Frustrated by GPU costs, so I built my own cloud
Frustrated by GPU costs, so I built my own cloud
Hey HN, It was back in April 2023 when I was trying to build a voice agent with the Orpheus/Sesame CSM models it was my first time building something in AI and I absolutely hated the friction the platforms had to offer, mental prices, vast.ai instances not even connecting half of the time, and nothing in the platforms that said "you're welcome here" Out of frustration, I spoke to some providers, sorted out a fine list, and built Runcrate. We made the onboarding, UX, and overall experience as frictionless as possible. The billing is just like you'd pay OpenAI or Anthropic, no contracts, no reservations, no quotas, no more contacting support "Hey I need more GPUs pls" and then get rejected. Whatever is in the stock, its available for everyone. Oh and yeah, there's also a cool button that comes with all instances that spins up a vscode on the browser for you with a custom jupyter notebook extension installed. Frictionless, that's all I wanted in this space. I hope it'll genuinely help people looking to get started with AI & startups looking to cut their bills down. Would love your feedback pls Jish
Share cardActual performance
Launch Intel predictions
Analyze your own launch →Correct prediction on native model
Similar products
GPU cluster energy costs made simple
The Poor Man's Guide to Cloud GPU Selection
Some hacks to optimize cloud costs
Optimize cloud costs
Optimize your cloud, minimize your costs
Gource visualizations rendered without a GPU
Lambda Echelon GPU Cluster
GPU PaaS
Slurmq – GPU quota enforcement for Slurm
GPU Accelerated PDAL