We are in the process of building a public cloud on our own hardware, with a key focus on GPU-enabled services for our customers. Despite the importance of AI/ML, our Fly GPU Machines product hasn’t gained the traction we hoped for. Security concerns and technical challenges with Nvidia drivers have been major hurdles. Developers seem more interested in LLMs (Large Language Models) than GPUs for their applications. Our GPU Fly Machines have not been as successful as we anticipated, but we have learned valuable lessons about making bets and acquiring assets for our startup. We are scaling back our GPU ambitions and focusing on our core product offerings.
https://fly.io/blog/wrong-about-gpu/