Case Study: Blend cuts AI inference costs by 85% with SaladCloud

Blend cuts AI inference cost by 85% on Salad while running 3X more scale

Blend, an AI copilot for e-commerce that creates professional product photos, faced prohibitively high AI inference costs and severe GPU scalability challenges on major cloud providers and local vendors. Their rapid growth and reliance on powerful GPUs for diffusion models led to unreliable performance and resource constraints that threatened their operations.

The company switched its workloads to SaladCloud, leveraging their distributed network of consumer GPUs. With engineering support from SaladCloud, Blend improved system reliability and achieved a lower P95 latency. The results were transformative: Blend now runs at 3X its previous scale for an 85% reduction in cost compared to major hyperscalers, while also improving performance.

Blend

Jamsheed Kamardeen

Chief Technology Officer