Poll
I'm in AI at AWS, I think our org is doing really well, but that being said, we don't drive the industry, we are just a big player in it
Amazon enables AI, that's all they need to do to be successful in the space
Seconding this. GPU instances are constantly out of capacity and heavily utilized for what I can only assume is AI training (or someone's very impractical and expensive cloud gaming desktop). More offerings coming out all the time
Amazon makes more money from AI on the cloud than others. They missed out on Alexa crap. And not having GitHub competitor. This last part is a huge miss since the developer ecosystem is there and itād align so well with AWS. But one company canāt do everything and Amazon is doing very well.
AWS Bedrock is coming
An interesting opinion I saw casting doubt on Amazonās AI strategy was this: https://www.semianalysis.com/p/amazons-cloud-crisis-how-aws-will Good article worth reading but here are the main points from ChatGPT (note that while I work at Nvidia, these are my own/articleās opinions and I donāt work with Amazon/AWS blah blah blah): - Amazon is significantly behind in developing AI infrastructure, platforms, and software services compared to competitors like Google and OpenAI. - The company's services suite has limited adoption outside of simple, cheap image nets. - Amazon lacks expertise and leadership in AI, including areas like LLM, DLRM, and Image/Video models. - The company's custom infrastructure is inefficiently set up for AI, and they lack knowledge of forefront hardware innovation. - Amazon's lack of access to enterprise and consumer data hampers its ability to focus on AI. - Data lock-in is less significant with AI, as models can be easily transferred to other clouds for compute-heavy inference tasks. - Amazon's in-house Trainium and Inferentia chips are not competitive for large language models, even with heavy discounting. - Amazon's networking choices, including Nitro and Elastic Fabric Adaptor (EFA), are limiting for AI clusters and hinder scalability, performance, and memory transactions. - The company's refusal to adopt solutions like Nvidia's NVLink and join the CXL consortium further hampers AI capabilities. - Amazon's current AI cloud computing strategy lacks the necessary capabilities and foresight to succeed in the market.
Amazon culture is built for execution not innovation. So, nothing creative comes out of there.
The one tbd part is how it will affect their cloud offering.