(am ex Amazon, and yes Amazon is really really bad, burned out for a year, I could feel the gears in my brain grinding down and doing permanent damage)
For AWS the cycle always seems to be new product comes out, has a great presentation and white paper, and then when the company tries to implement it, they find a bunch of cracks in the service, or hidden cost traps. Or worse, they don't find either until it's too late.
For example on an interview I asked a hiring manager about a time something went not well and how they resolved it, and as soon as she started talking about moving to DynamoDB I started chuckling, because I knew where the story was going. You test out DynamoDB and it works great, then you finally get it to production and find out there's something not performing correctly, and that's when you usually discover you can't actually alter your data structures in production. You either need to completely redo everything, or use the handy money slider to pay AWS more money for reasonable performance. They saw the Oracle DB "throw us money til it works" model and figured out how to automate it.
Similar story with Lambda, where they buried the lede on cold starts (especially for VPCs), and gave a wink and nod to using heartbeat requests to keep them warm. Which works right up until you're actually trying to scale the thing, which was supposed to be the whole point. It appears to mostly be resolved now with the move to Nitro and NIC pools for VPCs, but they let people be guinea pigs for that service for years with promises they knew were false.
I don't have personal stories for GCP, but my coworkers note similar experiences from their previous forays at other companies. Either way, unless you're extremely smart about it, the myriad of services both offer will lock you in, hard, and it's too easy to keep locking yourself in. It would be a herculean effort for some companies to extricate themselves at this point.
That's quite surprising to me - if FAANG's are the standard, Amazon seem like one of the most technically competent (as well as competence and excellence being properly valued there). Yet OP is describing the opposite.
At least Amazon Prime does what it says it will. Amazon works for the consumer. I don't buy from them physically anymore, but they are the only FAANG that still seems to offer some value.
I think they're all going to crash. Oh, and Tesla too XD
I will never work for Amazon given what I have heard about teams there; but what I am worried about is the impact on AWS, which is the main cloud provider at my work, and has been having more than their usual share of catastrophic region-wide problems this year. If this continues, we might have to rethink and move at least some of our systems to GCP to do some stability comparisons.