Thunder Compute (YC S24) — Self-host AI/ML with the world's cheapest GPU cloud

Hi, guys! This is Carl, co-founder of Thunder Compute. After going through YC this summer and two years of building we're thrilled to show off our product. ⚡ Thunder Compute is the best way for scrappy teams to get cloud GPUs for AI/ML. Our mission is to give everyone access to the top-of-the-line hardware used by frontier labs like OpenAI. ✨ Why Thunder Compute? 1. Thunder Compute has the world's cheapest GPUs. By over 40%. For example, the lowest-cost budget providers today sell NVIDIA A100s for ~$700/month. We sell the same GPU for $415/month. 2. Thunder Compute has a dead-simple user experience. With our templates, it only takes a few clicks to start up an instance pre-installed with your AI tools. We provide simple GUI and CLI options for users to create instances without the complexity of AWS. 3. We have active support and a great community. If anything goes wrong, you can directly message our team on Discord (https://discord.gg/nwuETS9jJK). This may seem trivial, but the same service on AWS costs thousands of dollars per month. I promise it will come in handy when you're troubleshooting the latest Flux-v158.23 model and the docs are in German. 💸 How are you so cheap? The price seems too good to be true. We invented a new type of GPU virtualization which eliminates idle time, using each GPU 6x more efficiently. This lets us undercut all of our competitors while sourcing GPUs from data centers with top-tier reliability (AWS and GCP). 🔮 What's Next? Today, Thunder Compute is optimized for development workflows. Over time we plan to build out tooling for production inference and large-scale training. 🚀 Get Started Today! If you are an indie dev or work with a small startup, Thunder Compute is the best cloud platform for your next project. Try out Thunder today with $20/month of free credit.