If you need to spawn 10-100+ instances for heavy computation tasks - then AWS and alikes are the only choice.<p>@$2k/mo+/instance - for smaller scale research it is way cheaper to buy high performance GPU card(s) yourself and use them without unpredictable recurring cost issues. Computers are cheap.<p><a href="https://www.reddit.com/r/MachineLearning/comments/45vepw/choose_a_good_cuda_card_looking_for_your_advice/" rel="nofollow">https://www.reddit.com/r/MachineLearning/comments/45vepw/cho...</a><p><a href="http://www.videocardbenchmark.net/high_end_gpus.html" rel="nofollow">http://www.videocardbenchmark.net/high_end_gpus.html</a>