What are your biggest challenges with cloud computing for AI/ML tasks?

Hi everyone, I’m an ML researcher exploring the challenges AI/ML practitioners face with current cloud computing platforms, and I’d love to hear your experiences and views. Here are a few guiding questions: 1. Do costs limit your ability to scale or run experiments? 2. Have you faced challenges accessing specific hardware like GPUs or CPUs? 3. Are you concerned about data privacy or vendor lock-in with centralized providers? 4. What features or solutions would make your cloud computing experience better? Also, I’m curious about your thoughts on distributed computing platforms (e.g., BOINC, Golem). These platforms crowdsource computing power from individuals’ devices instead of relying on centralized cloud servers. Have you ever considered or used such platforms? Do you see them as viable alternatives to traditional cloud providers? Feel free to share your frustrations, wishlists, or alternatives you’ve tried. Your insights will really help me understand this space better! Thanks in advance for sharing!