.1% this week across major cloud providers, led by ALVEO_U30's 31.2% decline.
Top Movers
| GPU | Blended Price | WoW Change | Class |
|---|---|---|---|
| ALVEO_U30 | $0.03/hr | ▼ 31.2% | General |
| TRAINIUM2 | $2.37/hr | ▲ 15.6% | Training |
| GAUDI | $0.78/hr | ▲ 14.1% | General |
| TRAINIUM | $0.98/hr | ▲ 13.3% | Training |
| RTX PRO 6000 | $0.55/hr | ▲ 12.1% | General |
| MI25 | $0.22/hr | ▼ 8.8% | General |
| INFERENTIA2 | $0.08/hr | ▼ 8.4% | Inference |
Blended pricing = average of spot, on-demand, and 1-year reserved rates across major cloud providers.
Training vs Inference
Training-class GPU pricing rose this week (avg $4.21/hr, +1.6% WoW), while inference-class pricing fell (avg $0.51/hr, -1.3% WoW).
The training-to-inference price ratio stands at 8.2x — widening compared to last week. The elevated spread suggests strong demand for training compute relative to inference, consistent with ongoing large model training activity.
Regional Spotlight: South America
South America trades at a 34% premium to global averages this week, with 19 GPU types available across the region. The most expensive GPUs in the region are GB200 ($23.02/hr), H200 ($14.82/hr), H100 ($7.67/hr). The 34% regional premium reflects emerging infrastructure and limited provider competition.
For detailed pricing data across all South America sub-regions, see the full regional profile.
Implications
For cloud buyers: Europe continues to offer the lowest average GPU pricing ($1.99/hr blended average). For workloads with regional flexibility, the gap between Europe and Middle East is $2.06/hr — a 104% premium. Compare regional pricing →
For semiconductor analysts: GPU pricing trends remain broadly stable this week. H100 (-0.9% WoW) and MI300X (-0.6% WoW) are tracking within normal ranges. Blackwell (B200) blended pricing at $7.19/hr (+0.7% WoW) provides an early read on next-generation adoption curves. View all GPU profiles →
For GPU investors: Stable pricing supports predictable returns for existing deployments. Model scenarios with the GPU ROI Calculator →